From a1e2f63e64875a29e8c01a7ae17f5744680167a5 Mon Sep 17 00:00:00 2001 From: Sebastian Thiel Date: Mon, 15 Nov 2010 11:37:14 +0100 Subject: submodule: Fleshed out interface, and a partial test which is not yet usable. It showed that the ConfigParser needs some work. If the root is set, it also needs to refer to the root_commit instead of to the root-tree, as it will have to decide whether it works on the working tree's version of the .gitmodules file or the one in the repository --- lib/git/config.py | 24 ++++++++- lib/git/ext/gitdb | 2 +- lib/git/objects/submodule.py | 115 +++++++++++++++++++++++++++++++++++++++++-- lib/git/remote.py | 30 ++--------- 4 files changed, 140 insertions(+), 31 deletions(-) (limited to 'lib/git') diff --git a/lib/git/config.py b/lib/git/config.py index 09bad0b6..e919838b 100644 --- a/lib/git/config.py +++ b/lib/git/config.py @@ -15,7 +15,7 @@ import cStringIO from git.odict import OrderedDict from git.util import LockFile -__all__ = ('GitConfigParser', ) +__all__ = ('GitConfigParser', 'SectionConstraint') class MetaParserBuilder(type): """Utlity class wrapping base-class methods into decorators that assure read-only properties""" @@ -63,7 +63,29 @@ def set_dirty_and_flush_changes(non_const_func): flush_changes.__name__ = non_const_func.__name__ return flush_changes + +class SectionConstraint(object): + """Constrains a ConfigParser to only option commands which are constrained to + always use the section we have been initialized with. + + It supports all ConfigParser methods that operate on an option""" + __slots__ = ("_config", "_section_name") + _valid_attrs_ = ("get_value", "set_value", "get", "set", "getint", "getfloat", "getboolean", "has_option") + def __init__(self, config, section): + self._config = config + self._section_name = section + + def __getattr__(self, attr): + if attr in self._valid_attrs_: + return lambda *args, **kwargs: self._call_config(attr, *args, **kwargs) + return super(SectionConstraint,self).__getattribute__(attr) + + def _call_config(self, method, *args, **kwargs): + """Call the configuration at the given method which must take a section name + as first argument""" + return getattr(self._config, method)(self._section_name, *args, **kwargs) + class GitConfigParser(cp.RawConfigParser, object): """Implements specifics required to read git style configuration files. diff --git a/lib/git/ext/gitdb b/lib/git/ext/gitdb index 78665b13..2ddc5bad 160000 --- a/lib/git/ext/gitdb +++ b/lib/git/ext/gitdb @@ -1 +1 @@ -Subproject commit 78665b13ff4125f4ce3e5311d040c027bdc92a9a +Subproject commit 2ddc5bad224d8f545ef3bb2ab3df98dfe063c5b6 diff --git a/lib/git/objects/submodule.py b/lib/git/objects/submodule.py index 1f571a48..b0fd0e35 100644 --- a/lib/git/objects/submodule.py +++ b/lib/git/objects/submodule.py @@ -1,4 +1,8 @@ import base +from cStringIO import StringIO +from git.config import GitConfigParser +from git.util import join_path_native +from git.exc import InvalidGitRepositoryError, NoSuchPathError __all__ = ("Submodule", ) @@ -7,10 +11,115 @@ class Submodule(base.IndexObject): represents a commit in the submodule's repository which is to be checked out at the path of this instance. The submodule type does not have a string type associated with it, as it exists - solely as a marker in the tree and index""" + solely as a marker in the tree and index. + + All methods work in bare and non-bare repositories.""" + + kModulesFile = '.gitmodules' # this is a bogus type for base class compatability type = 'submodule' - # TODO: Add functions to retrieve a repo for the submodule, to allow - # its initiailization and handling + __slots__ = ('_root_tree', '_url', '_ref') + + def _set_cache_(self, attr): + if attr == 'size': + raise ValueError("Submodules do not have a size as they do not refer to anything in this repository") + elif attr == '_root_tree': + # set a default value, which is the root tree of the current head + self._root_tree = self.repo.tree() + elif attr in ('path', '_url', '_ref'): + reader = self.config_reader() + # default submodule values + self._path = reader.get_value('path') + self._url = reader.get_value('url') + # git-python extension values - optional + self._ref = reader.get_value('ref', 'master') + else: + super(Submodule, self)._set_cache_(attr) + # END handle attribute name + + def _fp_config(self): + """:return: Configuration file as StringIO - we only access it through the respective blob's data""" + return StringIO(self._root_tree[self.kModulesFile].datastream.read()) + + def _config_parser(self, read_only): + """:return: Config Parser constrained to our submodule in read or write mode""" + parser = GitConfigParser(self._fp_config(), read_only = read_only) + return SectionConstraint(parser, 'submodule "%s"' % self.path) + + #{ Edit Interface + + @classmethod + def add(cls, repo, path, url, skip_init=False): + """Add a new submodule to the given repository. This will alter the index + as well as the .gitmodules file, but will not create a new commit. + :param repo: Repository instance which should receive the submodule + :param path: repository-relative path at which the submodule should be located + It will be created as required during the repository initialization. + :param url: git-clone compatible URL, see git-clone reference for more information + :param skip_init: if True, the new repository will not be cloned to its location. + :return: The newly created submodule instance""" + + def set_root_tree(self, root_tree): + """Set this instance to use the given tree which is supposed to contain the + .gitmodules blob. + :param root_tree: Tree'ish reference pointing at the root_tree + :raise ValueError: if the root_tree didn't contain the .gitmodules blob.""" + tree = self.repo.tree(root_tree) + if self.kModulesFile not in tree: + raise ValueError("Tree %s did not contain the %s file" % (root_tree, self.kModulesFile)) + # END handle exceptions + self._root_tree = tree + + # clear the possibly changing values + del(self.path) + del(self._ref) + del(self._url) + + def config_writer(self): + """:return: a config writer instance allowing you to read and write the data + belonging to this submodule into the .gitmodules file.""" + return self._config_parser(read_only=False) + + #} END edit interface + + #{ Query Interface + + def module(self): + """:return: Repo instance initialized from the repository at our submodule path + :raise InvalidGitRepositoryError: if a repository was not available""" + if self.repo.bare: + raise InvalidGitRepositoryError("Cannot retrieve module repository in bare parent repositories") + # END handle bare mode + + repo_path = join_path_native(self.repo.working_tree_dir, self.path) + try: + return Repo(repo_path) + except (InvalidGitRepositoryError, NoSuchPathError): + raise InvalidGitRepositoryError("No valid repository at %s" % self.path) + # END handle exceptions + + def ref(self): + """:return: The reference's name that we are to checkout""" + return self._ref + + def url(self): + """:return: The url to the repository which our module-repository refers to""" + return self._url + + def root_tree(self): + """:return: Tree instance referring to the tree which contains the .gitmodules file + we are to use + :note: will always point to the current head's root tree if it was not set explicitly""" + return self._root_tree + + def config_reader(self): + """:return: ConfigReader instance which allows you to qurey the configuration values + of this submodule, as provided by the .gitmodules file + :note: The config reader will actually read the data directly from the repository + and thus does not need nor care about your working tree. + :note: Should be cached by the caller and only kept as long as needed""" + return self._config_parser.read_only(read_only=True) + + #} END query interface diff --git a/lib/git/remote.py b/lib/git/remote.py index 52dd787d..135e37d7 100644 --- a/lib/git/remote.py +++ b/lib/git/remote.py @@ -7,7 +7,8 @@ from exc import GitCommandError from objects import Commit -from ConfigParser import NoOptionError +from ConfigParser import NoOptionError +from config import SectionConstraint from git.util import ( LazyMixin, @@ -30,29 +31,6 @@ import os __all__ = ('RemoteProgress', 'PushInfo', 'FetchInfo', 'Remote') -class _SectionConstraint(object): - """Constrains a ConfigParser to only option commands which are constrained to - always use the section we have been initialized with. - - It supports all ConfigParser methods that operate on an option""" - __slots__ = ("_config", "_section_name") - _valid_attrs_ = ("get_value", "set_value", "get", "set", "getint", "getfloat", "getboolean", "has_option") - - def __init__(self, config, section): - self._config = config - self._section_name = section - - def __getattr__(self, attr): - if attr in self._valid_attrs_: - return lambda *args, **kwargs: self._call_config(attr, *args, **kwargs) - return super(_SectionConstraint,self).__getattribute__(attr) - - def _call_config(self, method, *args, **kwargs): - """Call the configuration at the given method which must take a section name - as first argument""" - return getattr(self._config, method)(self._section_name, *args, **kwargs) - - class RemoteProgress(object): """ Handler providing an interface to parse progress information emitted by git-push @@ -449,7 +427,7 @@ class Remote(LazyMixin, Iterable): def _set_cache_(self, attr): if attr == "_config_reader": - self._config_reader = _SectionConstraint(self.repo.config_reader(), self._config_section_name()) + self._config_reader = SectionConstraint(self.repo.config_reader(), self._config_section_name()) else: super(Remote, self)._set_cache_(attr) @@ -735,4 +713,4 @@ class Remote(LazyMixin, Iterable): # clear our cache to assure we re-read the possibly changed configuration del(self._config_reader) - return _SectionConstraint(writer, self._config_section_name()) + return SectionConstraint(writer, self._config_section_name()) -- cgit v1.2.1 From 4d36f8ff4d1274a8815e932285ad6dbd6b2888af Mon Sep 17 00:00:00 2001 From: Sebastian Thiel Date: Mon, 15 Nov 2010 12:13:59 +0100 Subject: Improved GitConfigurationParser to better deal with streams and the corresponding locks. Submodule class now operates on parent_commits, the configuration is either streamed from the repository or written directly into a blob ( or file ) dependending on whether we have a working tree checkout or not which matches our parent_commit --- lib/git/config.py | 12 +++++--- lib/git/objects/submodule.py | 70 +++++++++++++++++++++++++++++--------------- 2 files changed, 54 insertions(+), 28 deletions(-) (limited to 'lib/git') diff --git a/lib/git/config.py b/lib/git/config.py index e919838b..8541dc0e 100644 --- a/lib/git/config.py +++ b/lib/git/config.py @@ -271,9 +271,9 @@ class GitConfigParser(cp.RawConfigParser, object): if not hasattr(file_object, "seek"): try: fp = open(file_object) + close_fp = True except IOError,e: continue - close_fp = True # END fp handling try: @@ -308,17 +308,21 @@ class GitConfigParser(cp.RawConfigParser, object): :raise IOError: if this is a read-only writer instance or if we could not obtain a file lock""" self._assure_writable("write") - self._lock._obtain_lock() - fp = self._file_or_files close_fp = False + # we have a physical file on disk, so get a lock + if isinstance(fp, (basestring, file)): + self._lock._obtain_lock() + # END get lock for physical files + if not hasattr(fp, "seek"): fp = open(self._file_or_files, "w") close_fp = True else: fp.seek(0) + # END handle stream or file # WRITE DATA try: @@ -390,7 +394,7 @@ class GitConfigParser(cp.RawConfigParser, object): return valuestr @needs_values - @set_dirty_and_flush_changes + @set_dirty_and_flush_changes def set_value(self, section, option, value): """Sets the given option in section to the given value. It will create the section if required, and will not throw as opposed to the default diff --git a/lib/git/objects/submodule.py b/lib/git/objects/submodule.py index b0fd0e35..b9bcfc07 100644 --- a/lib/git/objects/submodule.py +++ b/lib/git/objects/submodule.py @@ -6,6 +6,13 @@ from git.exc import InvalidGitRepositoryError, NoSuchPathError __all__ = ("Submodule", ) +class SubmoduleConfigParser(GitConfigParser): + """Catches calls to _write, and updates the .gitmodules blob in the index + with the new data, if we have written into a stream. Otherwise it will + add the local file to the index to make it correspond with the working tree.""" + _mutating_methods_ = tuple() + + class Submodule(base.IndexObject): """Implements access to a git submodule. They are special in that their sha represents a commit in the submodule's repository which is to be checked out @@ -20,14 +27,14 @@ class Submodule(base.IndexObject): # this is a bogus type for base class compatability type = 'submodule' - __slots__ = ('_root_tree', '_url', '_ref') + __slots__ = ('_parent_commit', '_url', '_ref') def _set_cache_(self, attr): if attr == 'size': raise ValueError("Submodules do not have a size as they do not refer to anything in this repository") - elif attr == '_root_tree': + elif attr == '_parent_commit': # set a default value, which is the root tree of the current head - self._root_tree = self.repo.tree() + self._parent_commit = self.repo.commit() elif attr in ('path', '_url', '_ref'): reader = self.config_reader() # default submodule values @@ -39,13 +46,26 @@ class Submodule(base.IndexObject): super(Submodule, self)._set_cache_(attr) # END handle attribute name - def _fp_config(self): + def _sio_modules(self): """:return: Configuration file as StringIO - we only access it through the respective blob's data""" - return StringIO(self._root_tree[self.kModulesFile].datastream.read()) + sio = StringIO(self._parent_commit.tree[self.kModulesFile].datastream.read()) + sio.name = self.kModulesFile + return sio def _config_parser(self, read_only): """:return: Config Parser constrained to our submodule in read or write mode""" - parser = GitConfigParser(self._fp_config(), read_only = read_only) + parent_matches_head = self.repo.head.commit == self._parent_commit + if not self.repo.bare and parent_matches_head: + fp_module = self.kModulesFile + else: + fp_module = self._sio_modules() + # END handle non-bare working tree + + if not read_only and not parent_matches_head: + raise ValueError("Cannot write blobs of 'historical' submodule configurations") + # END handle writes of historical submodules + + parser = GitConfigParser(fp_module, read_only = read_only) return SectionConstraint(parser, 'submodule "%s"' % self.path) #{ Edit Interface @@ -61,21 +81,24 @@ class Submodule(base.IndexObject): :param skip_init: if True, the new repository will not be cloned to its location. :return: The newly created submodule instance""" - def set_root_tree(self, root_tree): - """Set this instance to use the given tree which is supposed to contain the - .gitmodules blob. - :param root_tree: Tree'ish reference pointing at the root_tree - :raise ValueError: if the root_tree didn't contain the .gitmodules blob.""" - tree = self.repo.tree(root_tree) - if self.kModulesFile not in tree: - raise ValueError("Tree %s did not contain the %s file" % (root_tree, self.kModulesFile)) + def set_parent_commit(self, commit): + """Set this instance to use the given commit whose tree is supposed to + contain the .gitmodules blob. + :param commit: Commit'ish reference pointing at the root_tree + :raise ValueError: if the commit's tree didn't contain the .gitmodules blob.""" + pcommit = self.repo.commit(commit) + if self.kModulesFile not in pcommit.tree: + raise ValueError("Tree of commit %s did not contain the %s file" % (commit, self.kModulesFile)) # END handle exceptions - self._root_tree = tree + self._parent_commit = pcommit - # clear the possibly changing values - del(self.path) - del(self._ref) - del(self._url) + # clear the possibly changed values + for name in ('path', '_ref', '_url'): + try: + delattr(self, name) + except AttributeError: + pass + # END for each name to delete def config_writer(self): """:return: a config writer instance allowing you to read and write the data @@ -108,11 +131,10 @@ class Submodule(base.IndexObject): """:return: The url to the repository which our module-repository refers to""" return self._url - def root_tree(self): - """:return: Tree instance referring to the tree which contains the .gitmodules file - we are to use - :note: will always point to the current head's root tree if it was not set explicitly""" - return self._root_tree + def parent_commit(self): + """:return: Commit instance with the tree containing the .gitmodules file + :note: will always point to the current head's commit if it was not set explicitly""" + return self._parent_commit def config_reader(self): """:return: ConfigReader instance which allows you to qurey the configuration values -- cgit v1.2.1 From 00ce31ad308ff4c7ef874d2fa64374f47980c85c Mon Sep 17 00:00:00 2001 From: Sebastian Thiel Date: Mon, 15 Nov 2010 16:53:12 +0100 Subject: Objects: Constructor now manually checks and sets the input arguments to the local cache - previously a procedural approach was used, which was less code, but slower too. Especially in case of CommitObjects unrolling the loop manually makes a difference. Submodule: Implemented query methods and did a bit of testing. More is to come, but the test works for now. As special addition, the submodule implementation uses the section name as submodule ID even though it seems to be just the path. This allows to make renames easier --- lib/git/objects/base.py | 16 +--- lib/git/objects/commit.py | 25 ++++- lib/git/objects/submodule.py | 222 ++++++++++++++++++++++++++++++++++++------- lib/git/objects/tag.py | 13 ++- 4 files changed, 227 insertions(+), 49 deletions(-) (limited to 'lib/git') diff --git a/lib/git/objects/base.py b/lib/git/objects/base.py index 41862ac2..82c2589c 100644 --- a/lib/git/objects/base.py +++ b/lib/git/objects/base.py @@ -62,17 +62,6 @@ class Object(LazyMixin): inst.size = oinfo.size return inst - def _set_self_from_args_(self, args_dict): - """Initialize attributes on self from the given dict that was retrieved - from locals() in the calling method. - - Will only set an attribute on self if the corresponding value in args_dict - is not None""" - for attr, val in args_dict.items(): - if attr != "self" and val is not None: - setattr( self, attr, val ) - # END set all non-None attributes - def _set_cache_(self, attr): """Retrieve object information""" if attr == "size": @@ -140,7 +129,10 @@ class IndexObject(Object): Path may not be set of the index object has been created directly as it cannot be retrieved without knowing the parent tree.""" super(IndexObject, self).__init__(repo, binsha) - self._set_self_from_args_(locals()) + if mode is not None: + self.mode = mode + if path is not None: + self.path = path def __hash__(self): """:return: diff --git a/lib/git/objects/commit.py b/lib/git/objects/commit.py index 58c82da2..ae22fb76 100644 --- a/lib/git/objects/commit.py +++ b/lib/git/objects/commit.py @@ -108,7 +108,26 @@ class Commit(base.Object, Iterable, Diffable, Traversable, Serializable): super(Commit,self).__init__(repo, binsha) if tree is not None: assert isinstance(tree, Tree), "Tree needs to be a Tree instance, was %s" % type(tree) - self._set_self_from_args_(locals()) + if tree is not None: + self.tree = tree + if author is not None: + self.author = author + if authored_date is not None: + self.authored_date = authored_date + if author_tz_offset is not None: + self.author_tz_offset = author_tz_offset + if committer is not None: + self.committer = committer + if committed_date is not None: + self.committed_date = committed_date + if committer_tz_offset is not None: + self.committer_tz_offset = committer_tz_offset + if message is not None: + self.message = message + if parents is not None: + self.parents = parents + if encoding is not None: + self.encoding = encoding @classmethod def _get_intermediate_items(cls, commit): @@ -434,7 +453,7 @@ class Commit(base.Object, Iterable, Diffable, Traversable, Serializable): try: self.author.name = self.author.name.decode(self.encoding) except UnicodeDecodeError: - print >> sys.stderr, "Failed to decode author name: %s" % self.author.name + print >> sys.stderr, "Failed to decode author name '%s' using encoding %s" % (self.author.name, self.encoding) # END handle author's encoding # a stream from our data simply gives us the plain message @@ -443,7 +462,7 @@ class Commit(base.Object, Iterable, Diffable, Traversable, Serializable): try: self.message = self.message.decode(self.encoding) except UnicodeDecodeError: - print >> sys.stderr, "Failed to decode message: %s" % self.message + print >> sys.stderr, "Failed to decode message '%s' using encoding %s" % (self.message, self.encoding) # END exception handling return self diff --git a/lib/git/objects/submodule.py b/lib/git/objects/submodule.py index b9bcfc07..1aa0cfb5 100644 --- a/lib/git/objects/submodule.py +++ b/lib/git/objects/submodule.py @@ -1,11 +1,29 @@ import base -from cStringIO import StringIO -from git.config import GitConfigParser +from StringIO import StringIO # need a dict to set bloody .name field +from git.util import Iterable +from git.config import GitConfigParser, SectionConstraint from git.util import join_path_native from git.exc import InvalidGitRepositoryError, NoSuchPathError +import os + __all__ = ("Submodule", ) +#{ Utilities + +def sm_section(path): + """:return: section title used in .gitmodules configuration file""" + return 'submodule "%s"' % path + +def sm_name(section): + """:return: name of the submodule as parsed from the section name""" + section = section.strip() + return section[11:-1] +#} END utilities + + +#{ Classes + class SubmoduleConfigParser(GitConfigParser): """Catches calls to _write, and updates the .gitmodules blob in the index with the new data, if we have written into a stream. Otherwise it will @@ -13,7 +31,7 @@ class SubmoduleConfigParser(GitConfigParser): _mutating_methods_ = tuple() -class Submodule(base.IndexObject): +class Submodule(base.IndexObject, Iterable): """Implements access to a git submodule. They are special in that their sha represents a commit in the submodule's repository which is to be checked out at the path of this instance. @@ -22,12 +40,32 @@ class Submodule(base.IndexObject): All methods work in bare and non-bare repositories.""" - kModulesFile = '.gitmodules' + _id_attribute_ = "path" + k_modules_file = '.gitmodules' + k_ref_option = 'ref' + k_ref_default = 'master' # this is a bogus type for base class compatability type = 'submodule' - __slots__ = ('_parent_commit', '_url', '_ref') + __slots__ = ('_parent_commit', '_url', '_ref', '_name') + + def __init__(self, repo, binsha, mode=None, path=None, name = None, parent_commit=None, url=None, ref=None): + """Initialize this instance with its attributes. We only document the ones + that differ from ``IndexObject`` + :param binsha: binary sha referring to a commit in the remote repository, see url parameter + :param parent_commit: see set_parent_commit() + :param url: The url to the remote repository which is the submodule + :param ref: Reference to checkout when cloning the remote repository""" + super(Submodule, self).__init__(repo, binsha, mode, path) + if parent_commit is not None: + self._parent_commit = parent_commit + if url is not None: + self._url = url + if ref is not None: + self._ref = ref + if name is not None: + self._name = name def _set_cache_(self, attr): if attr == 'size': @@ -38,35 +76,63 @@ class Submodule(base.IndexObject): elif attr in ('path', '_url', '_ref'): reader = self.config_reader() # default submodule values - self._path = reader.get_value('path') + self.path = reader.get_value('path') self._url = reader.get_value('url') # git-python extension values - optional - self._ref = reader.get_value('ref', 'master') + self._ref = reader.get_value(self.k_ref_option, self.k_ref_default) + elif attr == '_name': + raise AttributeError("Cannot retrieve the name of a submodule if it was not set initially") else: super(Submodule, self)._set_cache_(attr) # END handle attribute name - - def _sio_modules(self): - """:return: Configuration file as StringIO - we only access it through the respective blob's data""" - sio = StringIO(self._parent_commit.tree[self.kModulesFile].datastream.read()) - sio.name = self.kModulesFile - return sio - def _config_parser(self, read_only): - """:return: Config Parser constrained to our submodule in read or write mode""" - parent_matches_head = self.repo.head.commit == self._parent_commit - if not self.repo.bare and parent_matches_head: - fp_module = self.kModulesFile + def __eq__(self, other): + """Compare with another submodule""" + return self.path == other.path and self.url == other.url and super(Submodule, self).__eq__(other) + + def __ne__(self, other): + """Compare with another submodule for inequality""" + return not (self == other) + + @classmethod + def _config_parser(cls, repo, parent_commit, read_only): + """:return: Config Parser constrained to our submodule in read or write mode + :raise IOError: If the .gitmodules file cannot be found, either locally or in the repository + at the given parent commit. Otherwise the exception would be delayed until the first + access of the config parser""" + parent_matches_head = repo.head.commit == parent_commit + if not repo.bare and parent_matches_head: + fp_module = cls.k_modules_file + fp_module_path = os.path.join(repo.working_tree_dir, fp_module) + if not os.path.isfile(fp_module_path): + raise IOError("%s file was not accessible" % fp_module_path) + # END handle existance else: - fp_module = self._sio_modules() + try: + fp_module = cls._sio_modules(parent_commit) + except KeyError: + raise IOError("Could not find %s file in the tree of parent commit %s" % (cls.k_modules_file, parent_commit)) + # END handle exceptions # END handle non-bare working tree if not read_only and not parent_matches_head: raise ValueError("Cannot write blobs of 'historical' submodule configurations") # END handle writes of historical submodules - parser = GitConfigParser(fp_module, read_only = read_only) - return SectionConstraint(parser, 'submodule "%s"' % self.path) + return GitConfigParser(fp_module, read_only = read_only) + + + @classmethod + def _sio_modules(cls, parent_commit): + """:return: Configuration file as StringIO - we only access it through the respective blob's data""" + sio = StringIO(parent_commit.tree[cls.k_modules_file].data_stream.read()) + sio.name = cls.k_modules_file + return sio + + def _config_parser_constrained(self, read_only): + """:return: Config Parser constrained to our submodule in read or write mode""" + parser = self._config_parser(self.repo, self._parent_commit, read_only) + return SectionConstraint(parser, sm_section(self.name)) #{ Edit Interface @@ -81,29 +147,52 @@ class Submodule(base.IndexObject): :param skip_init: if True, the new repository will not be cloned to its location. :return: The newly created submodule instance""" - def set_parent_commit(self, commit): + def set_parent_commit(self, commit, check=True): """Set this instance to use the given commit whose tree is supposed to contain the .gitmodules blob. :param commit: Commit'ish reference pointing at the root_tree - :raise ValueError: if the commit's tree didn't contain the .gitmodules blob.""" + :param check: if True, relatively expensive checks will be performed to verify + validity of the submodule. + :raise ValueError: if the commit's tree didn't contain the .gitmodules blob. + :raise ValueError: if the parent commit didn't store this submodule under the + current path""" pcommit = self.repo.commit(commit) - if self.kModulesFile not in pcommit.tree: - raise ValueError("Tree of commit %s did not contain the %s file" % (commit, self.kModulesFile)) + pctree = pcommit.tree + if self.k_modules_file not in pctree: + raise ValueError("Tree of commit %s did not contain the %s file" % (commit, self.k_modules_file)) # END handle exceptions + + prev_pc = self._parent_commit self._parent_commit = pcommit + if check: + parser = self._config_parser(self.repo, self._parent_commit, read_only=True) + if not parser.has_section(sm_section(self.name)): + self._parent_commit = prev_pc + raise ValueError("Submodule at path %r did not exist in parent commit %s" % (self.path, commit)) + # END handle submodule did not exist + # END handle checking mode + + # update our sha, it could have changed + self.binsha = pctree[self.path].binsha + # clear the possibly changed values for name in ('path', '_ref', '_url'): try: delattr(self, name) except AttributeError: pass + # END try attr deletion # END for each name to delete def config_writer(self): """:return: a config writer instance allowing you to read and write the data - belonging to this submodule into the .gitmodules file.""" - return self._config_parser(read_only=False) + belonging to this submodule into the .gitmodules file. + + :raise ValueError: if trying to get a writer on a parent_commit which does not + match the current head commit + :raise IOError: If the .gitmodules file/blob could not be read""" + return self._config_parser_constrained(read_only=False) #} END edit interface @@ -111,37 +200,104 @@ class Submodule(base.IndexObject): def module(self): """:return: Repo instance initialized from the repository at our submodule path - :raise InvalidGitRepositoryError: if a repository was not available""" + :raise InvalidGitRepositoryError: if a repository was not available. This could + also mean that it was not yet initialized""" + # late import to workaround circular dependencies + from git.repo import Repo + if self.repo.bare: raise InvalidGitRepositoryError("Cannot retrieve module repository in bare parent repositories") # END handle bare mode repo_path = join_path_native(self.repo.working_tree_dir, self.path) try: - return Repo(repo_path) + repo = Repo(repo_path) + if repo != self.repo: + return repo + # END handle repo uninitialized except (InvalidGitRepositoryError, NoSuchPathError): raise InvalidGitRepositoryError("No valid repository at %s" % self.path) + else: + raise InvalidGitRepositoryError("Repository at %r was not yet checked out" % repo_path) # END handle exceptions - + + @property def ref(self): """:return: The reference's name that we are to checkout""" return self._ref - + + @property def url(self): """:return: The url to the repository which our module-repository refers to""" return self._url + @property def parent_commit(self): """:return: Commit instance with the tree containing the .gitmodules file :note: will always point to the current head's commit if it was not set explicitly""" return self._parent_commit + + @property + def name(self): + """:return: The name of this submodule. It is used to identify it within the + .gitmodules file. + :note: by default, the name is the path at which to find the submodule, but + in git-python it should be a unique identifier similar to the identifiers + used for remotes, which allows to change the path of the submodule + easily + """ + return self._name def config_reader(self): """:return: ConfigReader instance which allows you to qurey the configuration values of this submodule, as provided by the .gitmodules file :note: The config reader will actually read the data directly from the repository and thus does not need nor care about your working tree. - :note: Should be cached by the caller and only kept as long as needed""" - return self._config_parser.read_only(read_only=True) + :note: Should be cached by the caller and only kept as long as needed + :raise IOError: If the .gitmodules file/blob could not be read""" + return self._config_parser_constrained(read_only=True) #} END query interface + + #{ Iterable Interface + + @classmethod + def iter_items(cls, repo, parent_commit='HEAD'): + """:return: iterator yielding Submodule instances available in the given repository""" + pc = repo.commit(parent_commit) # parent commit instance + try: + parser = cls._config_parser(repo, pc, read_only=True) + except IOError: + raise StopIteration + # END handle empty iterator + + rt = pc.tree # root tree + + for sms in parser.sections(): + n = sm_name(sms) + p = parser.get_value(sms, 'path') + u = parser.get_value(sms, 'url') + r = cls.k_ref_default + if parser.has_option(sms, cls.k_ref_option): + r = parser.get_value(sms, cls.k_ref_option) + # END handle optional information + + # get the binsha + try: + sm = rt[p] + except KeyError: + raise InvalidGitRepositoryError("Gitmodule path %r did not exist in revision of parent commit %s" % (p, parent_commit)) + # END handle critical error + + # fill in remaining info - saves time as it doesn't have to be parsed again + sm._name = n + sm._parent_commit = pc + sm._ref = r + sm._url = u + + yield sm + # END for each section + + #} END iterable interface + +#} END classes diff --git a/lib/git/objects/tag.py b/lib/git/objects/tag.py index ea480fc2..c7d02abe 100644 --- a/lib/git/objects/tag.py +++ b/lib/git/objects/tag.py @@ -33,7 +33,18 @@ class TagObject(base.Object): :param tagged_tz_offset: int_seconds_west_of_utc is the timezone that the authored_date is in, in a format similar to time.altzone""" super(TagObject, self).__init__(repo, binsha ) - self._set_self_from_args_(locals()) + if object is not None: + self.object = object + if tag is not None: + self.tag = tag + if tagger is not None: + self.tagger = tagger + if tagged_date is not None: + self.tagged_date = tagged_date + if tagger_tz_offset is not None: + self.tagger_tz_offset = tagger_tz_offset + if message is not None: + self.message = message def _set_cache_(self, attr): """Cache all our attributes at once""" -- cgit v1.2.1 From f97653aa06cf84bcf160be3786b6fce49ef52961 Mon Sep 17 00:00:00 2001 From: Sebastian Thiel Date: Mon, 15 Nov 2010 18:42:44 +0100 Subject: Repo: added submodule query and iteration methods similar to the ones provided for Remotes, including test --- lib/git/objects/submodule.py | 46 +++++++++++++++++++++++++++++++++++++++++--- lib/git/objects/util.py | 8 ++++++++ lib/git/repo/base.py | 22 ++++++++++++++++++++- lib/git/util.py | 3 +++ 4 files changed, 75 insertions(+), 4 deletions(-) (limited to 'lib/git') diff --git a/lib/git/objects/submodule.py b/lib/git/objects/submodule.py index 1aa0cfb5..eda95115 100644 --- a/lib/git/objects/submodule.py +++ b/lib/git/objects/submodule.py @@ -1,13 +1,15 @@ import base +from util import Traversable from StringIO import StringIO # need a dict to set bloody .name field from git.util import Iterable from git.config import GitConfigParser, SectionConstraint from git.util import join_path_native from git.exc import InvalidGitRepositoryError, NoSuchPathError +import stat import os -__all__ = ("Submodule", ) +__all__ = ("Submodule", "RootModule") #{ Utilities @@ -31,7 +33,7 @@ class SubmoduleConfigParser(GitConfigParser): _mutating_methods_ = tuple() -class Submodule(base.IndexObject, Iterable): +class Submodule(base.IndexObject, Iterable, Traversable): """Implements access to a git submodule. They are special in that their sha represents a commit in the submodule's repository which is to be checked out at the path of this instance. @@ -40,10 +42,11 @@ class Submodule(base.IndexObject, Iterable): All methods work in bare and non-bare repositories.""" - _id_attribute_ = "path" + _id_attribute_ = "name" k_modules_file = '.gitmodules' k_ref_option = 'ref' k_ref_default = 'master' + k_def_mode = stat.S_IFDIR | stat.S_IFLNK # submodules are directories with link-status # this is a bogus type for base class compatability type = 'submodule' @@ -86,6 +89,14 @@ class Submodule(base.IndexObject, Iterable): super(Submodule, self)._set_cache_(attr) # END handle attribute name + def _get_intermediate_items(self, item): + """:return: all the submodules of our module repository""" + try: + return type(self).list_items(item.module()) + except InvalidGitRepositoryError: + return list() + # END handle intermeditate items + def __eq__(self, other): """Compare with another submodule""" return self.path == other.path and self.url == other.url and super(Submodule, self).__eq__(other) @@ -107,6 +118,7 @@ class Submodule(base.IndexObject, Iterable): if not os.path.isfile(fp_module_path): raise IOError("%s file was not accessible" % fp_module_path) # END handle existance + fp_module = fp_module_path else: try: fp_module = cls._sio_modules(parent_commit) @@ -300,4 +312,32 @@ class Submodule(base.IndexObject, Iterable): #} END iterable interface + +class RootModule(Submodule): + """A (virtual) Root of all submodules in the given repository. It can be used + to more easily traverse all submodules of the master repository""" + + __slots__ = tuple() + + k_root_name = '__ROOT__' + + def __init__(self, repo): + # repo, binsha, mode=None, path=None, name = None, parent_commit=None, url=None, ref=None) + super(RootModule, self).__init__( + repo, + binsha = self.NULL_BIN_SHA, + mode = self.k_def_mode, + path = '', + name = self.k_root_name, + parent_commit = repo.head.commit, + url = '', + ref = self.k_ref_default + ) + + + #{ Interface + def module(self): + """:return: the actual repository containing the submodules""" + return self.repo + #} END interface #} END classes diff --git a/lib/git/objects/util.py b/lib/git/objects/util.py index 21833080..9a54e031 100644 --- a/lib/git/objects/util.py +++ b/lib/git/objects/util.py @@ -4,6 +4,8 @@ # This module is part of GitPython and is released under # the BSD License: http://www.opensource.org/licenses/bsd-license.php """Module for general utility functions""" +from git.util import IterableList + import re from collections import deque as Deque import platform @@ -273,6 +275,12 @@ class Traversable(object): """ raise NotImplementedError("To be implemented in subclass") + def list_traverse(self, *args, **kwargs): + """:return: IterableList with the results of the traversal as produced by + traverse()""" + out = IterableList(self._id_attribute_) + out.extend(self.traverse(*args, **kwargs)) + return out def traverse( self, predicate = lambda i,d: True, prune = lambda i,d: False, depth = -1, branch_first=True, diff --git a/lib/git/repo/base.py b/lib/git/repo/base.py index 790b1283..3a395af0 100644 --- a/lib/git/repo/base.py +++ b/lib/git/repo/base.py @@ -6,7 +6,6 @@ from git.exc import InvalidGitRepositoryError, NoSuchPathError from git.cmd import Git -from git.objects import Actor from git.refs import * from git.index import IndexFile from git.objects import * @@ -222,6 +221,27 @@ class Repo(object): """:return: Remote with the specified name :raise ValueError: if no remote with such a name exists""" return Remote(self, name) + + @property + def submodules(self): + """:return: git.IterableList(Submodule, ...) of direct submodules""" + return self.list_submodules(recursive=False) + + def submodule(self, name): + """:return: Submodule with the given name + :raise ValueError: If no such submodule exists""" + try: + return self.submodules[name] + except IndexError: + raise ValueError("Didn't find submodule named %r" % name) + # END exception handling + + def list_submodules(self, recursive=False): + """A list if Submodule objects available in this repository + :param recursive: If True, submodules of submodules (and so forth) will be + returned as well as part of a depth-first traversal + :return: ``git.IterableList(Submodule, ...)""" + return RootModule(self).list_traverse(ignore_self=1, depth = recursive and -1 or 1) @property def tags(self): diff --git a/lib/git/util.py b/lib/git/util.py index fcb50585..b77e7904 100644 --- a/lib/git/util.py +++ b/lib/git/util.py @@ -296,6 +296,9 @@ class IterableList(list): def __init__(self, id_attr, prefix=''): self._id_attr = id_attr self._prefix = prefix + if not isinstance(id_attr, basestring): + raise ValueError("First parameter must be a string identifying the name-property. Extend the list after initialization") + # END help debugging ! def __getattr__(self, attr): attr = self._prefix + attr -- cgit v1.2.1 From 624556eae1c292a1dc283d9dca1557e28abe8ee3 Mon Sep 17 00:00:00 2001 From: Sebastian Thiel Date: Mon, 15 Nov 2010 19:03:53 +0100 Subject: Optimized test-decorators, by completely removing with_bare_rw_repo, which was mainly copy-paste from with_rw_repo, what a shame --- lib/git/objects/base.py | 5 ++++- lib/git/repo/base.py | 13 ++++++------- 2 files changed, 10 insertions(+), 8 deletions(-) (limited to 'lib/git') diff --git a/lib/git/objects/base.py b/lib/git/objects/base.py index 82c2589c..b8cec47f 100644 --- a/lib/git/objects/base.py +++ b/lib/git/objects/base.py @@ -114,7 +114,10 @@ class Object(LazyMixin): class IndexObject(Object): """Base for all objects that can be part of the index file , namely Tree, Blob and SubModule objects""" - __slots__ = ("path", "mode") + __slots__ = ("path", "mode") + + # for compatability with iterable lists + _id_attribute_ = 'path' def __init__(self, repo, binsha, mode=None, path=None): """Initialize a newly instanced IndexObject diff --git a/lib/git/repo/base.py b/lib/git/repo/base.py index 3a395af0..0355b062 100644 --- a/lib/git/repo/base.py +++ b/lib/git/repo/base.py @@ -225,7 +225,7 @@ class Repo(object): @property def submodules(self): """:return: git.IterableList(Submodule, ...) of direct submodules""" - return self.list_submodules(recursive=False) + return Submodule.list_items(self) def submodule(self, name): """:return: Submodule with the given name @@ -236,12 +236,11 @@ class Repo(object): raise ValueError("Didn't find submodule named %r" % name) # END exception handling - def list_submodules(self, recursive=False): - """A list if Submodule objects available in this repository - :param recursive: If True, submodules of submodules (and so forth) will be - returned as well as part of a depth-first traversal - :return: ``git.IterableList(Submodule, ...)""" - return RootModule(self).list_traverse(ignore_self=1, depth = recursive and -1 or 1) + def iter_submodules(self, *args, **kwargs): + """An iterator yielding Submodule instances, see Traversable interface + for a description of args and kwargs + :return: Iterator""" + return RootModule(self).traverse(*args, **kwargs) @property def tags(self): -- cgit v1.2.1 From ceee7d7e0d98db12067744ac3cd0ab3a49602457 Mon Sep 17 00:00:00 2001 From: Sebastian Thiel Date: Mon, 15 Nov 2010 22:12:05 +0100 Subject: Added partial implementation of update, but realized that using refs in general may be contradicting if a tag is given there, as well as a commit sha of the submodule. Hence it should really be only a branch --- lib/git/objects/submodule.py | 60 +++++++++++++++++++++++++++++++++++++++++++- lib/git/refs.py | 2 ++ 2 files changed, 61 insertions(+), 1 deletion(-) (limited to 'lib/git') diff --git a/lib/git/objects/submodule.py b/lib/git/objects/submodule.py index eda95115..12610abd 100644 --- a/lib/git/objects/submodule.py +++ b/lib/git/objects/submodule.py @@ -56,6 +56,7 @@ class Submodule(base.IndexObject, Iterable, Traversable): def __init__(self, repo, binsha, mode=None, path=None, name = None, parent_commit=None, url=None, ref=None): """Initialize this instance with its attributes. We only document the ones that differ from ``IndexObject`` + :param repo: Our parent repository :param binsha: binary sha referring to a commit in the remote repository, see url parameter :param parent_commit: see set_parent_commit() :param url: The url to the remote repository which is the submodule @@ -105,6 +106,10 @@ class Submodule(base.IndexObject, Iterable, Traversable): """Compare with another submodule for inequality""" return not (self == other) + def __hash__(self): + """Hash this instance using its logical id, not the sha""" + return hash(self._name) + @classmethod def _config_parser(cls, repo, parent_commit, read_only): """:return: Config Parser constrained to our submodule in read or write mode @@ -159,6 +164,57 @@ class Submodule(base.IndexObject, Iterable, Traversable): :param skip_init: if True, the new repository will not be cloned to its location. :return: The newly created submodule instance""" + def update(self, recursive=False, init=True): + """Update the repository of this submodule to point to the checkout + we point at with the binsha of this instance. + :param recursive: if True, we will operate recursively and update child- + modules as well. + :param init: if True, the module repository will be cloned into place if necessary + :note: does nothing in bare repositories + :return: self""" + if self.repo.bare: + return self + #END pass in bare mode + + try: + mrepo = self.module() + except InvalidGitRepositoryError: + if not init: + return self + # END early abort if init is not allowed + import git + + # there is no git-repository yet - but delete empty paths + module_path = join_path_native(self.repo.working_tree_dir, self.path) + if os.path.isdir(module_path): + try: + os.rmdir(module_path) + except OSError: + raise OSError("Module directory at %r does already exist and is non-empty" % module_path) + # END handle OSError + # END handle directory removal + + # don't check it out at first + mrepo = git.Repo.clone_from(self.url, self.path, n=True) + # ref can be a tag or a branch - we can checkout branches, but not tags + # tag_ref = git.TagReference(mrepo, TagReference.to_full_path(self.ref)) + if tag_ref.is_valid(): + #if tag_ref.commit + mrepo.git.checkout(tag_ref) + else: + # assume it is a branch and try it + mrepo.git.checkout(self.hexsha, b=self.ref) + #if mrepo.head.ref.name != self.ref: + # mrepo.head.ref = git.Head(mrepo, git.Head.to_full_path(self.ref + #END handle initalization + + # TODO: handle ref-path + if mrepo.head.commit.binsha != self.binsha: + mrepo.git.checkout(self.binsha) + # END handle checkout + + return self + def set_parent_commit(self, commit, check=True): """Set this instance to use the given commit whose tree is supposed to contain the .gitmodules blob. @@ -167,7 +223,8 @@ class Submodule(base.IndexObject, Iterable, Traversable): validity of the submodule. :raise ValueError: if the commit's tree didn't contain the .gitmodules blob. :raise ValueError: if the parent commit didn't store this submodule under the - current path""" + current path + :return: self""" pcommit = self.repo.commit(commit) pctree = pcommit.tree if self.k_modules_file not in pctree: @@ -196,6 +253,7 @@ class Submodule(base.IndexObject, Iterable, Traversable): pass # END try attr deletion # END for each name to delete + return self def config_writer(self): """:return: a config writer instance allowing you to read and write the data diff --git a/lib/git/refs.py b/lib/git/refs.py index af7284ff..3dc73d03 100644 --- a/lib/git/refs.py +++ b/lib/git/refs.py @@ -489,6 +489,8 @@ class SymbolicReference(object): @classmethod def from_path(cls, repo, path): """ + :param path: full .git-directory-relative path name to the Reference to instantiate + :note: use to_full_path() if you only have a partial path of a known Reference Type :return: Instance of type Reference, Head, or Tag depending on the given path""" -- cgit v1.2.1 From d4fd7fca515ba9b088a7c811292f76f47d16cd7b Mon Sep 17 00:00:00 2001 From: Sebastian Thiel Date: Tue, 16 Nov 2010 00:18:13 +0100 Subject: Submodule now only supports branches to be given as hint that will svn-external like behaviour. Implemented first version of update, which works for now, but probably needs to see more features --- lib/git/config.py | 30 ++++---- lib/git/objects/submodule.py | 164 ++++++++++++++++++++++++++++++------------- lib/git/objects/util.py | 2 +- 3 files changed, 134 insertions(+), 62 deletions(-) (limited to 'lib/git') diff --git a/lib/git/config.py b/lib/git/config.py index 8541dc0e..073efd63 100644 --- a/lib/git/config.py +++ b/lib/git/config.py @@ -23,19 +23,23 @@ class MetaParserBuilder(type): """ Equip all base-class methods with a needs_values decorator, and all non-const methods with a set_dirty_and_flush_changes decorator in addition to that.""" - mutating_methods = clsdict['_mutating_methods_'] - for base in bases: - methods = ( t for t in inspect.getmembers(base, inspect.ismethod) if not t[0].startswith("_") ) - for name, method in methods: - if name in clsdict: - continue - method_with_values = needs_values(method) - if name in mutating_methods: - method_with_values = set_dirty_and_flush_changes(method_with_values) - # END mutating methods handling - - clsdict[name] = method_with_values - # END for each base + kmm = '_mutating_methods_' + if kmm in clsdict: + mutating_methods = clsdict[kmm] + for base in bases: + methods = ( t for t in inspect.getmembers(base, inspect.ismethod) if not t[0].startswith("_") ) + for name, method in methods: + if name in clsdict: + continue + method_with_values = needs_values(method) + if name in mutating_methods: + method_with_values = set_dirty_and_flush_changes(method_with_values) + # END mutating methods handling + + clsdict[name] = method_with_values + # END for each name/method pair + # END for each base + # END if mutating methods configuration is set new_type = super(MetaParserBuilder, metacls).__new__(metacls, name, bases, clsdict) return new_type diff --git a/lib/git/objects/submodule.py b/lib/git/objects/submodule.py index 12610abd..86aba49c 100644 --- a/lib/git/objects/submodule.py +++ b/lib/git/objects/submodule.py @@ -8,6 +8,8 @@ from git.exc import InvalidGitRepositoryError, NoSuchPathError import stat import os +import sys +import weakref __all__ = ("Submodule", "RootModule") @@ -27,11 +29,43 @@ def sm_name(section): #{ Classes class SubmoduleConfigParser(GitConfigParser): - """Catches calls to _write, and updates the .gitmodules blob in the index + """ + Catches calls to _write, and updates the .gitmodules blob in the index with the new data, if we have written into a stream. Otherwise it will - add the local file to the index to make it correspond with the working tree.""" - _mutating_methods_ = tuple() + add the local file to the index to make it correspond with the working tree. + Additionally, the cache must be cleared + """ + def __init__(self, *args, **kwargs): + self._smref = None + super(SubmoduleConfigParser, self).__init__(*args, **kwargs) + + #{ Interface + def set_submodule(self, submodule): + """Set this instance's submodule. It must be called before + the first write operation begins""" + self._smref = weakref.ref(submodule) + + def flush_to_index(self): + """Flush changes in our configuration file to the index""" + assert self._smref is not None + # should always have a file here + assert not isinstance(self._file_or_files, StringIO) + + sm = self._smref() + if sm is not None: + sm.repo.index.add([sm.k_modules_file]) + sm._clear_cache() + # END handle weakref + + #} END interface + + #{ Overridden Methods + def write(self): + rval = super(SubmoduleConfigParser, self).write() + self.flush_to_index() + return rval + # END overridden methods class Submodule(base.IndexObject, Iterable, Traversable): """Implements access to a git submodule. They are special in that their sha @@ -44,16 +78,16 @@ class Submodule(base.IndexObject, Iterable, Traversable): _id_attribute_ = "name" k_modules_file = '.gitmodules' - k_ref_option = 'ref' - k_ref_default = 'master' + k_head_option = 'branch' + k_head_default = 'master' k_def_mode = stat.S_IFDIR | stat.S_IFLNK # submodules are directories with link-status # this is a bogus type for base class compatability type = 'submodule' - __slots__ = ('_parent_commit', '_url', '_ref', '_name') + __slots__ = ('_parent_commit', '_url', '_branch', '_name', '__weakref__') - def __init__(self, repo, binsha, mode=None, path=None, name = None, parent_commit=None, url=None, ref=None): + def __init__(self, repo, binsha, mode=None, path=None, name = None, parent_commit=None, url=None, branch=None): """Initialize this instance with its attributes. We only document the ones that differ from ``IndexObject`` :param repo: Our parent repository @@ -66,8 +100,8 @@ class Submodule(base.IndexObject, Iterable, Traversable): self._parent_commit = parent_commit if url is not None: self._url = url - if ref is not None: - self._ref = ref + if branch is not None: + self._branch = branch if name is not None: self._name = name @@ -77,13 +111,13 @@ class Submodule(base.IndexObject, Iterable, Traversable): elif attr == '_parent_commit': # set a default value, which is the root tree of the current head self._parent_commit = self.repo.commit() - elif attr in ('path', '_url', '_ref'): + elif attr in ('path', '_url', '_branch'): reader = self.config_reader() # default submodule values self.path = reader.get_value('path') self._url = reader.get_value('url') # git-python extension values - optional - self._ref = reader.get_value(self.k_ref_option, self.k_ref_default) + self._branch = reader.get_value(self.k_head_option, self.k_head_default) elif attr == '_name': raise AttributeError("Cannot retrieve the name of a submodule if it was not set initially") else: @@ -132,12 +166,21 @@ class Submodule(base.IndexObject, Iterable, Traversable): # END handle exceptions # END handle non-bare working tree - if not read_only and not parent_matches_head: + if not read_only and (repo.bare or not parent_matches_head): raise ValueError("Cannot write blobs of 'historical' submodule configurations") # END handle writes of historical submodules - return GitConfigParser(fp_module, read_only = read_only) + return SubmoduleConfigParser(fp_module, read_only = read_only) + def _clear_cache(self): + # clear the possibly changed values + for name in ('path', '_branch', '_url'): + try: + delattr(self, name) + except AttributeError: + pass + # END try attr deletion + # END for each name to delete @classmethod def _sio_modules(cls, parent_commit): @@ -149,6 +192,7 @@ class Submodule(base.IndexObject, Iterable, Traversable): def _config_parser_constrained(self, read_only): """:return: Config Parser constrained to our submodule in read or write mode""" parser = self._config_parser(self.repo, self._parent_commit, read_only) + parser.set_submodule(self) return SectionConstraint(parser, sm_section(self.name)) #{ Edit Interface @@ -178,6 +222,9 @@ class Submodule(base.IndexObject, Iterable, Traversable): try: mrepo = self.module() + for remote in mrepo.remotes: + remote.fetch() + #END fetch new data except InvalidGitRepositoryError: if not init: return self @@ -194,25 +241,42 @@ class Submodule(base.IndexObject, Iterable, Traversable): # END handle OSError # END handle directory removal - # don't check it out at first - mrepo = git.Repo.clone_from(self.url, self.path, n=True) - # ref can be a tag or a branch - we can checkout branches, but not tags - # tag_ref = git.TagReference(mrepo, TagReference.to_full_path(self.ref)) - if tag_ref.is_valid(): - #if tag_ref.commit - mrepo.git.checkout(tag_ref) - else: - # assume it is a branch and try it - mrepo.git.checkout(self.hexsha, b=self.ref) - #if mrepo.head.ref.name != self.ref: - # mrepo.head.ref = git.Head(mrepo, git.Head.to_full_path(self.ref + # don't check it out at first - nonetheless it will create a local + # branch according to the remote-HEAD if possible + mrepo = git.Repo.clone_from(self.url, module_path, n=True) + + # see whether we have a valid branch to checkout + try: + remote_branch = mrepo.remotes.origin.refs[self.branch] + local_branch = git.Head(mrepo, git.Head.to_full_path(self.branch)) + if not local_branch.is_valid(): + mrepo.git.checkout(remote_branch, b=self.branch) + # END initial checkout + branch creation + # make sure we are not detached + mrepo.head.ref = local_branch + except IndexError: + print >> sys.stderr, "Warning: Failed to checkout tracking branch %s" % self.branch + #END handle tracking branch #END handle initalization - # TODO: handle ref-path - if mrepo.head.commit.binsha != self.binsha: - mrepo.git.checkout(self.binsha) + # if the commit to checkout is on the current branch, merge the branch + if mrepo.head.is_detached: + if mrepo.head.commit.binsha != self.binsha: + mrepo.git.checkout(self.hexsha) + # END checkout commit + else: + # TODO: allow to specify a rebase, merge, or reset + # TODO: Warn if the hexsha forces the tracking branch off the remote + # branch - this should be prevented when setting the branch option + mrepo.head.reset(self.hexsha, index=True, working_tree=True) # END handle checkout + if recursive: + for submodule in self.iter_items(self.module()): + submodule.update(recursive, init) + # END handle recursive update + # END for each submodule + return self def set_parent_commit(self, commit, check=True): @@ -245,14 +309,8 @@ class Submodule(base.IndexObject, Iterable, Traversable): # update our sha, it could have changed self.binsha = pctree[self.path].binsha - # clear the possibly changed values - for name in ('path', '_ref', '_url'): - try: - delattr(self, name) - except AttributeError: - pass - # END try attr deletion - # END for each name to delete + self._clear_cache() + return self def config_writer(self): @@ -262,6 +320,8 @@ class Submodule(base.IndexObject, Iterable, Traversable): :raise ValueError: if trying to get a writer on a parent_commit which does not match the current head commit :raise IOError: If the .gitmodules file/blob could not be read""" + if self.repo.bare: + raise InvalidGitRepositoryError("Cannot change submodule configuration in a bare repository") return self._config_parser_constrained(read_only=False) #} END edit interface @@ -279,24 +339,28 @@ class Submodule(base.IndexObject, Iterable, Traversable): raise InvalidGitRepositoryError("Cannot retrieve module repository in bare parent repositories") # END handle bare mode - repo_path = join_path_native(self.repo.working_tree_dir, self.path) + module_path = self.module_path() try: - repo = Repo(repo_path) + repo = Repo(module_path) if repo != self.repo: return repo # END handle repo uninitialized except (InvalidGitRepositoryError, NoSuchPathError): raise InvalidGitRepositoryError("No valid repository at %s" % self.path) else: - raise InvalidGitRepositoryError("Repository at %r was not yet checked out" % repo_path) + raise InvalidGitRepositoryError("Repository at %r was not yet checked out" % module_path) # END handle exceptions + + def module_path(self): + """:return: full path to the root of our module. It is relative to the filesystem root""" + return join_path_native(self.repo.working_tree_dir, self.path) @property - def ref(self): - """:return: The reference's name that we are to checkout""" - return self._ref + def branch(self): + """:return: The branch name that we are to checkout""" + return self._branch - @property + @property def url(self): """:return: The url to the repository which our module-repository refers to""" return self._url @@ -347,9 +411,9 @@ class Submodule(base.IndexObject, Iterable, Traversable): n = sm_name(sms) p = parser.get_value(sms, 'path') u = parser.get_value(sms, 'url') - r = cls.k_ref_default - if parser.has_option(sms, cls.k_ref_option): - r = parser.get_value(sms, cls.k_ref_option) + b = cls.k_head_default + if parser.has_option(sms, cls.k_head_option): + b = parser.get_value(sms, cls.k_head_option) # END handle optional information # get the binsha @@ -362,7 +426,7 @@ class Submodule(base.IndexObject, Iterable, Traversable): # fill in remaining info - saves time as it doesn't have to be parsed again sm._name = n sm._parent_commit = pc - sm._ref = r + sm._branch = b sm._url = u yield sm @@ -389,10 +453,14 @@ class RootModule(Submodule): name = self.k_root_name, parent_commit = repo.head.commit, url = '', - ref = self.k_ref_default + branch = self.k_head_default ) + def _clear_cache(self): + """May not do anything""" + pass + #{ Interface def module(self): """:return: the actual repository containing the submodules""" diff --git a/lib/git/objects/util.py b/lib/git/objects/util.py index 9a54e031..81544e26 100644 --- a/lib/git/objects/util.py +++ b/lib/git/objects/util.py @@ -343,7 +343,7 @@ class Traversable(object): if prune( rval, d ): continue - skipStartItem = ignore_self and ( item == self ) + skipStartItem = ignore_self and ( item is self ) if not skipStartItem and predicate( rval, d ): yield rval -- cgit v1.2.1 From af5abca21b56fcf641ff916bd567680888c364aa Mon Sep 17 00:00:00 2001 From: Sebastian Thiel Date: Tue, 16 Nov 2010 09:30:10 +0100 Subject: Added a few utility methods and improved the test. Refs need an improvement though to allow easy configuration of branch-specific settings --- lib/git/objects/submodule.py | 60 +++++++++++++++++++++++++++++++------------- 1 file changed, 43 insertions(+), 17 deletions(-) (limited to 'lib/git') diff --git a/lib/git/objects/submodule.py b/lib/git/objects/submodule.py index 86aba49c..d58e07a9 100644 --- a/lib/git/objects/submodule.py +++ b/lib/git/objects/submodule.py @@ -251,6 +251,17 @@ class Submodule(base.IndexObject, Iterable, Traversable): local_branch = git.Head(mrepo, git.Head.to_full_path(self.branch)) if not local_branch.is_valid(): mrepo.git.checkout(remote_branch, b=self.branch) + else: + # have a valid branch, but no checkout - make sure we can figure + # that out by marking the commit with a null_sha + # have to write it directly as .commit = NULLSHA tries to resolve the sha + ref = mrepo.head.ref + refpath = join_path_native(mrepo.git_dir, ref.to_full_path(ref.path)) + refdir = os.path.dirname(refpath) + if not os.path.isdir(refdir): + os.makedirs(refdir) + #END handle directory + open(refpath, 'w').write(self.NULL_HEX_SHA) # END initial checkout + branch creation # make sure we are not detached mrepo.head.ref = local_branch @@ -259,24 +270,24 @@ class Submodule(base.IndexObject, Iterable, Traversable): #END handle tracking branch #END handle initalization - # if the commit to checkout is on the current branch, merge the branch - if mrepo.head.is_detached: - if mrepo.head.commit.binsha != self.binsha: + # update the working tree + if mrepo.head.commit.binsha != self.binsha: + if mrepo.head.is_detached: mrepo.git.checkout(self.hexsha) - # END checkout commit - else: - # TODO: allow to specify a rebase, merge, or reset - # TODO: Warn if the hexsha forces the tracking branch off the remote - # branch - this should be prevented when setting the branch option - mrepo.head.reset(self.hexsha, index=True, working_tree=True) - # END handle checkout - - if recursive: - for submodule in self.iter_items(self.module()): - submodule.update(recursive, init) - # END handle recursive update - # END for each submodule - + else: + # TODO: allow to specify a rebase, merge, or reset + # TODO: Warn if the hexsha forces the tracking branch off the remote + # branch - this should be prevented when setting the branch option + mrepo.head.reset(self.hexsha, index=True, working_tree=True) + # END handle checkout + + if recursive: + for submodule in self.iter_items(self.module()): + submodule.update(recursive, init) + # END handle recursive update + # END for each submodule + # END update to new commit only if needed + return self def set_parent_commit(self, commit, check=True): @@ -354,6 +365,15 @@ class Submodule(base.IndexObject, Iterable, Traversable): def module_path(self): """:return: full path to the root of our module. It is relative to the filesystem root""" return join_path_native(self.repo.working_tree_dir, self.path) + + def module_exists(self): + """:return: True if our module exists and is a valid git repository. See module() method""" + try: + self.module() + return True + except InvalidGitRepositoryError: + return False + # END handle exception @property def branch(self): @@ -391,6 +411,12 @@ class Submodule(base.IndexObject, Iterable, Traversable): :raise IOError: If the .gitmodules file/blob could not be read""" return self._config_parser_constrained(read_only=True) + def children(self): + """:return: IterableList(Submodule, ...) an iterable list of submodules instances + which are children of this submodule + :raise InvalidGitRepositoryError: if the submodule is not checked-out""" + return self._get_intermediate_items(self) + #} END query interface #{ Iterable Interface -- cgit v1.2.1 From 9f73e8ba55f33394161b403bf7b8c2e0e05f47b0 Mon Sep 17 00:00:00 2001 From: Sebastian Thiel Date: Tue, 16 Nov 2010 11:05:31 +0100 Subject: remote: added methods to set and query the tracking branch status of normal heads, including test. Config: SectionConstraint was updated with additional callable methods, the complete ConfigParser interface should be covered now Remote: refs methods is much more efficient now as it will set the search path to the directory containing the remote refs - previously it used the remotes/ base directory and pruned the search result --- lib/git/config.py | 3 ++- lib/git/refs.py | 78 +++++++++++++++++++++++++++++++++++++++++++++++++++++++ lib/git/remote.py | 6 +---- 3 files changed, 81 insertions(+), 6 deletions(-) (limited to 'lib/git') diff --git a/lib/git/config.py b/lib/git/config.py index 073efd63..0528f318 100644 --- a/lib/git/config.py +++ b/lib/git/config.py @@ -74,7 +74,8 @@ class SectionConstraint(object): It supports all ConfigParser methods that operate on an option""" __slots__ = ("_config", "_section_name") - _valid_attrs_ = ("get_value", "set_value", "get", "set", "getint", "getfloat", "getboolean", "has_option") + _valid_attrs_ = ("get_value", "set_value", "get", "set", "getint", "getfloat", "getboolean", "has_option", + "remove_section", "remove_option", "options") def __init__(self, config, section): self._config = config diff --git a/lib/git/refs.py b/lib/git/refs.py index 3dc73d03..39c5ff29 100644 --- a/lib/git/refs.py +++ b/lib/git/refs.py @@ -29,6 +29,11 @@ from gitdb.util import ( hex_to_bin ) +from config import ( + GitConfigParser, + SectionConstraint + ) + from exc import GitCommandError __all__ = ("SymbolicReference", "Reference", "HEAD", "Head", "TagReference", @@ -701,6 +706,8 @@ class Head(Reference): >>> head.commit.hexsha '1c09f116cbc2cb4100fb6935bb162daa4723f455'""" _common_path_default = "refs/heads" + k_config_remote = "remote" + k_config_remote_ref = "merge" # branch to merge from remote @classmethod def create(cls, repo, path, commit='HEAD', force=False, **kwargs): @@ -747,6 +754,44 @@ class Head(Reference): flag = "-D" repo.git.branch(flag, *heads) + + def set_tracking_branch(self, remote_reference): + """Configure this branch to track the given remote reference. This will alter + this branch's configuration accordingly. + :param remote_reference: The remote reference to track or None to untrack + any references + :return: self""" + if remote_reference is not None and not isinstance(remote_reference, RemoteReference): + raise ValueError("Incorrect parameter type: %r" % remote_reference) + # END handle type + + writer = self.config_writer() + if remote_reference is None: + writer.remove_option(self.k_config_remote) + writer.remove_option(self.k_config_remote_ref) + if len(writer.options()) == 0: + writer.remove_section() + # END handle remove section + else: + writer.set_value(self.k_config_remote, remote_reference.remote_name) + writer.set_value(self.k_config_remote_ref, Head.to_full_path(remote_reference.remote_head)) + # END handle ref value + + return self + + + def tracking_branch(self): + """:return: The remote_reference we are tracking, or None if we are + not a tracking branch""" + reader = self.config_reader() + if reader.has_option(self.k_config_remote) and reader.has_option(self.k_config_remote_ref): + ref = Head(self.repo, Head.to_full_path(reader.get_value(self.k_config_remote_ref))) + remote_refpath = RemoteReference.to_full_path(join_path(reader.get_value(self.k_config_remote), ref.name)) + return RemoteReference(self.repo, remote_refpath) + # END handle have tracking branch + + # we are not a tracking branch + return None def rename(self, new_path, force=False): """Rename self to a new path @@ -800,6 +845,29 @@ class Head(Reference): self.repo.git.checkout(self, **kwargs) return self.repo.active_branch + #{ Configruation + + def _config_parser(self, read_only): + if read_only: + parser = self.repo.config_reader() + else: + parser = self.repo.config_writer() + # END handle parser instance + + return SectionConstraint(parser, 'branch "%s"' % self.name) + + def config_reader(self): + """:return: A configuration parser instance constrained to only read + this instance's values""" + return self._config_parser(read_only=True) + + def config_writer(self): + """:return: A configuration writer instance with read-and write acccess + to options of this head""" + return self._config_parser(read_only=False) + + #} END configuration + class TagReference(Reference): """Class representing a lightweight tag reference which either points to a commit @@ -893,6 +961,16 @@ class RemoteReference(Head): """Represents a reference pointing to a remote head.""" _common_path_default = "refs/remotes" + + @classmethod + def iter_items(cls, repo, common_path = None, remote=None): + """Iterate remote references, and if given, constrain them to the given remote""" + common_path = common_path or cls._common_path_default + if remote is not None: + common_path = join_path(common_path, str(remote)) + # END handle remote constraint + return super(RemoteReference, cls).iter_items(repo, common_path) + @property def remote_name(self): """ diff --git a/lib/git/remote.py b/lib/git/remote.py index 135e37d7..5124c603 100644 --- a/lib/git/remote.py +++ b/lib/git/remote.py @@ -468,11 +468,7 @@ class Remote(LazyMixin, Iterable): you to omit the remote path portion, i.e.:: remote.refs.master # yields RemoteReference('/refs/remotes/origin/master')""" out_refs = IterableList(RemoteReference._id_attribute_, "%s/" % self.name) - for ref in RemoteReference.list_items(self.repo): - if ref.remote_name == self.name: - out_refs.append(ref) - # END if names match - # END for each ref + out_refs.extend(RemoteReference.list_items(self.repo, remote=self.name)) assert out_refs, "Remote %s did not have any references" % self.name return out_refs -- cgit v1.2.1 From 21b4db556619db2ef25f0e0d90fef7e38e6713e5 Mon Sep 17 00:00:00 2001 From: Sebastian Thiel Date: Tue, 16 Nov 2010 11:27:23 +0100 Subject: Improved efficiency of the submodule.update process, improved test --- lib/git/objects/submodule.py | 46 ++++++++++++++++++++++++++------------------ 1 file changed, 27 insertions(+), 19 deletions(-) (limited to 'lib/git') diff --git a/lib/git/objects/submodule.py b/lib/git/objects/submodule.py index d58e07a9..72ab6360 100644 --- a/lib/git/objects/submodule.py +++ b/lib/git/objects/submodule.py @@ -144,6 +144,9 @@ class Submodule(base.IndexObject, Iterable, Traversable): """Hash this instance using its logical id, not the sha""" return hash(self._name) + def __str__(self): + return self._name + @classmethod def _config_parser(cls, repo, parent_commit, read_only): """:return: Config Parser constrained to our submodule in read or write mode @@ -250,20 +253,24 @@ class Submodule(base.IndexObject, Iterable, Traversable): remote_branch = mrepo.remotes.origin.refs[self.branch] local_branch = git.Head(mrepo, git.Head.to_full_path(self.branch)) if not local_branch.is_valid(): - mrepo.git.checkout(remote_branch, b=self.branch) - else: - # have a valid branch, but no checkout - make sure we can figure - # that out by marking the commit with a null_sha - # have to write it directly as .commit = NULLSHA tries to resolve the sha - ref = mrepo.head.ref - refpath = join_path_native(mrepo.git_dir, ref.to_full_path(ref.path)) - refdir = os.path.dirname(refpath) - if not os.path.isdir(refdir): - os.makedirs(refdir) - #END handle directory - open(refpath, 'w').write(self.NULL_HEX_SHA) + # Setup a tracking configuration - branch doesn't need to + # exist to do that + local_branch.set_tracking_branch(remote_branch) + #END handle local branch + + # have a valid branch, but no checkout - make sure we can figure + # that out by marking the commit with a null_sha + # have to write it directly as .commit = NULLSHA tries to resolve the sha + # This will bring the branch into existance + refpath = join_path_native(mrepo.git_dir, local_branch.path) + refdir = os.path.dirname(refpath) + if not os.path.isdir(refdir): + os.makedirs(refdir) + #END handle directory + open(refpath, 'w').write(self.NULL_HEX_SHA) # END initial checkout + branch creation - # make sure we are not detached + + # make sure HEAD is not detached mrepo.head.ref = local_branch except IndexError: print >> sys.stderr, "Warning: Failed to checkout tracking branch %s" % self.branch @@ -280,13 +287,14 @@ class Submodule(base.IndexObject, Iterable, Traversable): # branch - this should be prevented when setting the branch option mrepo.head.reset(self.hexsha, index=True, working_tree=True) # END handle checkout - - if recursive: - for submodule in self.iter_items(self.module()): - submodule.update(recursive, init) - # END handle recursive update - # END for each submodule # END update to new commit only if needed + + # HANDLE RECURSION + if recursive: + for submodule in self.iter_items(self.module()): + submodule.update(recursive, init) + # END handle recursive update + # END for each submodule return self -- cgit v1.2.1 From 78d2cd65b8b778f3b0cfef5268b0684314ca22ef Mon Sep 17 00:00:00 2001 From: Sebastian Thiel Date: Tue, 16 Nov 2010 12:11:29 +0100 Subject: implemented update to_last_revision option including test. Its now possible to update submodules such as svn-externals --- lib/git/objects/submodule.py | 43 +++++++++++++++++++++++++++++++++++++------ 1 file changed, 37 insertions(+), 6 deletions(-) (limited to 'lib/git') diff --git a/lib/git/objects/submodule.py b/lib/git/objects/submodule.py index 72ab6360..116c53f1 100644 --- a/lib/git/objects/submodule.py +++ b/lib/git/objects/submodule.py @@ -211,18 +211,26 @@ class Submodule(base.IndexObject, Iterable, Traversable): :param skip_init: if True, the new repository will not be cloned to its location. :return: The newly created submodule instance""" - def update(self, recursive=False, init=True): + def update(self, recursive=False, init=True, to_latest_revision=False): """Update the repository of this submodule to point to the checkout we point at with the binsha of this instance. :param recursive: if True, we will operate recursively and update child- modules as well. :param init: if True, the module repository will be cloned into place if necessary + :param to_latest_revision: if True, the submodule's sha will be ignored during checkout. + Instead, the remote will be fetched, and the local tracking branch updated. + This only works if we have a local tracking branch, which is the case + if the remote repository had a master branch, or of the 'branch' option + was specified for this submodule and the branch existed remotely :note: does nothing in bare repositories :return: self""" if self.repo.bare: return self #END pass in bare mode + + # ASSURE REPO IS PRESENT AND UPTODATE + ##################################### try: mrepo = self.module() for remote in mrepo.remotes: @@ -277,22 +285,45 @@ class Submodule(base.IndexObject, Iterable, Traversable): #END handle tracking branch #END handle initalization + + # DETERMINE SHAS TO CHECKOUT + ############################ + binsha = self.binsha + hexsha = self.hexsha + is_detached = mrepo.head.is_detached + if to_latest_revision: + msg_base = "Cannot update to latest revision in repository at %r as " % mrepo.working_dir + if not is_detached: + rref = mrepo.head.ref.tracking_branch() + if rref is not None: + rcommit = rref.commit + binsha = rcommit.binsha + hexsha = rcommit.hexsha + else: + print >> sys.stderr, "%s a tracking branch was not set for local branch '%s'" % (msg_base, mrepo.head.ref) + # END handle remote ref + else: + print >> sys.stderr, "%s there was no local tracking branch" % msg_base + # END handle detached head + # END handle to_latest_revision option + # update the working tree - if mrepo.head.commit.binsha != self.binsha: - if mrepo.head.is_detached: - mrepo.git.checkout(self.hexsha) + if mrepo.head.commit.binsha != binsha: + if is_detached: + mrepo.git.checkout(hexsha) else: # TODO: allow to specify a rebase, merge, or reset # TODO: Warn if the hexsha forces the tracking branch off the remote # branch - this should be prevented when setting the branch option - mrepo.head.reset(self.hexsha, index=True, working_tree=True) + mrepo.head.reset(hexsha, index=True, working_tree=True) # END handle checkout # END update to new commit only if needed # HANDLE RECURSION + ################## if recursive: for submodule in self.iter_items(self.module()): - submodule.update(recursive, init) + submodule.update(recursive, init, to_latest_revision) # END handle recursive update # END for each submodule -- cgit v1.2.1 From 3d061a1a506b71234f783628ba54a7bdf79bbce9 Mon Sep 17 00:00:00 2001 From: Sebastian Thiel Date: Tue, 16 Nov 2010 17:16:44 +0100 Subject: Implemented deletion of submodules including proper tests --- lib/git/objects/submodule.py | 126 +++++++++++++++++++++++++++++++++++++++++-- 1 file changed, 123 insertions(+), 3 deletions(-) (limited to 'lib/git') diff --git a/lib/git/objects/submodule.py b/lib/git/objects/submodule.py index 116c53f1..9e8abbd4 100644 --- a/lib/git/objects/submodule.py +++ b/lib/git/objects/submodule.py @@ -10,14 +10,15 @@ import stat import os import sys import weakref +import shutil __all__ = ("Submodule", "RootModule") #{ Utilities -def sm_section(path): +def sm_section(name): """:return: section title used in .gitmodules configuration file""" - return 'submodule "%s"' % path + return 'submodule "%s"' % name def sm_name(section): """:return: name of the submodule as parsed from the section name""" @@ -223,6 +224,7 @@ class Submodule(base.IndexObject, Iterable, Traversable): if the remote repository had a master branch, or of the 'branch' option was specified for this submodule and the branch existed remotely :note: does nothing in bare repositories + :note: method is definitely not atomic if recurisve is True :return: self""" if self.repo.bare: return self @@ -329,6 +331,111 @@ class Submodule(base.IndexObject, Iterable, Traversable): return self + def remove(self, module=True, force=False, configuration=True, dry_run=False): + """Remove this submodule from the repository. This will remove our entry + from the .gitmodules file and the entry in the .git/config file. + :param module: If True, the module we point to will be deleted + as well. If the module is currently on a commit which is not part + of any branch in the remote, if the currently checked out branch + is ahead of its tracking branch, if you have modifications in the + working tree, or untracked files, + In case the removal of the repository fails for these reasons, the + submodule status will not have been altered. + If this submodule has child-modules on its own, these will be deleted + prior to touching the own module. + :param force: Enforces the deletion of the module even though it contains + modifications. This basically enforces a brute-force file system based + deletion. + :param configuration: if True, the submodule is deleted from the configuration, + otherwise it isn't. Although this should be enabled most of the times, + this flag enables you to safely delete the repository of your submodule. + :param dry_run: if True, we will not actually do anything, but throw the errors + we would usually throw + :note: doesn't work in bare repositories + :raise InvalidGitRepositoryError: thrown if the repository cannot be deleted + :raise OSError: if directories or files could not be removed""" + if self.repo.bare: + raise InvalidGitRepositoryError("Cannot delete a submodule in bare repository") + # END handle bare mode + + if not (module + configuration): + raise ValueError("Need to specify to delete at least the module, or the configuration") + # END handle params + + # DELETE MODULE REPOSITORY + ########################## + if module and self.module_exists(): + if force: + # take the fast lane and just delete everything in our module path + # TODO: If we run into permission problems, we have a highly inconsistent + # state. Delete the .git folders last, start with the submodules first + mp = self.module_path() + method = None + if os.path.islink(mp): + method = os.remove + elif os.path.isdir(mp): + method = shutil.rmtree + elif os.path.exists(mp): + raise AssertionError("Cannot forcibly delete repository as it was neither a link, nor a directory") + #END handle brutal deletion + if not dry_run: + assert method + method(mp) + #END apply deletion method + else: + # verify we may delete our module + mod = self.module() + if mod.is_dirty(untracked_files=True): + raise InvalidGitRepositoryError("Cannot delete module at %s with any modifications, unless force is specified" % mod.working_tree_dir) + # END check for dirt + + # figure out whether we have new commits compared to the remotes + # NOTE: If the user pulled all the time, the remote heads might + # not have been updated, so commits coming from the remote look + # as if they come from us. But we stay strictly read-only and + # don't fetch beforhand. + for remote in mod.remotes: + num_branches_with_new_commits = 0 + rrefs = remote.refs + for rref in rrefs: + num_branches_with_new_commits = len(mod.git.cherry(rref)) != 0 + # END for each remote ref + # not a single remote branch contained all our commits + if num_branches_with_new_commits == len(rrefs): + raise InvalidGitRepositoryError("Cannot delete module at %s as there are new commits" % mod.working_tree_dir) + # END handle new commits + # END for each remote + + # gently remove all submodule repositories + for sm in self.children(): + sm.remove(module=True, force=False, configuration=False, dry_run=dry_run) + # END for each child-submodule + + # finally delete our own submodule + if not dry_run: + shutil.rmtree(mod.working_tree_dir) + # END delete tree if possible + # END handle force + # END handle module deletion + + # DELETE CONFIGURATION + ###################### + if configuration and not dry_run: + # first the index-entry + index = self.repo.index + try: + del(index.entries[index.entry_key(self.path, 0)]) + except KeyError: + pass + #END delete entry + index.write() + + # now git config - need the config intact, otherwise we can't query + # inforamtion anymore + self.repo.config_writer().remove_section(sm_section(self.name)) + self.config_writer().remove_section() + # END delete configuration + def set_parent_commit(self, commit, check=True): """Set this instance to use the given commit whose tree is supposed to contain the .gitmodules blob. @@ -410,10 +517,23 @@ class Submodule(base.IndexObject, Iterable, Traversable): try: self.module() return True - except InvalidGitRepositoryError: + except Exception: return False # END handle exception + def exists(self): + """:return: True if the submodule exists, False otherwise. Please note that + a submodule may exist (in the .gitmodules file) even though its module + doesn't exist""" + self._clear_cache() + try: + self.path + return True + except Exception: + # we raise if the path cannot be restored from configuration + return False + # END handle exceptions + @property def branch(self): """:return: The branch name that we are to checkout""" -- cgit v1.2.1 From 98e6edb546116cd98abdc3b37c6744e859bbde5c Mon Sep 17 00:00:00 2001 From: Sebastian Thiel Date: Tue, 16 Nov 2010 19:01:27 +0100 Subject: Initial implementation of submodule.add without any tests. These are to come next --- lib/git/index/base.py | 8 ++-- lib/git/objects/submodule.py | 94 ++++++++++++++++++++++++++++++++++++++++++-- 2 files changed, 95 insertions(+), 7 deletions(-) (limited to 'lib/git') diff --git a/lib/git/index/base.py b/lib/git/index/base.py index 05501ba1..a982d5c5 100644 --- a/lib/git/index/base.py +++ b/lib/git/index/base.py @@ -35,7 +35,8 @@ from git.exc import ( ) from git.objects import ( - Blob, + Blob, + Submodule, Tree, Object, Commit, @@ -553,7 +554,7 @@ class IndexFile(LazyMixin, diff.Diffable, Serializable): for item in items: if isinstance(item, basestring): paths.append(self._to_relative_path(item)) - elif isinstance(item, Blob): + elif isinstance(item, (Blob, Submodule)): entries.append(BaseIndexEntry.from_blob(item)) elif isinstance(item, BaseIndexEntry): entries.append(item) @@ -588,7 +589,7 @@ class IndexFile(LazyMixin, diff.Diffable, Serializable): They are added at stage 0 - - Blob object + - Blob or Submodule object Blobs are added as they are assuming a valid mode is set. The file they refer to may or may not exist in the file system, but must be a path relative to our repository. @@ -612,6 +613,7 @@ class IndexFile(LazyMixin, diff.Diffable, Serializable): explicitly set. Please note that Index Entries require binary sha's. :param force: + **CURRENTLY INEFFECTIVE** If True, otherwise ignored or excluded files will be added anyway. As opposed to the git-add command, we enable this flag by default diff --git a/lib/git/objects/submodule.py b/lib/git/objects/submodule.py index 9e8abbd4..93d47999 100644 --- a/lib/git/objects/submodule.py +++ b/lib/git/objects/submodule.py @@ -202,15 +202,101 @@ class Submodule(base.IndexObject, Iterable, Traversable): #{ Edit Interface @classmethod - def add(cls, repo, path, url, skip_init=False): + def add(cls, repo, name, path, url=None, branch=k_head_default, no_checkout=False): """Add a new submodule to the given repository. This will alter the index as well as the .gitmodules file, but will not create a new commit. + If the submodule already exists, no matter if the configuration differs + from the one provided, the existing submodule will be returned. :param repo: Repository instance which should receive the submodule - :param path: repository-relative path at which the submodule should be located + :param name: The name/identifier for the submodule + :param path: repository-relative or absolute path at which the submodule + should be located It will be created as required during the repository initialization. :param url: git-clone compatible URL, see git-clone reference for more information - :param skip_init: if True, the new repository will not be cloned to its location. - :return: The newly created submodule instance""" + If None, the repository is assumed to exist, and the url of the first + remote is taken instead. This is useful if you want to make an existing + repository a submodule of anotherone. + :param branch: branch at which the submodule should (later) be checked out. + The given branch must exist in the remote repository, and will be checked + out locally as a tracking branch. + It will only be written into the configuration if it differs from the + default. + :param no_checkout: if True, and if the repository has to be cloned manually, + no checkout will be performed + :return: The newly created submodule instance + :note: works atomically, such that no change will be done if the repository + update fails for instance""" + if repo.bare: + raise InvalidGitRepositoryError("Cannot add a submodule to bare repositories") + #END handle bare mode + + path = to_native_path_linux(path) + if path.endswith('/'): + path = path[:-1] + # END handle trailing slash + + sm = cls(repo, cls.NULL_BIN_SHA, cls.k_def_mode, path, name) + if sm.exists(): + # reretrieve submodule from tree + return repo.head.commit.tree[path] + # END handle existing + + branch = Head(repo, head.to_full_path(branch)) + has_module = sm.module_exists() + branch_is_default = branch.name == cls.k_head_default + if has_module and url is not None: + if url not in [r.url for r in sm.module().remotes]: + raise ValueError("Specified URL %s does not match any remote url of the repository at %s" % (url, sm.module_path())) + # END check url + # END verify urls match + + mrepo = None + if url is None: + if not has_module: + raise ValueError("A URL was not given and existing repository did not exsit at %s" % path) + # END check url + mrepo = sm.module() + urls = [r.url for r in mrepo.remotes] + if not urls: + raise ValueError("Didn't find any remote url in repository at %s" % sm.module_path()) + # END verify we have url + url = urls[0] + else: + # clone new repo + kwargs = {'n' : no_checkout} + if branch_is_default: + kwargs['b'] = str(branch) + # END setup checkout-branch + mrepo = git.Repo.clone_from(url, path, **kwargs) + # END verify url + + # update configuration and index + writer = sm.config_writer() + writer.set_value('url', url) + writer.set_value('path', path) + + sm._url = url + if not branch_is_default: + # store full path + writer.set_value(cls.k_head_option, branch.path) + sm._branch = branch + # END handle path + del(writer) + + # NOTE: Have to write the repo config file as well, otherwise + # the default implementation will be offended and not update the repository + # Maybe this is a good way to assure it doesn't get into our way, but + # we want to stay backwards compatible too ... . Its so redundant ! + repo.config_writer().set_value(sm_section(sm.name), 'url', url) + + # we deliberatly assume that our head matches our index ! + pcommit = repo.head.commit + sm._parent_commit = pcommit + sm.binsha = mrepo.head.commit.binsha + repo.index.add([sm], write=True) + + return sm + def update(self, recursive=False, init=True, to_latest_revision=False): """Update the repository of this submodule to point to the checkout -- cgit v1.2.1 From 33964afb47ce3af8a32e6613b0834e5f94bdfe68 Mon Sep 17 00:00:00 2001 From: Sebastian Thiel Date: Tue, 16 Nov 2010 21:07:33 +0100 Subject: Added tests for all failure modes of submodule add ( except for one ), and fixed a few issues on the way --- lib/git/objects/submodule.py | 40 +++++++++++++++++++++++++++++----------- 1 file changed, 29 insertions(+), 11 deletions(-) (limited to 'lib/git') diff --git a/lib/git/objects/submodule.py b/lib/git/objects/submodule.py index 93d47999..44dc9b02 100644 --- a/lib/git/objects/submodule.py +++ b/lib/git/objects/submodule.py @@ -1,11 +1,11 @@ import base from util import Traversable from StringIO import StringIO # need a dict to set bloody .name field -from git.util import Iterable +from git.util import Iterable, join_path_native, to_native_path_linux from git.config import GitConfigParser, SectionConstraint -from git.util import join_path_native from git.exc import InvalidGitRepositoryError, NoSuchPathError import stat +import git import os import sys @@ -87,6 +87,7 @@ class Submodule(base.IndexObject, Iterable, Traversable): type = 'submodule' __slots__ = ('_parent_commit', '_url', '_branch', '_name', '__weakref__') + _cache_attrs = ('path', '_url', '_branch') def __init__(self, repo, binsha, mode=None, path=None, name = None, parent_commit=None, url=None, branch=None): """Initialize this instance with its attributes. We only document the ones @@ -178,7 +179,7 @@ class Submodule(base.IndexObject, Iterable, Traversable): def _clear_cache(self): # clear the possibly changed values - for name in ('path', '_branch', '_url'): + for name in self._cache_attrs: try: delattr(self, name) except AttributeError: @@ -235,18 +236,19 @@ class Submodule(base.IndexObject, Iterable, Traversable): path = path[:-1] # END handle trailing slash + # INSTANTIATE INTERMEDIATE SM sm = cls(repo, cls.NULL_BIN_SHA, cls.k_def_mode, path, name) if sm.exists(): # reretrieve submodule from tree return repo.head.commit.tree[path] # END handle existing - branch = Head(repo, head.to_full_path(branch)) + branch = git.Head(repo, git.Head.to_full_path(branch)) has_module = sm.module_exists() branch_is_default = branch.name == cls.k_head_default if has_module and url is not None: if url not in [r.url for r in sm.module().remotes]: - raise ValueError("Specified URL %s does not match any remote url of the repository at %s" % (url, sm.module_path())) + raise ValueError("Specified URL '%s' does not match any remote url of the repository at '%s'" % (url, sm.module_path())) # END check url # END verify urls match @@ -611,14 +613,30 @@ class Submodule(base.IndexObject, Iterable, Traversable): """:return: True if the submodule exists, False otherwise. Please note that a submodule may exist (in the .gitmodules file) even though its module doesn't exist""" + # keep attributes for later, and restore them if we have no valid data + # this way we do not actually alter the state of the object + loc = locals() + for attr in self._cache_attrs: + if hasattr(self, attr): + loc[attr] = getattr(self, attr) + # END if we have the attribute cache + #END for each attr self._clear_cache() + try: - self.path - return True - except Exception: - # we raise if the path cannot be restored from configuration - return False - # END handle exceptions + try: + self.path + return True + except Exception: + return False + # END handle exceptions + finally: + for attr in self._cache_attrs: + if attr in loc: + setattr(self, attr, loc[attr]) + # END if we have a cache + # END reapply each attribute + # END handle object state consistency @property def branch(self): -- cgit v1.2.1 From 7b3ef45167e1c2f7d1b7507c13fcedd914f87da9 Mon Sep 17 00:00:00 2001 From: Sebastian Thiel Date: Tue, 16 Nov 2010 21:21:09 +0100 Subject: The submodule's branch is now a branch instance, not a plain string anymore --- lib/git/objects/submodule.py | 42 ++++++++++++++++++++++++------------------ 1 file changed, 24 insertions(+), 18 deletions(-) (limited to 'lib/git') diff --git a/lib/git/objects/submodule.py b/lib/git/objects/submodule.py index 44dc9b02..586ebeab 100644 --- a/lib/git/objects/submodule.py +++ b/lib/git/objects/submodule.py @@ -24,6 +24,10 @@ def sm_name(section): """:return: name of the submodule as parsed from the section name""" section = section.strip() return section[11:-1] + +def mkhead(repo, path): + """:return: New branch/head instance""" + return git.Head(repo, git.Head.to_full_path(path)) #} END utilities @@ -96,13 +100,14 @@ class Submodule(base.IndexObject, Iterable, Traversable): :param binsha: binary sha referring to a commit in the remote repository, see url parameter :param parent_commit: see set_parent_commit() :param url: The url to the remote repository which is the submodule - :param ref: Reference to checkout when cloning the remote repository""" + :param branch: Head instance to checkout when cloning the remote repository""" super(Submodule, self).__init__(repo, binsha, mode, path) if parent_commit is not None: self._parent_commit = parent_commit if url is not None: self._url = url if branch is not None: + assert isinstance(branch, git.Head) self._branch = branch if name is not None: self._name = name @@ -119,7 +124,7 @@ class Submodule(base.IndexObject, Iterable, Traversable): self.path = reader.get_value('path') self._url = reader.get_value('url') # git-python extension values - optional - self._branch = reader.get_value(self.k_head_option, self.k_head_default) + self._branch = mkhead(self.repo, reader.get_value(self.k_head_option, self.k_head_default)) elif attr == '_name': raise AttributeError("Cannot retrieve the name of a submodule if it was not set initially") else: @@ -203,7 +208,7 @@ class Submodule(base.IndexObject, Iterable, Traversable): #{ Edit Interface @classmethod - def add(cls, repo, name, path, url=None, branch=k_head_default, no_checkout=False): + def add(cls, repo, name, path, url=None, branch=None, no_checkout=False): """Add a new submodule to the given repository. This will alter the index as well as the .gitmodules file, but will not create a new commit. If the submodule already exists, no matter if the configuration differs @@ -220,8 +225,10 @@ class Submodule(base.IndexObject, Iterable, Traversable): :param branch: branch at which the submodule should (later) be checked out. The given branch must exist in the remote repository, and will be checked out locally as a tracking branch. - It will only be written into the configuration if it differs from the - default. + It will only be written into the configuration if it not None, which is + when the checked out branch will be the one the remote HEAD pointed to. + The result you get in these situation is somewhat fuzzy, and it is recommended + to specify at least 'master' here :param no_checkout: if True, and if the repository has to be cloned manually, no checkout will be performed :return: The newly created submodule instance @@ -243,9 +250,9 @@ class Submodule(base.IndexObject, Iterable, Traversable): return repo.head.commit.tree[path] # END handle existing - branch = git.Head(repo, git.Head.to_full_path(branch)) + br = mkhead(repo, branch or cls.k_head_default) has_module = sm.module_exists() - branch_is_default = branch.name == cls.k_head_default + branch_is_default = branch is None if has_module and url is not None: if url not in [r.url for r in sm.module().remotes]: raise ValueError("Specified URL '%s' does not match any remote url of the repository at '%s'" % (url, sm.module_path())) @@ -266,8 +273,8 @@ class Submodule(base.IndexObject, Iterable, Traversable): else: # clone new repo kwargs = {'n' : no_checkout} - if branch_is_default: - kwargs['b'] = str(branch) + if not branch_is_default: + kwargs['b'] = str(br) # END setup checkout-branch mrepo = git.Repo.clone_from(url, path, **kwargs) # END verify url @@ -280,8 +287,8 @@ class Submodule(base.IndexObject, Iterable, Traversable): sm._url = url if not branch_is_default: # store full path - writer.set_value(cls.k_head_option, branch.path) - sm._branch = branch + writer.set_value(cls.k_head_option, br.path) + sm._branch = br.path # END handle path del(writer) @@ -348,8 +355,8 @@ class Submodule(base.IndexObject, Iterable, Traversable): # see whether we have a valid branch to checkout try: - remote_branch = mrepo.remotes.origin.refs[self.branch] - local_branch = git.Head(mrepo, git.Head.to_full_path(self.branch)) + remote_branch = mrepo.remotes.origin.refs[self.branch.name] + local_branch = self.branch if not local_branch.is_valid(): # Setup a tracking configuration - branch doesn't need to # exist to do that @@ -578,7 +585,6 @@ class Submodule(base.IndexObject, Iterable, Traversable): :raise InvalidGitRepositoryError: if a repository was not available. This could also mean that it was not yet initialized""" # late import to workaround circular dependencies - from git.repo import Repo if self.repo.bare: raise InvalidGitRepositoryError("Cannot retrieve module repository in bare parent repositories") @@ -586,7 +592,7 @@ class Submodule(base.IndexObject, Iterable, Traversable): module_path = self.module_path() try: - repo = Repo(module_path) + repo = git.Repo(module_path) if repo != self.repo: return repo # END handle repo uninitialized @@ -640,7 +646,7 @@ class Submodule(base.IndexObject, Iterable, Traversable): @property def branch(self): - """:return: The branch name that we are to checkout""" + """:return: The branch instance that we are to checkout""" return self._branch @property @@ -715,7 +721,7 @@ class Submodule(base.IndexObject, Iterable, Traversable): # fill in remaining info - saves time as it doesn't have to be parsed again sm._name = n sm._parent_commit = pc - sm._branch = b + sm._branch = mkhead(repo, b) sm._url = u yield sm @@ -742,7 +748,7 @@ class RootModule(Submodule): name = self.k_root_name, parent_commit = repo.head.commit, url = '', - branch = self.k_head_default + branch = mkhead(repo, self.k_head_default) ) -- cgit v1.2.1 From ef48ca5f54fe31536920ec4171596ff8468db5fe Mon Sep 17 00:00:00 2001 From: Sebastian Thiel Date: Wed, 17 Nov 2010 00:28:57 +0100 Subject: Added rest of submodule.add test code which should be pretty much 100% coverage for it --- lib/git/objects/submodule.py | 15 +++++++++++---- 1 file changed, 11 insertions(+), 4 deletions(-) (limited to 'lib/git') diff --git a/lib/git/objects/submodule.py b/lib/git/objects/submodule.py index 586ebeab..e07117a6 100644 --- a/lib/git/objects/submodule.py +++ b/lib/git/objects/submodule.py @@ -85,7 +85,7 @@ class Submodule(base.IndexObject, Iterable, Traversable): k_modules_file = '.gitmodules' k_head_option = 'branch' k_head_default = 'master' - k_def_mode = stat.S_IFDIR | stat.S_IFLNK # submodules are directories with link-status + k_default_mode = stat.S_IFDIR | stat.S_IFLNK # submodules are directories with link-status # this is a bogus type for base class compatability type = 'submodule' @@ -244,7 +244,7 @@ class Submodule(base.IndexObject, Iterable, Traversable): # END handle trailing slash # INSTANTIATE INTERMEDIATE SM - sm = cls(repo, cls.NULL_BIN_SHA, cls.k_def_mode, path, name) + sm = cls(repo, cls.NULL_BIN_SHA, cls.k_default_mode, path, name) if sm.exists(): # reretrieve submodule from tree return repo.head.commit.tree[path] @@ -712,10 +712,17 @@ class Submodule(base.IndexObject, Iterable, Traversable): # END handle optional information # get the binsha + index = repo.index try: sm = rt[p] except KeyError: - raise InvalidGitRepositoryError("Gitmodule path %r did not exist in revision of parent commit %s" % (p, parent_commit)) + # try the index, maybe it was just added + try: + entry = index.entries[index.entry_key(p, 0)] + sm = cls(repo, entry.binsha, entry.mode, entry.path) + except KeyError: + raise InvalidGitRepositoryError("Gitmodule path %r did not exist in revision of parent commit %s" % (p, parent_commit)) + # END handle keyerror # END handle critical error # fill in remaining info - saves time as it doesn't have to be parsed again @@ -743,7 +750,7 @@ class RootModule(Submodule): super(RootModule, self).__init__( repo, binsha = self.NULL_BIN_SHA, - mode = self.k_def_mode, + mode = self.k_default_mode, path = '', name = self.k_root_name, parent_commit = repo.head.commit, -- cgit v1.2.1 From e84d05f4bbf7090a9802e9cd198d1c383974cb12 Mon Sep 17 00:00:00 2001 From: Sebastian Thiel Date: Wed, 17 Nov 2010 11:17:05 +0100 Subject: Repo: scetched out submodule_update --- lib/git/repo/base.py | 63 +++++++++++++++++++++++++++++++++++++++++++++++++++- 1 file changed, 62 insertions(+), 1 deletion(-) (limited to 'lib/git') diff --git a/lib/git/repo/base.py b/lib/git/repo/base.py index 0355b062..d1a41f3a 100644 --- a/lib/git/repo/base.py +++ b/lib/git/repo/base.py @@ -222,6 +222,8 @@ class Repo(object): :raise ValueError: if no remote with such a name exists""" return Remote(self, name) + #{ Submodules + @property def submodules(self): """:return: git.IterableList(Submodule, ...) of direct submodules""" @@ -240,7 +242,66 @@ class Repo(object): """An iterator yielding Submodule instances, see Traversable interface for a description of args and kwargs :return: Iterator""" - return RootModule(self).traverse(*args, **kwargs) + return RootModule(self).traverse(*args, **kwargs) + + def submodule_update(self, previous_commit=None, force_remove=False, to_latest_revision=False): + """Update the submodules of this repository to the current HEAD commit. + This method behaves smartly by determining changes of the path of a submodules + repository, next to changes to the to-be-checked-out commit or the branch to be + checked out. This works if the submodules ID does not change. + Additionally it will detect addition and removal of submodules, which will be handled + gracefully. + + :param previous_commit: If set to a commit'ish, the commit we should use + as the previous commit the HEAD pointed to before it was set to the commit it points to now. + If None, it defaults to ORIG_HEAD otherwise, or the parent of the current + commit if it is not given + :param force_remove: If submodules have been deleted, they will be forcibly removed. + Otherwise the update may fail if a submodule's repository cannot be deleted as + changes have been made to it (see Submodule.update() for more information) + :param to_latest_revision: If True, instead of checking out the revision pointed to + by this submodule's sha, the checked out tracking branch will be merged with the + newest remote branch fetched from the repository's origin""" + if self.bare: + raise InvalidGitRepositoryError("Cannot update submodules in bare repositories") + # END handle bare + + # HANDLE COMMITS + ################## + cur_commit = self.head.commit + if previous_commit is None: + symref = SymbolicReference(self, SymbolicReference.to_full_path('ORIG_HEAD')) + try: + previous_commit = symref.commit + except Exception: + pcommits = cur_commit.parents + if pcommits: + previous_commit = pcommits[0] + else: + # in this special case, we just diff against ourselve, which + # means exactly no change + previous_commit = cur_commit + # END handle initial commit + # END no ORIG_HEAD + else: + previous_commit = self.commit(previous_commit) # obtain commit object + # END handle previous commit + + sms = self.submodules() + + # HANDLE REMOVALS + + # HANDLE PATH RENAMES + + # FINALLY UPDATE ALL ACTUAL SUBMODULES + ########################################## + if previous_commit == cur_commit: + for sm in sms: + sm.update(recursive=True, init=True, to_latest_revision=to_latest_revision) + # END for each submodule to update + # END handle commits are equal + + #}END submodules @property def tags(self): -- cgit v1.2.1 From b03933057df80ea9f860cc616eb7733f140f866e Mon Sep 17 00:00:00 2001 From: Sebastian Thiel Date: Wed, 17 Nov 2010 14:52:01 +0100 Subject: index: Sped up reading and writing of the index file by reducing the amount of attribute lookups considerably --- lib/git/index/base.py | 4 +-- lib/git/index/fun.py | 68 ++++++++++++++++++++++++--------------------------- 2 files changed, 34 insertions(+), 38 deletions(-) (limited to 'lib/git') diff --git a/lib/git/index/base.py b/lib/git/index/base.py index a982d5c5..05caa06d 100644 --- a/lib/git/index/base.py +++ b/lib/git/index/base.py @@ -750,7 +750,7 @@ class IndexFile(LazyMixin, diff.Diffable, Serializable): may be absolute or relative paths, entries or blobs""" paths = list() for item in items: - if isinstance(item, (BaseIndexEntry,Blob)): + if isinstance(item, (BaseIndexEntry,(Blob, Submodule))): paths.append(self._to_relative_path(item.path)) elif isinstance(item, basestring): paths.append(self._to_relative_path(item)) @@ -777,7 +777,7 @@ class IndexFile(LazyMixin, diff.Diffable, Serializable): The path string may include globs, such as *.c. - - Blob object + - Blob Object Only the path portion is used in this case. - BaseIndexEntry or compatible type diff --git a/lib/git/index/fun.py b/lib/git/index/fun.py index 48c4fa74..b05344a8 100644 --- a/lib/git/index/fun.py +++ b/lib/git/index/fun.py @@ -53,22 +53,6 @@ def stat_mode_to_index_mode(mode): return S_IFREG | 0644 | (mode & 0100) # blobs with or without executable bit -def write_cache_entry(entry, stream): - """Write the given entry to the stream""" - beginoffset = stream.tell() - write = stream.write - write(entry[4]) # ctime - write(entry[5]) # mtime - path = entry[3] - plen = len(path) & CE_NAMEMASK # path length - assert plen == len(path), "Path %s too long to fit into index" % entry[3] - flags = plen | entry[2] - write(pack(">LLLLLL20sH", entry[6], entry[7], entry[0], - entry[8], entry[9], entry[10], entry[1], flags)) - write(path) - real_size = ((stream.tell() - beginoffset + 8) & ~7) - write("\0" * ((beginoffset + real_size) - stream.tell())) - def write_cache(entries, stream, extension_data=None, ShaStreamCls=IndexFileSHA1Writer): """Write the cache represented by entries to a stream @@ -83,15 +67,29 @@ def write_cache(entries, stream, extension_data=None, ShaStreamCls=IndexFileSHA1 a 4 byte identifier, followed by its size ( 4 bytes )""" # wrap the stream into a compatible writer stream = ShaStreamCls(stream) + + tell = stream.tell + write = stream.write # header version = 2 - stream.write("DIRC") - stream.write(pack(">LL", version, len(entries))) + write("DIRC") + write(pack(">LL", version, len(entries))) # body for entry in entries: - write_cache_entry(entry, stream) + beginoffset = tell() + write(entry[4]) # ctime + write(entry[5]) # mtime + path = entry[3] + plen = len(path) & CE_NAMEMASK # path length + assert plen == len(path), "Path %s too long to fit into index" % entry[3] + flags = plen | entry[2] + write(pack(">LLLLLL20sH", entry[6], entry[7], entry[0], + entry[8], entry[9], entry[10], entry[1], flags)) + write(path) + real_size = ((tell() - beginoffset + 8) & ~7) + write("\0" * ((beginoffset + real_size) - tell())) # END for each entry # write previously cached extensions data @@ -101,21 +99,6 @@ def write_cache(entries, stream, extension_data=None, ShaStreamCls=IndexFileSHA1 # write the sha over the content stream.write_sha() -def read_entry(stream): - """Return: One entry of the given stream""" - beginoffset = stream.tell() - read = stream.read - ctime = unpack(">8s", read(8))[0] - mtime = unpack(">8s", read(8))[0] - (dev, ino, mode, uid, gid, size, sha, flags) = \ - unpack(">LLLLLL20sH", read(20 + 4 * 6 + 2)) - path_size = flags & CE_NAMEMASK - path = read(path_size) - - real_size = ((stream.tell() - beginoffset + 8) & ~7) - data = read((beginoffset + real_size) - stream.tell()) - return IndexEntry((mode, sha, flags, path, ctime, mtime, dev, ino, uid, gid, size)) - def read_header(stream): """Return tuple(version_long, num_entries) from the given stream""" type_id = stream.read(4) @@ -147,10 +130,23 @@ def read_cache(stream): version, num_entries = read_header(stream) count = 0 entries = dict() + + read = stream.read + tell = stream.tell while count < num_entries: - entry = read_entry(stream) + beginoffset = tell() + ctime = unpack(">8s", read(8))[0] + mtime = unpack(">8s", read(8))[0] + (dev, ino, mode, uid, gid, size, sha, flags) = \ + unpack(">LLLLLL20sH", read(20 + 4 * 6 + 2)) + path_size = flags & CE_NAMEMASK + path = read(path_size) + + real_size = ((tell() - beginoffset + 8) & ~7) + data = read((beginoffset + real_size) - tell()) + entry = IndexEntry((mode, sha, flags, path, ctime, mtime, dev, ino, uid, gid, size)) # entry_key would be the method to use, but we safe the effort - entries[(entry.path, entry.stage)] = entry + entries[(path, entry.stage)] = entry count += 1 # END for each entry -- cgit v1.2.1 From a1e6234c27abf041e4c8cd1a799950e7cd9104f6 Mon Sep 17 00:00:00 2001 From: Sebastian Thiel Date: Wed, 17 Nov 2010 15:24:48 +0100 Subject: Inital implementation of Submodule.move including a very simple and to-be-improved test --- lib/git/config.py | 5 ++ lib/git/index/fun.py | 3 +- lib/git/objects/submodule.py | 145 ++++++++++++++++++++++++++++++++++++------- 3 files changed, 130 insertions(+), 23 deletions(-) (limited to 'lib/git') diff --git a/lib/git/config.py b/lib/git/config.py index 0528f318..f1a8832e 100644 --- a/lib/git/config.py +++ b/lib/git/config.py @@ -91,6 +91,11 @@ class SectionConstraint(object): as first argument""" return getattr(self._config, method)(self._section_name, *args, **kwargs) + @property + def config(self): + """return: Configparser instance we constrain""" + return self._config + class GitConfigParser(cp.RawConfigParser, object): """Implements specifics required to read git style configuration files. diff --git a/lib/git/index/fun.py b/lib/git/index/fun.py index b05344a8..87fdf1a9 100644 --- a/lib/git/index/fun.py +++ b/lib/git/index/fun.py @@ -30,6 +30,7 @@ from typ import ( CE_NAMEMASK, CE_STAGESHIFT ) +CE_NAMEMASK_INV = ~CE_NAMEMASK from util import ( pack, @@ -84,7 +85,7 @@ def write_cache(entries, stream, extension_data=None, ShaStreamCls=IndexFileSHA1 path = entry[3] plen = len(path) & CE_NAMEMASK # path length assert plen == len(path), "Path %s too long to fit into index" % entry[3] - flags = plen | entry[2] + flags = plen | (entry[2] & CE_NAMEMASK_INV) # clear possible previous values write(pack(">LLLLLL20sH", entry[6], entry[7], entry[0], entry[8], entry[9], entry[10], entry[1], flags)) write(path) diff --git a/lib/git/objects/submodule.py b/lib/git/objects/submodule.py index e07117a6..8a1ab6af 100644 --- a/lib/git/objects/submodule.py +++ b/lib/git/objects/submodule.py @@ -28,6 +28,19 @@ def sm_name(section): def mkhead(repo, path): """:return: New branch/head instance""" return git.Head(repo, git.Head.to_full_path(path)) + +def unbare_repo(func): + """Methods with this decorator raise InvalidGitRepositoryError if they + encounter a bare repository""" + def wrapper(self, *args, **kwargs): + if self.repo.bare: + raise InvalidGitRepositoryError("Method '%s' cannot operate on bare repositories" % func.__name__) + #END bare method + return func(self, *args, **kwargs) + # END wrapper + wrapper.__name__ = func.__name__ + return wrapper + #} END utilities @@ -39,10 +52,14 @@ class SubmoduleConfigParser(GitConfigParser): with the new data, if we have written into a stream. Otherwise it will add the local file to the index to make it correspond with the working tree. Additionally, the cache must be cleared + + Please note that no mutating method will work in bare mode """ def __init__(self, *args, **kwargs): self._smref = None + self._index = None + self._auto_write = True super(SubmoduleConfigParser, self).__init__(*args, **kwargs) #{ Interface @@ -59,7 +76,11 @@ class SubmoduleConfigParser(GitConfigParser): sm = self._smref() if sm is not None: - sm.repo.index.add([sm.k_modules_file]) + index = self._index + if index is None: + index = sm.repo.index + # END handle index + index.add([sm.k_modules_file], write=self._auto_write) sm._clear_cache() # END handle weakref @@ -102,6 +123,7 @@ class Submodule(base.IndexObject, Iterable, Traversable): :param url: The url to the remote repository which is the submodule :param branch: Head instance to checkout when cloning the remote repository""" super(Submodule, self).__init__(repo, binsha, mode, path) + self.size = 0 if parent_commit is not None: self._parent_commit = parent_commit if url is not None: @@ -113,9 +135,7 @@ class Submodule(base.IndexObject, Iterable, Traversable): self._name = name def _set_cache_(self, attr): - if attr == 'size': - raise ValueError("Submodules do not have a size as they do not refer to anything in this repository") - elif attr == '_parent_commit': + if attr == '_parent_commit': # set a default value, which is the root tree of the current head self._parent_commit = self.repo.commit() elif attr in ('path', '_url', '_branch'): @@ -235,8 +255,8 @@ class Submodule(base.IndexObject, Iterable, Traversable): :note: works atomically, such that no change will be done if the repository update fails for instance""" if repo.bare: - raise InvalidGitRepositoryError("Cannot add a submodule to bare repositories") - #END handle bare mode + raise InvalidGitRepositoryError("Cannot add submodules to bare repositories") + # END handle bare repos path = to_native_path_linux(path) if path.endswith('/'): @@ -280,7 +300,8 @@ class Submodule(base.IndexObject, Iterable, Traversable): # END verify url # update configuration and index - writer = sm.config_writer() + index = sm.repo.index + writer = sm.config_writer(index=index, write=False) writer.set_value('url', url) writer.set_value('path', path) @@ -302,11 +323,10 @@ class Submodule(base.IndexObject, Iterable, Traversable): pcommit = repo.head.commit sm._parent_commit = pcommit sm.binsha = mrepo.head.commit.binsha - repo.index.add([sm], write=True) + index.add([sm], write=True) return sm - def update(self, recursive=False, init=True, to_latest_revision=False): """Update the repository of this submodule to point to the checkout we point at with the binsha of this instance. @@ -426,6 +446,85 @@ class Submodule(base.IndexObject, Iterable, Traversable): return self + @unbare_repo + def move(self, module_path): + """Move the submodule to a another module path. This involves physically moving + the repository at our current path, changing the configuration, as well as + adjusting our index entry accordingly. + :param module_path: the path to which to move our module, given as + repository-relative path. Intermediate directories will be created + accordingly. If the path already exists, it must be empty. + Trailling (back)slashes are removed automatically + :return: self + :raise ValueError: if the module path existed and was not empty, or was a file + :note: Currently the method is not atomic, and it could leave the repository + in an inconsistent state if a sub-step fails for some reason + """ + module_path = to_native_path_linux(module_path) + if module_path.endswith('/'): + module_path = module_path[:-1] + # END handle trailing slash + + # VERIFY DESTINATION + if module_path == self.path: + return self + #END handle no change + + dest_path = join_path_native(self.repo.working_tree_dir, module_path) + if os.path.isfile(dest_path): + raise ValueError("Cannot move repository onto a file: %s" % dest_path) + # END handle target files + + # remove existing destination + if os.path.exists(dest_path): + if len(os.listdir(dest_path)): + raise ValueError("Destination module directory was not empty") + #END handle non-emptyness + + if os.path.islink(dest_path): + os.remove(dest_path) + else: + os.rmdir(dest_path) + #END handle link + else: + # recreate parent directories + # NOTE: renames() does that now + pass + #END handle existance + + # move the module into place if possible + cur_path = self.module_path() + if os.path.exists(cur_path): + os.renames(cur_path, dest_path) + #END move physical module + + # NOTE: from now on, we would have to undo the rename ! + + # rename the index entry - have to manipulate the index directly as + # git-mv cannot be used on submodules ... yeah + index = self.repo.index + try: + ekey = index.entry_key(self.path, 0) + entry = index.entries[ekey] + del(index.entries[ekey]) + nentry = git.IndexEntry(entry[:3]+(module_path,)+entry[4:]) + ekey = index.entry_key(module_path, 0) + index.entries[ekey] = nentry + except KeyError: + raise ValueError("Submodule's entry at %r did not exist" % (self.path)) + #END handle submodule doesn't exist + + # update configuration + writer = self.config_writer(index=index) # auto-write + writer.set_value('path', module_path) + self.path = module_path + del(writer) + + return self + + + + @unbare_repo def remove(self, module=True, force=False, configuration=True, dry_run=False): """Remove this submodule from the repository. This will remove our entry from the .gitmodules file and the entry in the .git/config file. @@ -449,10 +548,6 @@ class Submodule(base.IndexObject, Iterable, Traversable): :note: doesn't work in bare repositories :raise InvalidGitRepositoryError: thrown if the repository cannot be deleted :raise OSError: if directories or files could not be removed""" - if self.repo.bare: - raise InvalidGitRepositoryError("Cannot delete a submodule in bare repository") - # END handle bare mode - if not (module + configuration): raise ValueError("Need to specify to delete at least the module, or the configuration") # END handle params @@ -565,31 +660,37 @@ class Submodule(base.IndexObject, Iterable, Traversable): return self - def config_writer(self): + @unbare_repo + def config_writer(self, index=None, write=True): """:return: a config writer instance allowing you to read and write the data belonging to this submodule into the .gitmodules file. + :param index: if not None, an IndexFile instance which should be written. + defaults to the index of the Submodule's parent repository. + :param write: if True, the index will be written each time a configuration + value changes. + :note: the parameters allow for a more efficient writing of the index, + as you can pass in a modified index on your own, prevent automatic writing, + and write yourself once the whole operation is complete :raise ValueError: if trying to get a writer on a parent_commit which does not match the current head commit :raise IOError: If the .gitmodules file/blob could not be read""" - if self.repo.bare: - raise InvalidGitRepositoryError("Cannot change submodule configuration in a bare repository") - return self._config_parser_constrained(read_only=False) + writer = self._config_parser_constrained(read_only=False) + if index is not None: + writer.config._index = index + writer.config._auto_write = write + return writer #} END edit interface #{ Query Interface + @unbare_repo def module(self): """:return: Repo instance initialized from the repository at our submodule path :raise InvalidGitRepositoryError: if a repository was not available. This could also mean that it was not yet initialized""" # late import to workaround circular dependencies - - if self.repo.bare: - raise InvalidGitRepositoryError("Cannot retrieve module repository in bare parent repositories") - # END handle bare mode - module_path = self.module_path() try: repo = git.Repo(module_path) -- cgit v1.2.1 From 609a46a72764dc71104aa5d7b1ca5f53d4237a75 Mon Sep 17 00:00:00 2001 From: Sebastian Thiel Date: Wed, 17 Nov 2010 17:15:24 +0100 Subject: submodule: removed module_path method as it is implemented in the abspath property alrdeady Improved submodule move tests --- lib/git/objects/submodule.py | 16 +++++----------- 1 file changed, 5 insertions(+), 11 deletions(-) (limited to 'lib/git') diff --git a/lib/git/objects/submodule.py b/lib/git/objects/submodule.py index 8a1ab6af..51453820 100644 --- a/lib/git/objects/submodule.py +++ b/lib/git/objects/submodule.py @@ -275,7 +275,7 @@ class Submodule(base.IndexObject, Iterable, Traversable): branch_is_default = branch is None if has_module and url is not None: if url not in [r.url for r in sm.module().remotes]: - raise ValueError("Specified URL '%s' does not match any remote url of the repository at '%s'" % (url, sm.module_path())) + raise ValueError("Specified URL '%s' does not match any remote url of the repository at '%s'" % (url, sm.abspath)) # END check url # END verify urls match @@ -287,7 +287,7 @@ class Submodule(base.IndexObject, Iterable, Traversable): mrepo = sm.module() urls = [r.url for r in mrepo.remotes] if not urls: - raise ValueError("Didn't find any remote url in repository at %s" % sm.module_path()) + raise ValueError("Didn't find any remote url in repository at %s" % sm.abspath) # END verify we have url url = urls[0] else: @@ -493,7 +493,7 @@ class Submodule(base.IndexObject, Iterable, Traversable): #END handle existance # move the module into place if possible - cur_path = self.module_path() + cur_path = self.abspath if os.path.exists(cur_path): os.renames(cur_path, dest_path) #END move physical module @@ -522,8 +522,6 @@ class Submodule(base.IndexObject, Iterable, Traversable): return self - - @unbare_repo def remove(self, module=True, force=False, configuration=True, dry_run=False): """Remove this submodule from the repository. This will remove our entry @@ -559,7 +557,7 @@ class Submodule(base.IndexObject, Iterable, Traversable): # take the fast lane and just delete everything in our module path # TODO: If we run into permission problems, we have a highly inconsistent # state. Delete the .git folders last, start with the submodules first - mp = self.module_path() + mp = self.abspath method = None if os.path.islink(mp): method = os.remove @@ -691,7 +689,7 @@ class Submodule(base.IndexObject, Iterable, Traversable): :raise InvalidGitRepositoryError: if a repository was not available. This could also mean that it was not yet initialized""" # late import to workaround circular dependencies - module_path = self.module_path() + module_path = self.abspath try: repo = git.Repo(module_path) if repo != self.repo: @@ -703,10 +701,6 @@ class Submodule(base.IndexObject, Iterable, Traversable): raise InvalidGitRepositoryError("Repository at %r was not yet checked out" % module_path) # END handle exceptions - def module_path(self): - """:return: full path to the root of our module. It is relative to the filesystem root""" - return join_path_native(self.repo.working_tree_dir, self.path) - def module_exists(self): """:return: True if our module exists and is a valid git repository. See module() method""" try: -- cgit v1.2.1 From 7cc4d748a132377ffe63534e9777d7541a3253c5 Mon Sep 17 00:00:00 2001 From: Sebastian Thiel Date: Wed, 17 Nov 2010 21:33:33 +0100 Subject: repo: Added create_submodule method which fits into the tradition of offering a create_* method for most important entities. Moved implementation of smart update method to the RootModule implementation, where it may do special things without requiring an interface for everything --- lib/git/repo/base.py | 71 +++++++++++----------------------------------------- 1 file changed, 14 insertions(+), 57 deletions(-) (limited to 'lib/git') diff --git a/lib/git/repo/base.py b/lib/git/repo/base.py index d1a41f3a..aa00d028 100644 --- a/lib/git/repo/base.py +++ b/lib/git/repo/base.py @@ -226,7 +226,8 @@ class Repo(object): @property def submodules(self): - """:return: git.IterableList(Submodule, ...) of direct submodules""" + """:return: git.IterableList(Submodule, ...) of direct submodules + available from the current head""" return Submodule.list_items(self) def submodule(self, name): @@ -238,68 +239,24 @@ class Repo(object): raise ValueError("Didn't find submodule named %r" % name) # END exception handling + def create_submodule(self, *args, **kwargs): + """Create a new submodule + :note: See the documentation of Submodule.add for a description of the + applicable parameters + :return: created submodules""" + return Submodule.add(self, *args, **kwargs) + def iter_submodules(self, *args, **kwargs): """An iterator yielding Submodule instances, see Traversable interface for a description of args and kwargs :return: Iterator""" return RootModule(self).traverse(*args, **kwargs) - def submodule_update(self, previous_commit=None, force_remove=False, to_latest_revision=False): - """Update the submodules of this repository to the current HEAD commit. - This method behaves smartly by determining changes of the path of a submodules - repository, next to changes to the to-be-checked-out commit or the branch to be - checked out. This works if the submodules ID does not change. - Additionally it will detect addition and removal of submodules, which will be handled - gracefully. - - :param previous_commit: If set to a commit'ish, the commit we should use - as the previous commit the HEAD pointed to before it was set to the commit it points to now. - If None, it defaults to ORIG_HEAD otherwise, or the parent of the current - commit if it is not given - :param force_remove: If submodules have been deleted, they will be forcibly removed. - Otherwise the update may fail if a submodule's repository cannot be deleted as - changes have been made to it (see Submodule.update() for more information) - :param to_latest_revision: If True, instead of checking out the revision pointed to - by this submodule's sha, the checked out tracking branch will be merged with the - newest remote branch fetched from the repository's origin""" - if self.bare: - raise InvalidGitRepositoryError("Cannot update submodules in bare repositories") - # END handle bare - - # HANDLE COMMITS - ################## - cur_commit = self.head.commit - if previous_commit is None: - symref = SymbolicReference(self, SymbolicReference.to_full_path('ORIG_HEAD')) - try: - previous_commit = symref.commit - except Exception: - pcommits = cur_commit.parents - if pcommits: - previous_commit = pcommits[0] - else: - # in this special case, we just diff against ourselve, which - # means exactly no change - previous_commit = cur_commit - # END handle initial commit - # END no ORIG_HEAD - else: - previous_commit = self.commit(previous_commit) # obtain commit object - # END handle previous commit - - sms = self.submodules() - - # HANDLE REMOVALS - - # HANDLE PATH RENAMES - - # FINALLY UPDATE ALL ACTUAL SUBMODULES - ########################################## - if previous_commit == cur_commit: - for sm in sms: - sm.update(recursive=True, init=True, to_latest_revision=to_latest_revision) - # END for each submodule to update - # END handle commits are equal + def submodule_update(self, *args, **kwargs): + """Update the submodules, keeping the repository consistent as it will + take the previous state into consideration. For more information, please + see the documentation of RootModule.update""" + return RootModule(self).update(*args, **kwargs) #}END submodules -- cgit v1.2.1 From 1687283c13caf7ff8d1959591541dff6a171ca1e Mon Sep 17 00:00:00 2001 From: Sebastian Thiel Date: Wed, 17 Nov 2010 22:38:10 +0100 Subject: RootModule.update: initial implementation of update method, which should be able to handle submodule removals, additions, path changes and branch changes. All this still needs to be tested though --- lib/git/objects/submodule.py | 210 +++++++++++++++++++++++++++++++++++++++---- lib/git/util.py | 8 ++ 2 files changed, 199 insertions(+), 19 deletions(-) (limited to 'lib/git') diff --git a/lib/git/objects/submodule.py b/lib/git/objects/submodule.py index 51453820..d31f1ec9 100644 --- a/lib/git/objects/submodule.py +++ b/lib/git/objects/submodule.py @@ -41,6 +41,17 @@ def unbare_repo(func): wrapper.__name__ = func.__name__ return wrapper +def find_remote_branch(remotes, branch): + """Find the remote branch matching the name of the given branch or raise InvalidGitRepositoryError""" + for remote in remotes: + try: + return remote.refs[branch.name] + except IndexError: + continue + # END exception handling + #END for remote + raise InvalidGitRepositoryError("Didn't find remote branch %r in any of the given remotes", branch + #} END utilities @@ -375,7 +386,8 @@ class Submodule(base.IndexObject, Iterable, Traversable): # see whether we have a valid branch to checkout try: - remote_branch = mrepo.remotes.origin.refs[self.branch.name] + # find a remote which has our branch - we try to be flexible + remote_branch = find_remote_branch(mrepo.remotes, self.branch) local_branch = self.branch if not local_branch.is_valid(): # Setup a tracking configuration - branch doesn't need to @@ -447,7 +459,7 @@ class Submodule(base.IndexObject, Iterable, Traversable): return self @unbare_repo - def move(self, module_path): + def move(self, module_path, module_only=False): """Move the submodule to a another module path. This involves physically moving the repository at our current path, changing the configuration, as well as adjusting our index entry accordingly. @@ -455,6 +467,10 @@ class Submodule(base.IndexObject, Iterable, Traversable): repository-relative path. Intermediate directories will be created accordingly. If the path already exists, it must be empty. Trailling (back)slashes are removed automatically + :param module_only: if True, only the repository managed by this submodule + will be moved, not the configuration. This will effectively + leave your repository in an inconsistent state unless the configuration + and index already point to the target location. :return: self :raise ValueError: if the module path existed and was not empty, or was a file :note: Currently the method is not atomic, and it could leave the repository @@ -475,6 +491,13 @@ class Submodule(base.IndexObject, Iterable, Traversable): raise ValueError("Cannot move repository onto a file: %s" % dest_path) # END handle target files + index = self.repo.index + tekey = index.entry_key(module_path, 0) + # if the target item already exists, fail + if not module_only and tekey in index.entries: + raise ValueError("Index entry for target path did alredy exist") + #END handle index key already there + # remove existing destination if os.path.exists(dest_path): if len(os.listdir(dest_path)): @@ -502,23 +525,23 @@ class Submodule(base.IndexObject, Iterable, Traversable): # rename the index entry - have to manipulate the index directly as # git-mv cannot be used on submodules ... yeah - index = self.repo.index - try: - ekey = index.entry_key(self.path, 0) - entry = index.entries[ekey] - del(index.entries[ekey]) - nentry = git.IndexEntry(entry[:3]+(module_path,)+entry[4:]) - ekey = index.entry_key(module_path, 0) - index.entries[ekey] = nentry - except KeyError: - raise ValueError("Submodule's entry at %r did not exist" % (self.path)) - #END handle submodule doesn't exist - - # update configuration - writer = self.config_writer(index=index) # auto-write - writer.set_value('path', module_path) - self.path = module_path - del(writer) + if not module_only: + try: + ekey = index.entry_key(self.path, 0) + entry = index.entries[ekey] + del(index.entries[ekey]) + nentry = git.IndexEntry(entry[:3]+(module_path,)+entry[4:]) + index.entries[tekey] = nentry + except KeyError: + raise ValueError("Submodule's entry at %r did not exist" % (self.path)) + #END handle submodule doesn't exist + + # update configuration + writer = self.config_writer(index=index) # auto-write + writer.set_value('path', module_path) + self.path = module_path + del(writer) + # END handle module_only return self @@ -543,6 +566,7 @@ class Submodule(base.IndexObject, Iterable, Traversable): this flag enables you to safely delete the repository of your submodule. :param dry_run: if True, we will not actually do anything, but throw the errors we would usually throw + :return: self :note: doesn't work in bare repositories :raise InvalidGitRepositoryError: thrown if the repository cannot be deleted :raise OSError: if directories or files could not be removed""" @@ -624,6 +648,8 @@ class Submodule(base.IndexObject, Iterable, Traversable): self.config_writer().remove_section() # END delete configuration + return self + def set_parent_commit(self, commit, check=True): """Set this instance to use the given commit whose tree is supposed to contain the .gitmodules blob. @@ -859,6 +885,152 @@ class RootModule(Submodule): pass #{ Interface + + def update(self, previous_commit=None, recursive=True, force_remove=False, init=True, to_latest_revision=False): + """Update the submodules of this repository to the current HEAD commit. + This method behaves smartly by determining changes of the path of a submodules + repository, next to changes to the to-be-checked-out commit or the branch to be + checked out. This works if the submodules ID does not change. + Additionally it will detect addition and removal of submodules, which will be handled + gracefully. + + :param previous_commit: If set to a commit'ish, the commit we should use + as the previous commit the HEAD pointed to before it was set to the commit it points to now. + If None, it defaults to ORIG_HEAD otherwise, or the parent of the current + commit if it is not given + :param recursive: if True, the children of submodules will be updated as well + using the same technique + :param force_remove: If submodules have been deleted, they will be forcibly removed. + Otherwise the update may fail if a submodule's repository cannot be deleted as + changes have been made to it (see Submodule.update() for more information) + :param init: If we encounter a new module which would need to be initialized, then do it. + :param to_latest_revision: If True, instead of checking out the revision pointed to + by this submodule's sha, the checked out tracking branch will be merged with the + newest remote branch fetched from the repository's origin""" + if self.repo.bare: + raise InvalidGitRepositoryError("Cannot update submodules in bare repositories") + # END handle bare + + repo = self.repo + + # HANDLE COMMITS + ################## + cur_commit = repo.head.commit + if previous_commit is None: + symref = SymbolicReference(repo, SymbolicReference.to_full_path('ORIG_HEAD')) + try: + previous_commit = symref.commit + except Exception: + pcommits = cur_commit.parents + if pcommits: + previous_commit = pcommits[0] + else: + # in this special case, we just diff against ourselve, which + # means exactly no change + previous_commit = cur_commit + # END handle initial commit + # END no ORIG_HEAD + else: + previous_commit = repo.commit(previous_commit) # obtain commit object + # END handle previous commit + + + # HANDLE REMOVALS + psms = type(self).list_items(repo, parent_commit=previous_commit) + sms = self.children() + spsms = set(psms) + ssms = set(sms) + + # HANDLE REMOVALS + ################### + for rsm in (spsms - ssms): + # fake it into thinking its at the current commit to allow deletion + # of previous module. Trigger the cache to be updated before that + #rsm.url + rsm._parent_commit = repo.head.commit + rsm.remove(configuration=False, module=True, force=force_remove) + # END for each removed submodule + + # HANDLE PATH RENAMES + url changes + branch changes + for csm in (spsms & ssms): + psm = psms[csm.name] + sm = sms[csm.name] + + if sm.path != psm.path and psm.module_exists(): + # move the module to the new path + psm.move(sm.path, module_only=True) + # END handle path changes + + if sm.module_exists(): + # handle url change + if sm.url != psm.url: + # Add the new remote, remove the old one + # This way, if the url just changes, the commits will not + # have to be re-retrieved + nn = '__new_origin__' + smm = sm.module() + rmts = smm.remotes + assert nn not in rmts + smr = smm.create_remote(nn, sm.url) + srm.fetch() + + # now delete the changed one + orig_name = None + for remote in rmts: + if remote.url == psm.url: + orig_name = remote.name + smm.delete_remote(remote) + break + # END if urls match + # END for each remote + + # rename the new remote back to what it was + # if we have not found any remote with the original url + # we may not have a name. This is a special case, + # and its okay to fail her + assert orig_name is not None, "Couldn't find original remote-repo at url %r" % psm.url + smr.rename(orig_name) + # END handle url + + if sm.branch != psm.branch: + # finally, create a new tracking branch which tracks the + # new remote branch + smm = sm.module() + smmr = smm.remotes + tbr = git.Head.create(smm, sm.branch.name) + tbr.set_tracking_branch(find_remote_branch(smmr, sm.branch)) + + # figure out whether the previous tracking branch contains + # new commits compared to the other one, if not we can + # delete it. + try: + tbr = find_remote_branch(smmr, psm.branch) + if len(smm.git.cherry(tbr, psm.branch)) == 0: + psm.branch.delete(smm, psm.branch) + #END delete original tracking branch if there are no changes + except InvalidGitRepositoryError: + # ignore it if the previous branch couldn't be found in the + # current remotes, this just means we can't handle it + pass + # END exception handling + #END handle branch + #END handle + # END for each common submodule + + # FINALLY UPDATE ALL ACTUAL SUBMODULES + ########################################## + for sm in sms: + sm.update(recursive=True, init=init, to_latest_revision=to_latest_revision) + + # update recursively depth first - question is which inconsitent + # state will be better in case it fails somewhere. Defective branch + # or defective depth + if recursive: + type(cls)(sm.module()).update(recursive=True, force_remove=force_remove, + init=init, to_latest_revision=to_latest_revision) + #END handle recursive + # END for each submodule to update + def module(self): """:return: the actual repository containing the submodules""" return self.repo diff --git a/lib/git/util.py b/lib/git/util.py index b77e7904..b7446953 100644 --- a/lib/git/util.py +++ b/lib/git/util.py @@ -316,6 +316,14 @@ class IterableList(list): return getattr(self, index) except AttributeError: raise IndexError( "No item found with id %r" % (self._prefix + index) ) + + def __contains__(self, item): + try: + self[item] + return True + except IndexError: + return False + # END handle exception class Iterable(object): -- cgit v1.2.1 From 7a320abc52307b4d4010166bd899ac75024ec9a7 Mon Sep 17 00:00:00 2001 From: Sebastian Thiel Date: Thu, 18 Nov 2010 09:20:15 +0100 Subject: commit: when creating a new commit and advancing the head, it will now write the ORIG_HEAD reference as well --- lib/git/objects/commit.py | 6 ++++++ lib/git/refs.py | 2 +- 2 files changed, 7 insertions(+), 1 deletion(-) (limited to 'lib/git') diff --git a/lib/git/objects/commit.py b/lib/git/objects/commit.py index ae22fb76..1aedaabf 100644 --- a/lib/git/objects/commit.py +++ b/lib/git/objects/commit.py @@ -365,7 +365,13 @@ class Commit(base.Object, Iterable, Diffable, Traversable, Serializable): new_commit.binsha = istream.binsha if head: + # need late import here, importing git at the very beginning throws + # as well ... + import git.refs try: + cur_commit = repo.head.commit + # Adjust the original head reference - force it + git.refs.SymbolicReference.create(repo, 'ORIG_HEAD', cur_commit, force=True) repo.head.commit = new_commit except ValueError: # head is not yet set to the ref our HEAD points to diff --git a/lib/git/refs.py b/lib/git/refs.py index 39c5ff29..399c4b78 100644 --- a/lib/git/refs.py +++ b/lib/git/refs.py @@ -277,7 +277,7 @@ class SymbolicReference(object): @classmethod def to_full_path(cls, path): """ - :return: string with a full path name which can be used to initialize + :return: string with a full repository-relative path which can be used to initialize a Reference instance, for instance by using ``Reference.from_path``""" if isinstance(path, SymbolicReference): path = path.path -- cgit v1.2.1 From 82849578e61a7dfb47fc76dcbe18b1e3b6a36951 Mon Sep 17 00:00:00 2001 From: Sebastian Thiel Date: Thu, 18 Nov 2010 10:40:16 +0100 Subject: ORIG_HEAD handling is now implemented in the ref-class itself, instead of being a special case of the commit method; includes tests util: Fixed iterable lists, which broke due to an incorrectly implemented __contains__ method --- lib/git/objects/commit.py | 3 --- lib/git/refs.py | 51 +++++++++++++++++++++++++++++++++++++++++------ lib/git/util.py | 8 -------- 3 files changed, 45 insertions(+), 17 deletions(-) (limited to 'lib/git') diff --git a/lib/git/objects/commit.py b/lib/git/objects/commit.py index 1aedaabf..a2b6c554 100644 --- a/lib/git/objects/commit.py +++ b/lib/git/objects/commit.py @@ -369,9 +369,6 @@ class Commit(base.Object, Iterable, Diffable, Traversable, Serializable): # as well ... import git.refs try: - cur_commit = repo.head.commit - # Adjust the original head reference - force it - git.refs.SymbolicReference.create(repo, 'ORIG_HEAD', cur_commit, force=True) repo.head.commit = new_commit except ValueError: # head is not yet set to the ref our HEAD points to diff --git a/lib/git/refs.py b/lib/git/refs.py index 399c4b78..fcf5fd10 100644 --- a/lib/git/refs.py +++ b/lib/git/refs.py @@ -224,12 +224,30 @@ class SymbolicReference(object): # END end try string # END try commit attribute + # maintain the orig-head if we are currently checked-out + head = HEAD(self.repo) + try: + if head.ref == self: + try: + # TODO: implement this atomically, if we fail below, orig_head is at an incorrect spot + # Enforce the creation of ORIG_HEAD + SymbolicReference.create(self.repo, head.orig_head().name, self.commit, force=True) + except ValueError: + pass + #END exception handling + # END if we are checked-out + except TypeError: + pass + # END handle detached heads + # if we are writing a ref, use symbolic ref to get the reflog and more # checking - # Otherwise we detach it and have to do it manually + # Otherwise we detach it and have to do it manually. Besides, this works + # recursively automaitcally, but should be replaced with a python implementation + # soon if write_value.startswith('ref:'): self.repo.git.symbolic_ref(self.path, write_value[5:]) - return + return # END non-detached handling path = self._abs_path() @@ -243,10 +261,10 @@ class SymbolicReference(object): finally: fp.close() # END writing - - reference = property(_get_reference, _set_reference, doc="Returns the Reference we point to") - # alias + + # aliased reference + reference = property(_get_reference, _set_reference, doc="Returns the Reference we point to") ref = reference def is_valid(self): @@ -553,7 +571,6 @@ class Reference(SymbolicReference, LazyMixin, Iterable): :note: TypeChecking is done by the git command""" - # check for existence, touch it if required abs_path = self._abs_path() existed = True if not isfile(abs_path): @@ -618,6 +635,7 @@ class HEAD(SymbolicReference): """Special case of a Symbolic Reference as it represents the repository's HEAD reference.""" _HEAD_NAME = 'HEAD' + _ORIG_HEAD_NAME = 'ORIG_HEAD' __slots__ = tuple() def __init__(self, repo, path=_HEAD_NAME): @@ -625,6 +643,27 @@ class HEAD(SymbolicReference): raise ValueError("HEAD instance must point to %r, got %r" % (self._HEAD_NAME, path)) super(HEAD, self).__init__(repo, path) + def orig_head(self): + """:return: SymbolicReference pointing at the ORIG_HEAD, which is maintained + to contain the previous value of HEAD""" + return SymbolicReference(self.repo, self._ORIG_HEAD_NAME) + + def _set_reference(self, ref): + """If someone changes the reference through us, we must manually update + the ORIG_HEAD if we are detached. The underlying implementation can only + handle un-detached heads as it has to check whether the current head + is the checked-out one""" + if self.is_detached: + prev_commit = self.commit + super(HEAD, self)._set_reference(ref) + SymbolicReference.create(self.repo, self._ORIG_HEAD_NAME, prev_commit, force=True) + else: + super(HEAD, self)._set_reference(ref) + # END handle detached mode + + # aliased reference + reference = property(SymbolicReference._get_reference, _set_reference, doc="Returns the Reference we point to") + ref = reference def reset(self, commit='HEAD', index=True, working_tree = False, paths=None, **kwargs): diff --git a/lib/git/util.py b/lib/git/util.py index b7446953..c945e6a3 100644 --- a/lib/git/util.py +++ b/lib/git/util.py @@ -317,14 +317,6 @@ class IterableList(list): except AttributeError: raise IndexError( "No item found with id %r" % (self._prefix + index) ) - def __contains__(self, item): - try: - self[item] - return True - except IndexError: - return False - # END handle exception - class Iterable(object): """Defines an interface for iterable items which is to assure a uniform -- cgit v1.2.1 From 0c1834134ce177cdbd30a56994fcc4bf8f5be8b2 Mon Sep 17 00:00:00 2001 From: Sebastian Thiel Date: Thu, 18 Nov 2010 11:41:16 +0100 Subject: Added test-setup which can test all aspects of the (smart) update method --- lib/git/objects/submodule.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'lib/git') diff --git a/lib/git/objects/submodule.py b/lib/git/objects/submodule.py index d31f1ec9..7ef7f590 100644 --- a/lib/git/objects/submodule.py +++ b/lib/git/objects/submodule.py @@ -50,7 +50,7 @@ def find_remote_branch(remotes, branch): continue # END exception handling #END for remote - raise InvalidGitRepositoryError("Didn't find remote branch %r in any of the given remotes", branch + raise InvalidGitRepositoryError("Didn't find remote branch %r in any of the given remotes", branch) #} END utilities -- cgit v1.2.1 From c0990b2a6dd2e777b46c1685ddb985b3c0ef59a2 Mon Sep 17 00:00:00 2001 From: Sebastian Thiel Date: Thu, 18 Nov 2010 17:09:32 +0100 Subject: first update test succeeds, so it verifies that existing repositories can be moved later if the configuration changed, and actually it also verifies that the url-change is handled correctly (as we changed the url from the default to the local path) --- lib/git/objects/submodule.py | 162 ++++++++++++++++++++++++++----------------- 1 file changed, 100 insertions(+), 62 deletions(-) (limited to 'lib/git') diff --git a/lib/git/objects/submodule.py b/lib/git/objects/submodule.py index 7ef7f590..9fb8ce8f 100644 --- a/lib/git/objects/submodule.py +++ b/lib/git/objects/submodule.py @@ -104,6 +104,7 @@ class SubmoduleConfigParser(GitConfigParser): return rval # END overridden methods + class Submodule(base.IndexObject, Iterable, Traversable): """Implements access to a git submodule. They are special in that their sha represents a commit in the submodule's repository which is to be checked out @@ -172,7 +173,10 @@ class Submodule(base.IndexObject, Iterable, Traversable): def __eq__(self, other): """Compare with another submodule""" - return self.path == other.path and self.url == other.url and super(Submodule, self).__eq__(other) + # we may only compare by name as this should be the ID they are hashed with + # Otherwise this type wouldn't be hashable + # return self.path == other.path and self.url == other.url and super(Submodule, self).__eq__(other) + return self._name == other._name def __ne__(self, other): """Compare with another submodule for inequality""" @@ -185,6 +189,9 @@ class Submodule(base.IndexObject, Iterable, Traversable): def __str__(self): return self._name + def __repr__(self): + return "git.%s(name=%s, path=%s, url=%s, branch=%s)" % (type(self).__name__, self._name, self.path, self.url, self.branch) + @classmethod def _config_parser(cls, repo, parent_commit, read_only): """:return: Config Parser constrained to our submodule in read or write mode @@ -459,7 +466,7 @@ class Submodule(base.IndexObject, Iterable, Traversable): return self @unbare_repo - def move(self, module_path, module_only=False): + def move(self, module_path, configuration=True, module=True): """Move the submodule to a another module path. This involves physically moving the repository at our current path, changing the configuration, as well as adjusting our index entry accordingly. @@ -467,15 +474,21 @@ class Submodule(base.IndexObject, Iterable, Traversable): repository-relative path. Intermediate directories will be created accordingly. If the path already exists, it must be empty. Trailling (back)slashes are removed automatically - :param module_only: if True, only the repository managed by this submodule + :param configuration: if True, the configuration will be adjusted to let + the submodule point to the given path. + :param module: if True, the repository managed by this submodule will be moved, not the configuration. This will effectively leave your repository in an inconsistent state unless the configuration - and index already point to the target location. + and index already point to the target location. :return: self :raise ValueError: if the module path existed and was not empty, or was a file :note: Currently the method is not atomic, and it could leave the repository in an inconsistent state if a sub-step fails for some reason """ + if module + configuration < 1: + raise ValueError("You must specify to move at least the module or the configuration of the submodule") + #END handle input + module_path = to_native_path_linux(module_path) if module_path.endswith('/'): module_path = module_path[:-1] @@ -494,54 +507,64 @@ class Submodule(base.IndexObject, Iterable, Traversable): index = self.repo.index tekey = index.entry_key(module_path, 0) # if the target item already exists, fail - if not module_only and tekey in index.entries: + if configuration and tekey in index.entries: raise ValueError("Index entry for target path did alredy exist") #END handle index key already there # remove existing destination - if os.path.exists(dest_path): - if len(os.listdir(dest_path)): - raise ValueError("Destination module directory was not empty") - #END handle non-emptyness - - if os.path.islink(dest_path): - os.remove(dest_path) + if module: + if os.path.exists(dest_path): + if len(os.listdir(dest_path)): + raise ValueError("Destination module directory was not empty") + #END handle non-emptyness + + if os.path.islink(dest_path): + os.remove(dest_path) + else: + os.rmdir(dest_path) + #END handle link else: - os.rmdir(dest_path) - #END handle link - else: - # recreate parent directories - # NOTE: renames() does that now - pass - #END handle existance + # recreate parent directories + # NOTE: renames() does that now + pass + #END handle existance + # END handle module # move the module into place if possible cur_path = self.abspath - if os.path.exists(cur_path): + renamed_module = False + if module and os.path.exists(cur_path): os.renames(cur_path, dest_path) + renamed_module = True #END move physical module - # NOTE: from now on, we would have to undo the rename ! # rename the index entry - have to manipulate the index directly as # git-mv cannot be used on submodules ... yeah - if not module_only: - try: - ekey = index.entry_key(self.path, 0) - entry = index.entries[ekey] - del(index.entries[ekey]) - nentry = git.IndexEntry(entry[:3]+(module_path,)+entry[4:]) - index.entries[tekey] = nentry - except KeyError: - raise ValueError("Submodule's entry at %r did not exist" % (self.path)) - #END handle submodule doesn't exist - - # update configuration - writer = self.config_writer(index=index) # auto-write - writer.set_value('path', module_path) - self.path = module_path - del(writer) - # END handle module_only + try: + if configuration: + try: + ekey = index.entry_key(self.path, 0) + entry = index.entries[ekey] + del(index.entries[ekey]) + nentry = git.IndexEntry(entry[:3]+(module_path,)+entry[4:]) + index.entries[tekey] = nentry + except KeyError: + raise InvalidGitRepositoryError("Submodule's entry at %r did not exist" % (self.path)) + #END handle submodule doesn't exist + + # update configuration + writer = self.config_writer(index=index) # auto-write + writer.set_value('path', module_path) + self.path = module_path + del(writer) + # END handle configuration flag + except Exception: + if renamed_module: + os.renames(dest_path, cur_path) + # END undo module renaming + raise + #END handle undo rename return self @@ -917,7 +940,7 @@ class RootModule(Submodule): ################## cur_commit = repo.head.commit if previous_commit is None: - symref = SymbolicReference(repo, SymbolicReference.to_full_path('ORIG_HEAD')) + symref = repo.head.orig_head() try: previous_commit = symref.commit except Exception: @@ -936,8 +959,8 @@ class RootModule(Submodule): # HANDLE REMOVALS - psms = type(self).list_items(repo, parent_commit=previous_commit) - sms = self.children() + psms = self.list_items(repo, parent_commit=previous_commit) + sms = self.list_items(self.module()) spsms = set(psms) ssms = set(sms) @@ -958,7 +981,7 @@ class RootModule(Submodule): if sm.path != psm.path and psm.module_exists(): # move the module to the new path - psm.move(sm.path, module_only=True) + psm.move(sm.path, module=True, configuration=False) # END handle path changes if sm.module_exists(): @@ -970,26 +993,39 @@ class RootModule(Submodule): nn = '__new_origin__' smm = sm.module() rmts = smm.remotes - assert nn not in rmts - smr = smm.create_remote(nn, sm.url) - srm.fetch() - - # now delete the changed one - orig_name = None - for remote in rmts: - if remote.url == psm.url: - orig_name = remote.name - smm.delete_remote(remote) - break - # END if urls match - # END for each remote - # rename the new remote back to what it was - # if we have not found any remote with the original url - # we may not have a name. This is a special case, - # and its okay to fail her - assert orig_name is not None, "Couldn't find original remote-repo at url %r" % psm.url - smr.rename(orig_name) + # don't do anything if we already have the url we search in place + if len([r for r in rmts if r.url == sm.url]) == 0: + assert nn not in [r.name for r in rmts] + smr = smm.create_remote(nn, sm.url) + smr.fetch() + + # now delete the changed one + orig_name = None + for remote in rmts: + if remote.url == psm.url: + orig_name = remote.name + smm.delete_remote(remote) + break + # END if urls match + # END for each remote + + # if we didn't find a matching remote, but have exactly one, + # we can safely use this one + if len(rmts) == 1: + orig_name = rmts[0].name + smm.delete_remote(rmts[0]) + else: + # if we have not found any remote with the original url + # we may not have a name. This is a special case, + # and its okay to fail here + # Alternatively we could just generate a unique name + raise InvalidGitRepositoryError("Couldn't find original remote-repo at url %r" % psm.url) + # END only one remove + + # rename the new remote back to what it was + smr.rename(orig_name) + # END skip remote handling if new url already exists in module # END handle url if sm.branch != psm.branch: @@ -1020,11 +1056,13 @@ class RootModule(Submodule): # FINALLY UPDATE ALL ACTUAL SUBMODULES ########################################## for sm in sms: + # update the submodule using the default method sm.update(recursive=True, init=init, to_latest_revision=to_latest_revision) # update recursively depth first - question is which inconsitent # state will be better in case it fails somewhere. Defective branch - # or defective depth + # or defective depth. The RootSubmodule type will never process itself, + # which was done in the previous expression if recursive: type(cls)(sm.module()).update(recursive=True, force_remove=force_remove, init=init, to_latest_revision=to_latest_revision) -- cgit v1.2.1 From cf5eaddde33e983bc7b496f458bdd49154f6f498 Mon Sep 17 00:00:00 2001 From: Sebastian Thiel Date: Thu, 18 Nov 2010 20:44:21 +0100 Subject: Updated tests and implementation to verify functionality for handling submodule removals, as well as url changes --- lib/git/objects/submodule.py | 80 +++++++++++++++++++++++++++++++++++--------- lib/git/remote.py | 5 +++ 2 files changed, 69 insertions(+), 16 deletions(-) (limited to 'lib/git') diff --git a/lib/git/objects/submodule.py b/lib/git/objects/submodule.py index 9fb8ce8f..948a267f 100644 --- a/lib/git/objects/submodule.py +++ b/lib/git/objects/submodule.py @@ -958,7 +958,6 @@ class RootModule(Submodule): # END handle previous commit - # HANDLE REMOVALS psms = self.list_items(repo, parent_commit=previous_commit) sms = self.list_items(self.module()) spsms = set(psms) @@ -974,7 +973,9 @@ class RootModule(Submodule): rsm.remove(configuration=False, module=True, force=force_remove) # END for each removed submodule - # HANDLE PATH RENAMES + url changes + branch changes + # HANDLE PATH RENAMES + ##################### + # url changes + branch changes for csm in (spsms & ssms): psm = psms[csm.name] sm = sms[csm.name] @@ -996,35 +997,79 @@ class RootModule(Submodule): # don't do anything if we already have the url we search in place if len([r for r in rmts if r.url == sm.url]) == 0: + + assert nn not in [r.name for r in rmts] smr = smm.create_remote(nn, sm.url) smr.fetch() + # If we have a tracking branch, it should be available + # in the new remote as well. + if len([r for r in smr.refs if r.remote_head == sm.branch.name]) == 0: + raise ValueError("Submodule branch named %r was not available in new submodule remote at %r" % (sm.branch.name, sm.url)) + # END head is not detached + # now delete the changed one - orig_name = None + rmt_for_deletion = None for remote in rmts: if remote.url == psm.url: - orig_name = remote.name - smm.delete_remote(remote) + rmt_for_deletion = remote break # END if urls match # END for each remote # if we didn't find a matching remote, but have exactly one, # we can safely use this one - if len(rmts) == 1: - orig_name = rmts[0].name - smm.delete_remote(rmts[0]) - else: - # if we have not found any remote with the original url - # we may not have a name. This is a special case, - # and its okay to fail here - # Alternatively we could just generate a unique name - raise InvalidGitRepositoryError("Couldn't find original remote-repo at url %r" % psm.url) - # END only one remove + if rmt_for_deletion is None: + if len(rmts) == 1: + rmt_for_deletion = rmts[0] + else: + # if we have not found any remote with the original url + # we may not have a name. This is a special case, + # and its okay to fail here + # Alternatively we could just generate a unique name and leave all + # existing ones in place + raise InvalidGitRepositoryError("Couldn't find original remote-repo at url %r" % psm.url) + #END handle one single remote + # END handle check we found a remote + + orig_name = rmt_for_deletion.name + smm.delete_remote(rmt_for_deletion) + # NOTE: Currently we leave tags from the deleted remotes + # as well as separate tracking branches in the possibly totally + # changed repository ( someone could have changed the url to + # another project ). At some point, one might want to clean + # it up, but the danger is high to remove stuff the user + # has added explicitly # rename the new remote back to what it was smr.rename(orig_name) + + # early on, we verified that the our current tracking branch + # exists in the remote. Now we have to assure that the + # sha we point to is still contained in the new remote + # tracking branch. + smsha = sm.binsha + found = False + rref = smr.refs[self.branch.name] + for c in rref.commit.traverse(): + if c.binsha == smsha: + found = True + break + # END traverse all commits in search for sha + # END for each commit + + if not found: + # adjust our internal binsha to use the one of the remote + # this way, it will be checked out in the next step + # This will change the submodule relative to us, so + # the user will be able to commit the change easily + print >> sys.stderr, "WARNING: Current sha %s was not contained in the tracking branch at the new remote, setting it the the remote's tracking branch" % sm.hexsha + sm.binsha = rref.commit.binsha + #END reset binsha + + #NOTE: All checkout is performed by the base implementation of update + # END skip remote handling if new url already exists in module # END handle url @@ -1049,12 +1094,15 @@ class RootModule(Submodule): # current remotes, this just means we can't handle it pass # END exception handling + + #NOTE: All checkout is done in the base implementation of update + #END handle branch #END handle # END for each common submodule # FINALLY UPDATE ALL ACTUAL SUBMODULES - ########################################## + ###################################### for sm in sms: # update the submodule using the default method sm.update(recursive=True, init=init, to_latest_revision=to_latest_revision) diff --git a/lib/git/remote.py b/lib/git/remote.py index 5124c603..a06da222 100644 --- a/lib/git/remote.py +++ b/lib/git/remote.py @@ -28,6 +28,7 @@ from gitdb.util import join import re import os +import sys __all__ = ('RemoteProgress', 'PushInfo', 'FetchInfo', 'Remote') @@ -591,6 +592,10 @@ class Remote(LazyMixin, Iterable): for line in self._digest_process_messages(proc.stderr, progress): if line.startswith('From') or line.startswith('remote: Total'): continue + elif line.startswith('warning:'): + print >> sys.stderr, line + continue + # END handle special messages fetch_info_lines.append(line) # END for each line -- cgit v1.2.1 From 3f2d76ba8e6d004ff5849ed8c7c34f6a4ac2e1e3 Mon Sep 17 00:00:00 2001 From: Sebastian Thiel Date: Thu, 18 Nov 2010 21:36:01 +0100 Subject: Added test for branch changes - it appears to work well, at least as far as the restricted tests are concerned --- lib/git/objects/submodule.py | 19 ++++++++++++++----- 1 file changed, 14 insertions(+), 5 deletions(-) (limited to 'lib/git') diff --git a/lib/git/objects/submodule.py b/lib/git/objects/submodule.py index 948a267f..aa11909f 100644 --- a/lib/git/objects/submodule.py +++ b/lib/git/objects/submodule.py @@ -41,7 +41,7 @@ def unbare_repo(func): wrapper.__name__ = func.__name__ return wrapper -def find_remote_branch(remotes, branch): +def find_first_remote_branch(remotes, branch): """Find the remote branch matching the name of the given branch or raise InvalidGitRepositoryError""" for remote in remotes: try: @@ -394,7 +394,7 @@ class Submodule(base.IndexObject, Iterable, Traversable): # see whether we have a valid branch to checkout try: # find a remote which has our branch - we try to be flexible - remote_branch = find_remote_branch(mrepo.remotes, self.branch) + remote_branch = find_first_remote_branch(mrepo.remotes, self.branch) local_branch = self.branch if not local_branch.is_valid(): # Setup a tracking configuration - branch doesn't need to @@ -1078,14 +1078,23 @@ class RootModule(Submodule): # new remote branch smm = sm.module() smmr = smm.remotes - tbr = git.Head.create(smm, sm.branch.name) - tbr.set_tracking_branch(find_remote_branch(smmr, sm.branch)) + try: + tbr = git.Head.create(smm, sm.branch.name) + except git.GitCommandError, e: + if e.status != 128: + raise + #END handle something unexpected + + # ... or reuse the existing one + tbr = git.Head(smm, git.Head.to_full_path(sm.branch.name)) + #END assure tracking branch exists + tbr.set_tracking_branch(find_first_remote_branch(smmr, sm.branch)) # figure out whether the previous tracking branch contains # new commits compared to the other one, if not we can # delete it. try: - tbr = find_remote_branch(smmr, psm.branch) + tbr = find_first_remote_branch(smmr, psm.branch) if len(smm.git.cherry(tbr, psm.branch)) == 0: psm.branch.delete(smm, psm.branch) #END delete original tracking branch if there are no changes -- cgit v1.2.1 From ebe8f644e751c1b2115301c1a961bef14d2cce89 Mon Sep 17 00:00:00 2001 From: Sebastian Thiel Date: Thu, 18 Nov 2010 22:11:06 +0100 Subject: Added test for the recursive code path. --- lib/git/objects/submodule.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'lib/git') diff --git a/lib/git/objects/submodule.py b/lib/git/objects/submodule.py index aa11909f..c769b160 100644 --- a/lib/git/objects/submodule.py +++ b/lib/git/objects/submodule.py @@ -1121,7 +1121,7 @@ class RootModule(Submodule): # or defective depth. The RootSubmodule type will never process itself, # which was done in the previous expression if recursive: - type(cls)(sm.module()).update(recursive=True, force_remove=force_remove, + type(self)(sm.module()).update(recursive=True, force_remove=force_remove, init=init, to_latest_revision=to_latest_revision) #END handle recursive # END for each submodule to update -- cgit v1.2.1 From 4c34d5c3f2a4ed7194276a026e0ec6437d339c67 Mon Sep 17 00:00:00 2001 From: Sebastian Thiel Date: Thu, 18 Nov 2010 22:44:53 +0100 Subject: Split big submodule file into smaller files. Tried to manually get imports right, but its not yet tested --- lib/git/objects/submodule.py | 1133 --------------------------------- lib/git/objects/submodule/__init__.py | 3 + lib/git/objects/submodule/base.py | 794 +++++++++++++++++++++++ lib/git/objects/submodule/root.py | 259 ++++++++ lib/git/objects/submodule/util.py | 101 +++ 5 files changed, 1157 insertions(+), 1133 deletions(-) delete mode 100644 lib/git/objects/submodule.py create mode 100644 lib/git/objects/submodule/__init__.py create mode 100644 lib/git/objects/submodule/base.py create mode 100644 lib/git/objects/submodule/root.py create mode 100644 lib/git/objects/submodule/util.py (limited to 'lib/git') diff --git a/lib/git/objects/submodule.py b/lib/git/objects/submodule.py deleted file mode 100644 index c769b160..00000000 --- a/lib/git/objects/submodule.py +++ /dev/null @@ -1,1133 +0,0 @@ -import base -from util import Traversable -from StringIO import StringIO # need a dict to set bloody .name field -from git.util import Iterable, join_path_native, to_native_path_linux -from git.config import GitConfigParser, SectionConstraint -from git.exc import InvalidGitRepositoryError, NoSuchPathError -import stat -import git - -import os -import sys -import weakref -import shutil - -__all__ = ("Submodule", "RootModule") - -#{ Utilities - -def sm_section(name): - """:return: section title used in .gitmodules configuration file""" - return 'submodule "%s"' % name - -def sm_name(section): - """:return: name of the submodule as parsed from the section name""" - section = section.strip() - return section[11:-1] - -def mkhead(repo, path): - """:return: New branch/head instance""" - return git.Head(repo, git.Head.to_full_path(path)) - -def unbare_repo(func): - """Methods with this decorator raise InvalidGitRepositoryError if they - encounter a bare repository""" - def wrapper(self, *args, **kwargs): - if self.repo.bare: - raise InvalidGitRepositoryError("Method '%s' cannot operate on bare repositories" % func.__name__) - #END bare method - return func(self, *args, **kwargs) - # END wrapper - wrapper.__name__ = func.__name__ - return wrapper - -def find_first_remote_branch(remotes, branch): - """Find the remote branch matching the name of the given branch or raise InvalidGitRepositoryError""" - for remote in remotes: - try: - return remote.refs[branch.name] - except IndexError: - continue - # END exception handling - #END for remote - raise InvalidGitRepositoryError("Didn't find remote branch %r in any of the given remotes", branch) - -#} END utilities - - -#{ Classes - -class SubmoduleConfigParser(GitConfigParser): - """ - Catches calls to _write, and updates the .gitmodules blob in the index - with the new data, if we have written into a stream. Otherwise it will - add the local file to the index to make it correspond with the working tree. - Additionally, the cache must be cleared - - Please note that no mutating method will work in bare mode - """ - - def __init__(self, *args, **kwargs): - self._smref = None - self._index = None - self._auto_write = True - super(SubmoduleConfigParser, self).__init__(*args, **kwargs) - - #{ Interface - def set_submodule(self, submodule): - """Set this instance's submodule. It must be called before - the first write operation begins""" - self._smref = weakref.ref(submodule) - - def flush_to_index(self): - """Flush changes in our configuration file to the index""" - assert self._smref is not None - # should always have a file here - assert not isinstance(self._file_or_files, StringIO) - - sm = self._smref() - if sm is not None: - index = self._index - if index is None: - index = sm.repo.index - # END handle index - index.add([sm.k_modules_file], write=self._auto_write) - sm._clear_cache() - # END handle weakref - - #} END interface - - #{ Overridden Methods - def write(self): - rval = super(SubmoduleConfigParser, self).write() - self.flush_to_index() - return rval - # END overridden methods - - -class Submodule(base.IndexObject, Iterable, Traversable): - """Implements access to a git submodule. They are special in that their sha - represents a commit in the submodule's repository which is to be checked out - at the path of this instance. - The submodule type does not have a string type associated with it, as it exists - solely as a marker in the tree and index. - - All methods work in bare and non-bare repositories.""" - - _id_attribute_ = "name" - k_modules_file = '.gitmodules' - k_head_option = 'branch' - k_head_default = 'master' - k_default_mode = stat.S_IFDIR | stat.S_IFLNK # submodules are directories with link-status - - # this is a bogus type for base class compatability - type = 'submodule' - - __slots__ = ('_parent_commit', '_url', '_branch', '_name', '__weakref__') - _cache_attrs = ('path', '_url', '_branch') - - def __init__(self, repo, binsha, mode=None, path=None, name = None, parent_commit=None, url=None, branch=None): - """Initialize this instance with its attributes. We only document the ones - that differ from ``IndexObject`` - :param repo: Our parent repository - :param binsha: binary sha referring to a commit in the remote repository, see url parameter - :param parent_commit: see set_parent_commit() - :param url: The url to the remote repository which is the submodule - :param branch: Head instance to checkout when cloning the remote repository""" - super(Submodule, self).__init__(repo, binsha, mode, path) - self.size = 0 - if parent_commit is not None: - self._parent_commit = parent_commit - if url is not None: - self._url = url - if branch is not None: - assert isinstance(branch, git.Head) - self._branch = branch - if name is not None: - self._name = name - - def _set_cache_(self, attr): - if attr == '_parent_commit': - # set a default value, which is the root tree of the current head - self._parent_commit = self.repo.commit() - elif attr in ('path', '_url', '_branch'): - reader = self.config_reader() - # default submodule values - self.path = reader.get_value('path') - self._url = reader.get_value('url') - # git-python extension values - optional - self._branch = mkhead(self.repo, reader.get_value(self.k_head_option, self.k_head_default)) - elif attr == '_name': - raise AttributeError("Cannot retrieve the name of a submodule if it was not set initially") - else: - super(Submodule, self)._set_cache_(attr) - # END handle attribute name - - def _get_intermediate_items(self, item): - """:return: all the submodules of our module repository""" - try: - return type(self).list_items(item.module()) - except InvalidGitRepositoryError: - return list() - # END handle intermeditate items - - def __eq__(self, other): - """Compare with another submodule""" - # we may only compare by name as this should be the ID they are hashed with - # Otherwise this type wouldn't be hashable - # return self.path == other.path and self.url == other.url and super(Submodule, self).__eq__(other) - return self._name == other._name - - def __ne__(self, other): - """Compare with another submodule for inequality""" - return not (self == other) - - def __hash__(self): - """Hash this instance using its logical id, not the sha""" - return hash(self._name) - - def __str__(self): - return self._name - - def __repr__(self): - return "git.%s(name=%s, path=%s, url=%s, branch=%s)" % (type(self).__name__, self._name, self.path, self.url, self.branch) - - @classmethod - def _config_parser(cls, repo, parent_commit, read_only): - """:return: Config Parser constrained to our submodule in read or write mode - :raise IOError: If the .gitmodules file cannot be found, either locally or in the repository - at the given parent commit. Otherwise the exception would be delayed until the first - access of the config parser""" - parent_matches_head = repo.head.commit == parent_commit - if not repo.bare and parent_matches_head: - fp_module = cls.k_modules_file - fp_module_path = os.path.join(repo.working_tree_dir, fp_module) - if not os.path.isfile(fp_module_path): - raise IOError("%s file was not accessible" % fp_module_path) - # END handle existance - fp_module = fp_module_path - else: - try: - fp_module = cls._sio_modules(parent_commit) - except KeyError: - raise IOError("Could not find %s file in the tree of parent commit %s" % (cls.k_modules_file, parent_commit)) - # END handle exceptions - # END handle non-bare working tree - - if not read_only and (repo.bare or not parent_matches_head): - raise ValueError("Cannot write blobs of 'historical' submodule configurations") - # END handle writes of historical submodules - - return SubmoduleConfigParser(fp_module, read_only = read_only) - - def _clear_cache(self): - # clear the possibly changed values - for name in self._cache_attrs: - try: - delattr(self, name) - except AttributeError: - pass - # END try attr deletion - # END for each name to delete - - @classmethod - def _sio_modules(cls, parent_commit): - """:return: Configuration file as StringIO - we only access it through the respective blob's data""" - sio = StringIO(parent_commit.tree[cls.k_modules_file].data_stream.read()) - sio.name = cls.k_modules_file - return sio - - def _config_parser_constrained(self, read_only): - """:return: Config Parser constrained to our submodule in read or write mode""" - parser = self._config_parser(self.repo, self._parent_commit, read_only) - parser.set_submodule(self) - return SectionConstraint(parser, sm_section(self.name)) - - #{ Edit Interface - - @classmethod - def add(cls, repo, name, path, url=None, branch=None, no_checkout=False): - """Add a new submodule to the given repository. This will alter the index - as well as the .gitmodules file, but will not create a new commit. - If the submodule already exists, no matter if the configuration differs - from the one provided, the existing submodule will be returned. - :param repo: Repository instance which should receive the submodule - :param name: The name/identifier for the submodule - :param path: repository-relative or absolute path at which the submodule - should be located - It will be created as required during the repository initialization. - :param url: git-clone compatible URL, see git-clone reference for more information - If None, the repository is assumed to exist, and the url of the first - remote is taken instead. This is useful if you want to make an existing - repository a submodule of anotherone. - :param branch: branch at which the submodule should (later) be checked out. - The given branch must exist in the remote repository, and will be checked - out locally as a tracking branch. - It will only be written into the configuration if it not None, which is - when the checked out branch will be the one the remote HEAD pointed to. - The result you get in these situation is somewhat fuzzy, and it is recommended - to specify at least 'master' here - :param no_checkout: if True, and if the repository has to be cloned manually, - no checkout will be performed - :return: The newly created submodule instance - :note: works atomically, such that no change will be done if the repository - update fails for instance""" - if repo.bare: - raise InvalidGitRepositoryError("Cannot add submodules to bare repositories") - # END handle bare repos - - path = to_native_path_linux(path) - if path.endswith('/'): - path = path[:-1] - # END handle trailing slash - - # INSTANTIATE INTERMEDIATE SM - sm = cls(repo, cls.NULL_BIN_SHA, cls.k_default_mode, path, name) - if sm.exists(): - # reretrieve submodule from tree - return repo.head.commit.tree[path] - # END handle existing - - br = mkhead(repo, branch or cls.k_head_default) - has_module = sm.module_exists() - branch_is_default = branch is None - if has_module and url is not None: - if url not in [r.url for r in sm.module().remotes]: - raise ValueError("Specified URL '%s' does not match any remote url of the repository at '%s'" % (url, sm.abspath)) - # END check url - # END verify urls match - - mrepo = None - if url is None: - if not has_module: - raise ValueError("A URL was not given and existing repository did not exsit at %s" % path) - # END check url - mrepo = sm.module() - urls = [r.url for r in mrepo.remotes] - if not urls: - raise ValueError("Didn't find any remote url in repository at %s" % sm.abspath) - # END verify we have url - url = urls[0] - else: - # clone new repo - kwargs = {'n' : no_checkout} - if not branch_is_default: - kwargs['b'] = str(br) - # END setup checkout-branch - mrepo = git.Repo.clone_from(url, path, **kwargs) - # END verify url - - # update configuration and index - index = sm.repo.index - writer = sm.config_writer(index=index, write=False) - writer.set_value('url', url) - writer.set_value('path', path) - - sm._url = url - if not branch_is_default: - # store full path - writer.set_value(cls.k_head_option, br.path) - sm._branch = br.path - # END handle path - del(writer) - - # NOTE: Have to write the repo config file as well, otherwise - # the default implementation will be offended and not update the repository - # Maybe this is a good way to assure it doesn't get into our way, but - # we want to stay backwards compatible too ... . Its so redundant ! - repo.config_writer().set_value(sm_section(sm.name), 'url', url) - - # we deliberatly assume that our head matches our index ! - pcommit = repo.head.commit - sm._parent_commit = pcommit - sm.binsha = mrepo.head.commit.binsha - index.add([sm], write=True) - - return sm - - def update(self, recursive=False, init=True, to_latest_revision=False): - """Update the repository of this submodule to point to the checkout - we point at with the binsha of this instance. - :param recursive: if True, we will operate recursively and update child- - modules as well. - :param init: if True, the module repository will be cloned into place if necessary - :param to_latest_revision: if True, the submodule's sha will be ignored during checkout. - Instead, the remote will be fetched, and the local tracking branch updated. - This only works if we have a local tracking branch, which is the case - if the remote repository had a master branch, or of the 'branch' option - was specified for this submodule and the branch existed remotely - :note: does nothing in bare repositories - :note: method is definitely not atomic if recurisve is True - :return: self""" - if self.repo.bare: - return self - #END pass in bare mode - - - # ASSURE REPO IS PRESENT AND UPTODATE - ##################################### - try: - mrepo = self.module() - for remote in mrepo.remotes: - remote.fetch() - #END fetch new data - except InvalidGitRepositoryError: - if not init: - return self - # END early abort if init is not allowed - import git - - # there is no git-repository yet - but delete empty paths - module_path = join_path_native(self.repo.working_tree_dir, self.path) - if os.path.isdir(module_path): - try: - os.rmdir(module_path) - except OSError: - raise OSError("Module directory at %r does already exist and is non-empty" % module_path) - # END handle OSError - # END handle directory removal - - # don't check it out at first - nonetheless it will create a local - # branch according to the remote-HEAD if possible - mrepo = git.Repo.clone_from(self.url, module_path, n=True) - - # see whether we have a valid branch to checkout - try: - # find a remote which has our branch - we try to be flexible - remote_branch = find_first_remote_branch(mrepo.remotes, self.branch) - local_branch = self.branch - if not local_branch.is_valid(): - # Setup a tracking configuration - branch doesn't need to - # exist to do that - local_branch.set_tracking_branch(remote_branch) - #END handle local branch - - # have a valid branch, but no checkout - make sure we can figure - # that out by marking the commit with a null_sha - # have to write it directly as .commit = NULLSHA tries to resolve the sha - # This will bring the branch into existance - refpath = join_path_native(mrepo.git_dir, local_branch.path) - refdir = os.path.dirname(refpath) - if not os.path.isdir(refdir): - os.makedirs(refdir) - #END handle directory - open(refpath, 'w').write(self.NULL_HEX_SHA) - # END initial checkout + branch creation - - # make sure HEAD is not detached - mrepo.head.ref = local_branch - except IndexError: - print >> sys.stderr, "Warning: Failed to checkout tracking branch %s" % self.branch - #END handle tracking branch - #END handle initalization - - - # DETERMINE SHAS TO CHECKOUT - ############################ - binsha = self.binsha - hexsha = self.hexsha - is_detached = mrepo.head.is_detached - if to_latest_revision: - msg_base = "Cannot update to latest revision in repository at %r as " % mrepo.working_dir - if not is_detached: - rref = mrepo.head.ref.tracking_branch() - if rref is not None: - rcommit = rref.commit - binsha = rcommit.binsha - hexsha = rcommit.hexsha - else: - print >> sys.stderr, "%s a tracking branch was not set for local branch '%s'" % (msg_base, mrepo.head.ref) - # END handle remote ref - else: - print >> sys.stderr, "%s there was no local tracking branch" % msg_base - # END handle detached head - # END handle to_latest_revision option - - # update the working tree - if mrepo.head.commit.binsha != binsha: - if is_detached: - mrepo.git.checkout(hexsha) - else: - # TODO: allow to specify a rebase, merge, or reset - # TODO: Warn if the hexsha forces the tracking branch off the remote - # branch - this should be prevented when setting the branch option - mrepo.head.reset(hexsha, index=True, working_tree=True) - # END handle checkout - # END update to new commit only if needed - - # HANDLE RECURSION - ################## - if recursive: - for submodule in self.iter_items(self.module()): - submodule.update(recursive, init, to_latest_revision) - # END handle recursive update - # END for each submodule - - return self - - @unbare_repo - def move(self, module_path, configuration=True, module=True): - """Move the submodule to a another module path. This involves physically moving - the repository at our current path, changing the configuration, as well as - adjusting our index entry accordingly. - :param module_path: the path to which to move our module, given as - repository-relative path. Intermediate directories will be created - accordingly. If the path already exists, it must be empty. - Trailling (back)slashes are removed automatically - :param configuration: if True, the configuration will be adjusted to let - the submodule point to the given path. - :param module: if True, the repository managed by this submodule - will be moved, not the configuration. This will effectively - leave your repository in an inconsistent state unless the configuration - and index already point to the target location. - :return: self - :raise ValueError: if the module path existed and was not empty, or was a file - :note: Currently the method is not atomic, and it could leave the repository - in an inconsistent state if a sub-step fails for some reason - """ - if module + configuration < 1: - raise ValueError("You must specify to move at least the module or the configuration of the submodule") - #END handle input - - module_path = to_native_path_linux(module_path) - if module_path.endswith('/'): - module_path = module_path[:-1] - # END handle trailing slash - - # VERIFY DESTINATION - if module_path == self.path: - return self - #END handle no change - - dest_path = join_path_native(self.repo.working_tree_dir, module_path) - if os.path.isfile(dest_path): - raise ValueError("Cannot move repository onto a file: %s" % dest_path) - # END handle target files - - index = self.repo.index - tekey = index.entry_key(module_path, 0) - # if the target item already exists, fail - if configuration and tekey in index.entries: - raise ValueError("Index entry for target path did alredy exist") - #END handle index key already there - - # remove existing destination - if module: - if os.path.exists(dest_path): - if len(os.listdir(dest_path)): - raise ValueError("Destination module directory was not empty") - #END handle non-emptyness - - if os.path.islink(dest_path): - os.remove(dest_path) - else: - os.rmdir(dest_path) - #END handle link - else: - # recreate parent directories - # NOTE: renames() does that now - pass - #END handle existance - # END handle module - - # move the module into place if possible - cur_path = self.abspath - renamed_module = False - if module and os.path.exists(cur_path): - os.renames(cur_path, dest_path) - renamed_module = True - #END move physical module - - - # rename the index entry - have to manipulate the index directly as - # git-mv cannot be used on submodules ... yeah - try: - if configuration: - try: - ekey = index.entry_key(self.path, 0) - entry = index.entries[ekey] - del(index.entries[ekey]) - nentry = git.IndexEntry(entry[:3]+(module_path,)+entry[4:]) - index.entries[tekey] = nentry - except KeyError: - raise InvalidGitRepositoryError("Submodule's entry at %r did not exist" % (self.path)) - #END handle submodule doesn't exist - - # update configuration - writer = self.config_writer(index=index) # auto-write - writer.set_value('path', module_path) - self.path = module_path - del(writer) - # END handle configuration flag - except Exception: - if renamed_module: - os.renames(dest_path, cur_path) - # END undo module renaming - raise - #END handle undo rename - - return self - - @unbare_repo - def remove(self, module=True, force=False, configuration=True, dry_run=False): - """Remove this submodule from the repository. This will remove our entry - from the .gitmodules file and the entry in the .git/config file. - :param module: If True, the module we point to will be deleted - as well. If the module is currently on a commit which is not part - of any branch in the remote, if the currently checked out branch - is ahead of its tracking branch, if you have modifications in the - working tree, or untracked files, - In case the removal of the repository fails for these reasons, the - submodule status will not have been altered. - If this submodule has child-modules on its own, these will be deleted - prior to touching the own module. - :param force: Enforces the deletion of the module even though it contains - modifications. This basically enforces a brute-force file system based - deletion. - :param configuration: if True, the submodule is deleted from the configuration, - otherwise it isn't. Although this should be enabled most of the times, - this flag enables you to safely delete the repository of your submodule. - :param dry_run: if True, we will not actually do anything, but throw the errors - we would usually throw - :return: self - :note: doesn't work in bare repositories - :raise InvalidGitRepositoryError: thrown if the repository cannot be deleted - :raise OSError: if directories or files could not be removed""" - if not (module + configuration): - raise ValueError("Need to specify to delete at least the module, or the configuration") - # END handle params - - # DELETE MODULE REPOSITORY - ########################## - if module and self.module_exists(): - if force: - # take the fast lane and just delete everything in our module path - # TODO: If we run into permission problems, we have a highly inconsistent - # state. Delete the .git folders last, start with the submodules first - mp = self.abspath - method = None - if os.path.islink(mp): - method = os.remove - elif os.path.isdir(mp): - method = shutil.rmtree - elif os.path.exists(mp): - raise AssertionError("Cannot forcibly delete repository as it was neither a link, nor a directory") - #END handle brutal deletion - if not dry_run: - assert method - method(mp) - #END apply deletion method - else: - # verify we may delete our module - mod = self.module() - if mod.is_dirty(untracked_files=True): - raise InvalidGitRepositoryError("Cannot delete module at %s with any modifications, unless force is specified" % mod.working_tree_dir) - # END check for dirt - - # figure out whether we have new commits compared to the remotes - # NOTE: If the user pulled all the time, the remote heads might - # not have been updated, so commits coming from the remote look - # as if they come from us. But we stay strictly read-only and - # don't fetch beforhand. - for remote in mod.remotes: - num_branches_with_new_commits = 0 - rrefs = remote.refs - for rref in rrefs: - num_branches_with_new_commits = len(mod.git.cherry(rref)) != 0 - # END for each remote ref - # not a single remote branch contained all our commits - if num_branches_with_new_commits == len(rrefs): - raise InvalidGitRepositoryError("Cannot delete module at %s as there are new commits" % mod.working_tree_dir) - # END handle new commits - # END for each remote - - # gently remove all submodule repositories - for sm in self.children(): - sm.remove(module=True, force=False, configuration=False, dry_run=dry_run) - # END for each child-submodule - - # finally delete our own submodule - if not dry_run: - shutil.rmtree(mod.working_tree_dir) - # END delete tree if possible - # END handle force - # END handle module deletion - - # DELETE CONFIGURATION - ###################### - if configuration and not dry_run: - # first the index-entry - index = self.repo.index - try: - del(index.entries[index.entry_key(self.path, 0)]) - except KeyError: - pass - #END delete entry - index.write() - - # now git config - need the config intact, otherwise we can't query - # inforamtion anymore - self.repo.config_writer().remove_section(sm_section(self.name)) - self.config_writer().remove_section() - # END delete configuration - - return self - - def set_parent_commit(self, commit, check=True): - """Set this instance to use the given commit whose tree is supposed to - contain the .gitmodules blob. - :param commit: Commit'ish reference pointing at the root_tree - :param check: if True, relatively expensive checks will be performed to verify - validity of the submodule. - :raise ValueError: if the commit's tree didn't contain the .gitmodules blob. - :raise ValueError: if the parent commit didn't store this submodule under the - current path - :return: self""" - pcommit = self.repo.commit(commit) - pctree = pcommit.tree - if self.k_modules_file not in pctree: - raise ValueError("Tree of commit %s did not contain the %s file" % (commit, self.k_modules_file)) - # END handle exceptions - - prev_pc = self._parent_commit - self._parent_commit = pcommit - - if check: - parser = self._config_parser(self.repo, self._parent_commit, read_only=True) - if not parser.has_section(sm_section(self.name)): - self._parent_commit = prev_pc - raise ValueError("Submodule at path %r did not exist in parent commit %s" % (self.path, commit)) - # END handle submodule did not exist - # END handle checking mode - - # update our sha, it could have changed - self.binsha = pctree[self.path].binsha - - self._clear_cache() - - return self - - @unbare_repo - def config_writer(self, index=None, write=True): - """:return: a config writer instance allowing you to read and write the data - belonging to this submodule into the .gitmodules file. - - :param index: if not None, an IndexFile instance which should be written. - defaults to the index of the Submodule's parent repository. - :param write: if True, the index will be written each time a configuration - value changes. - :note: the parameters allow for a more efficient writing of the index, - as you can pass in a modified index on your own, prevent automatic writing, - and write yourself once the whole operation is complete - :raise ValueError: if trying to get a writer on a parent_commit which does not - match the current head commit - :raise IOError: If the .gitmodules file/blob could not be read""" - writer = self._config_parser_constrained(read_only=False) - if index is not None: - writer.config._index = index - writer.config._auto_write = write - return writer - - #} END edit interface - - #{ Query Interface - - @unbare_repo - def module(self): - """:return: Repo instance initialized from the repository at our submodule path - :raise InvalidGitRepositoryError: if a repository was not available. This could - also mean that it was not yet initialized""" - # late import to workaround circular dependencies - module_path = self.abspath - try: - repo = git.Repo(module_path) - if repo != self.repo: - return repo - # END handle repo uninitialized - except (InvalidGitRepositoryError, NoSuchPathError): - raise InvalidGitRepositoryError("No valid repository at %s" % self.path) - else: - raise InvalidGitRepositoryError("Repository at %r was not yet checked out" % module_path) - # END handle exceptions - - def module_exists(self): - """:return: True if our module exists and is a valid git repository. See module() method""" - try: - self.module() - return True - except Exception: - return False - # END handle exception - - def exists(self): - """:return: True if the submodule exists, False otherwise. Please note that - a submodule may exist (in the .gitmodules file) even though its module - doesn't exist""" - # keep attributes for later, and restore them if we have no valid data - # this way we do not actually alter the state of the object - loc = locals() - for attr in self._cache_attrs: - if hasattr(self, attr): - loc[attr] = getattr(self, attr) - # END if we have the attribute cache - #END for each attr - self._clear_cache() - - try: - try: - self.path - return True - except Exception: - return False - # END handle exceptions - finally: - for attr in self._cache_attrs: - if attr in loc: - setattr(self, attr, loc[attr]) - # END if we have a cache - # END reapply each attribute - # END handle object state consistency - - @property - def branch(self): - """:return: The branch instance that we are to checkout""" - return self._branch - - @property - def url(self): - """:return: The url to the repository which our module-repository refers to""" - return self._url - - @property - def parent_commit(self): - """:return: Commit instance with the tree containing the .gitmodules file - :note: will always point to the current head's commit if it was not set explicitly""" - return self._parent_commit - - @property - def name(self): - """:return: The name of this submodule. It is used to identify it within the - .gitmodules file. - :note: by default, the name is the path at which to find the submodule, but - in git-python it should be a unique identifier similar to the identifiers - used for remotes, which allows to change the path of the submodule - easily - """ - return self._name - - def config_reader(self): - """:return: ConfigReader instance which allows you to qurey the configuration values - of this submodule, as provided by the .gitmodules file - :note: The config reader will actually read the data directly from the repository - and thus does not need nor care about your working tree. - :note: Should be cached by the caller and only kept as long as needed - :raise IOError: If the .gitmodules file/blob could not be read""" - return self._config_parser_constrained(read_only=True) - - def children(self): - """:return: IterableList(Submodule, ...) an iterable list of submodules instances - which are children of this submodule - :raise InvalidGitRepositoryError: if the submodule is not checked-out""" - return self._get_intermediate_items(self) - - #} END query interface - - #{ Iterable Interface - - @classmethod - def iter_items(cls, repo, parent_commit='HEAD'): - """:return: iterator yielding Submodule instances available in the given repository""" - pc = repo.commit(parent_commit) # parent commit instance - try: - parser = cls._config_parser(repo, pc, read_only=True) - except IOError: - raise StopIteration - # END handle empty iterator - - rt = pc.tree # root tree - - for sms in parser.sections(): - n = sm_name(sms) - p = parser.get_value(sms, 'path') - u = parser.get_value(sms, 'url') - b = cls.k_head_default - if parser.has_option(sms, cls.k_head_option): - b = parser.get_value(sms, cls.k_head_option) - # END handle optional information - - # get the binsha - index = repo.index - try: - sm = rt[p] - except KeyError: - # try the index, maybe it was just added - try: - entry = index.entries[index.entry_key(p, 0)] - sm = cls(repo, entry.binsha, entry.mode, entry.path) - except KeyError: - raise InvalidGitRepositoryError("Gitmodule path %r did not exist in revision of parent commit %s" % (p, parent_commit)) - # END handle keyerror - # END handle critical error - - # fill in remaining info - saves time as it doesn't have to be parsed again - sm._name = n - sm._parent_commit = pc - sm._branch = mkhead(repo, b) - sm._url = u - - yield sm - # END for each section - - #} END iterable interface - - -class RootModule(Submodule): - """A (virtual) Root of all submodules in the given repository. It can be used - to more easily traverse all submodules of the master repository""" - - __slots__ = tuple() - - k_root_name = '__ROOT__' - - def __init__(self, repo): - # repo, binsha, mode=None, path=None, name = None, parent_commit=None, url=None, ref=None) - super(RootModule, self).__init__( - repo, - binsha = self.NULL_BIN_SHA, - mode = self.k_default_mode, - path = '', - name = self.k_root_name, - parent_commit = repo.head.commit, - url = '', - branch = mkhead(repo, self.k_head_default) - ) - - - def _clear_cache(self): - """May not do anything""" - pass - - #{ Interface - - def update(self, previous_commit=None, recursive=True, force_remove=False, init=True, to_latest_revision=False): - """Update the submodules of this repository to the current HEAD commit. - This method behaves smartly by determining changes of the path of a submodules - repository, next to changes to the to-be-checked-out commit or the branch to be - checked out. This works if the submodules ID does not change. - Additionally it will detect addition and removal of submodules, which will be handled - gracefully. - - :param previous_commit: If set to a commit'ish, the commit we should use - as the previous commit the HEAD pointed to before it was set to the commit it points to now. - If None, it defaults to ORIG_HEAD otherwise, or the parent of the current - commit if it is not given - :param recursive: if True, the children of submodules will be updated as well - using the same technique - :param force_remove: If submodules have been deleted, they will be forcibly removed. - Otherwise the update may fail if a submodule's repository cannot be deleted as - changes have been made to it (see Submodule.update() for more information) - :param init: If we encounter a new module which would need to be initialized, then do it. - :param to_latest_revision: If True, instead of checking out the revision pointed to - by this submodule's sha, the checked out tracking branch will be merged with the - newest remote branch fetched from the repository's origin""" - if self.repo.bare: - raise InvalidGitRepositoryError("Cannot update submodules in bare repositories") - # END handle bare - - repo = self.repo - - # HANDLE COMMITS - ################## - cur_commit = repo.head.commit - if previous_commit is None: - symref = repo.head.orig_head() - try: - previous_commit = symref.commit - except Exception: - pcommits = cur_commit.parents - if pcommits: - previous_commit = pcommits[0] - else: - # in this special case, we just diff against ourselve, which - # means exactly no change - previous_commit = cur_commit - # END handle initial commit - # END no ORIG_HEAD - else: - previous_commit = repo.commit(previous_commit) # obtain commit object - # END handle previous commit - - - psms = self.list_items(repo, parent_commit=previous_commit) - sms = self.list_items(self.module()) - spsms = set(psms) - ssms = set(sms) - - # HANDLE REMOVALS - ################### - for rsm in (spsms - ssms): - # fake it into thinking its at the current commit to allow deletion - # of previous module. Trigger the cache to be updated before that - #rsm.url - rsm._parent_commit = repo.head.commit - rsm.remove(configuration=False, module=True, force=force_remove) - # END for each removed submodule - - # HANDLE PATH RENAMES - ##################### - # url changes + branch changes - for csm in (spsms & ssms): - psm = psms[csm.name] - sm = sms[csm.name] - - if sm.path != psm.path and psm.module_exists(): - # move the module to the new path - psm.move(sm.path, module=True, configuration=False) - # END handle path changes - - if sm.module_exists(): - # handle url change - if sm.url != psm.url: - # Add the new remote, remove the old one - # This way, if the url just changes, the commits will not - # have to be re-retrieved - nn = '__new_origin__' - smm = sm.module() - rmts = smm.remotes - - # don't do anything if we already have the url we search in place - if len([r for r in rmts if r.url == sm.url]) == 0: - - - assert nn not in [r.name for r in rmts] - smr = smm.create_remote(nn, sm.url) - smr.fetch() - - # If we have a tracking branch, it should be available - # in the new remote as well. - if len([r for r in smr.refs if r.remote_head == sm.branch.name]) == 0: - raise ValueError("Submodule branch named %r was not available in new submodule remote at %r" % (sm.branch.name, sm.url)) - # END head is not detached - - # now delete the changed one - rmt_for_deletion = None - for remote in rmts: - if remote.url == psm.url: - rmt_for_deletion = remote - break - # END if urls match - # END for each remote - - # if we didn't find a matching remote, but have exactly one, - # we can safely use this one - if rmt_for_deletion is None: - if len(rmts) == 1: - rmt_for_deletion = rmts[0] - else: - # if we have not found any remote with the original url - # we may not have a name. This is a special case, - # and its okay to fail here - # Alternatively we could just generate a unique name and leave all - # existing ones in place - raise InvalidGitRepositoryError("Couldn't find original remote-repo at url %r" % psm.url) - #END handle one single remote - # END handle check we found a remote - - orig_name = rmt_for_deletion.name - smm.delete_remote(rmt_for_deletion) - # NOTE: Currently we leave tags from the deleted remotes - # as well as separate tracking branches in the possibly totally - # changed repository ( someone could have changed the url to - # another project ). At some point, one might want to clean - # it up, but the danger is high to remove stuff the user - # has added explicitly - - # rename the new remote back to what it was - smr.rename(orig_name) - - # early on, we verified that the our current tracking branch - # exists in the remote. Now we have to assure that the - # sha we point to is still contained in the new remote - # tracking branch. - smsha = sm.binsha - found = False - rref = smr.refs[self.branch.name] - for c in rref.commit.traverse(): - if c.binsha == smsha: - found = True - break - # END traverse all commits in search for sha - # END for each commit - - if not found: - # adjust our internal binsha to use the one of the remote - # this way, it will be checked out in the next step - # This will change the submodule relative to us, so - # the user will be able to commit the change easily - print >> sys.stderr, "WARNING: Current sha %s was not contained in the tracking branch at the new remote, setting it the the remote's tracking branch" % sm.hexsha - sm.binsha = rref.commit.binsha - #END reset binsha - - #NOTE: All checkout is performed by the base implementation of update - - # END skip remote handling if new url already exists in module - # END handle url - - if sm.branch != psm.branch: - # finally, create a new tracking branch which tracks the - # new remote branch - smm = sm.module() - smmr = smm.remotes - try: - tbr = git.Head.create(smm, sm.branch.name) - except git.GitCommandError, e: - if e.status != 128: - raise - #END handle something unexpected - - # ... or reuse the existing one - tbr = git.Head(smm, git.Head.to_full_path(sm.branch.name)) - #END assure tracking branch exists - - tbr.set_tracking_branch(find_first_remote_branch(smmr, sm.branch)) - # figure out whether the previous tracking branch contains - # new commits compared to the other one, if not we can - # delete it. - try: - tbr = find_first_remote_branch(smmr, psm.branch) - if len(smm.git.cherry(tbr, psm.branch)) == 0: - psm.branch.delete(smm, psm.branch) - #END delete original tracking branch if there are no changes - except InvalidGitRepositoryError: - # ignore it if the previous branch couldn't be found in the - # current remotes, this just means we can't handle it - pass - # END exception handling - - #NOTE: All checkout is done in the base implementation of update - - #END handle branch - #END handle - # END for each common submodule - - # FINALLY UPDATE ALL ACTUAL SUBMODULES - ###################################### - for sm in sms: - # update the submodule using the default method - sm.update(recursive=True, init=init, to_latest_revision=to_latest_revision) - - # update recursively depth first - question is which inconsitent - # state will be better in case it fails somewhere. Defective branch - # or defective depth. The RootSubmodule type will never process itself, - # which was done in the previous expression - if recursive: - type(self)(sm.module()).update(recursive=True, force_remove=force_remove, - init=init, to_latest_revision=to_latest_revision) - #END handle recursive - # END for each submodule to update - - def module(self): - """:return: the actual repository containing the submodules""" - return self.repo - #} END interface -#} END classes diff --git a/lib/git/objects/submodule/__init__.py b/lib/git/objects/submodule/__init__.py new file mode 100644 index 00000000..24663658 --- /dev/null +++ b/lib/git/objects/submodule/__init__.py @@ -0,0 +1,3 @@ + +from base import * +from root import * diff --git a/lib/git/objects/submodule/base.py b/lib/git/objects/submodule/base.py new file mode 100644 index 00000000..6cdc57a0 --- /dev/null +++ b/lib/git/objects/submodule/base.py @@ -0,0 +1,794 @@ +import git.objects.base +from util import * +from git.objects.util import Traversable +from StringIO import StringIO # need a dict to set bloody .name field +from git.util import Iterable, join_path_native, to_native_path_linux +from git.config import SectionConstraint +from git.exc import InvalidGitRepositoryError, NoSuchPathError +import stat +import git + +import os +import sys + +import shutil + +__all__ = ("Submodule", "RootModule") + + + +class Submodule(git.objects.base.IndexObject, Iterable, Traversable): + """Implements access to a git submodule. They are special in that their sha + represents a commit in the submodule's repository which is to be checked out + at the path of this instance. + The submodule type does not have a string type associated with it, as it exists + solely as a marker in the tree and index. + + All methods work in bare and non-bare repositories.""" + + _id_attribute_ = "name" + k_modules_file = '.gitmodules' + k_head_option = 'branch' + k_head_default = 'master' + k_default_mode = stat.S_IFDIR | stat.S_IFLNK # submodules are directories with link-status + + # this is a bogus type for base class compatability + type = 'submodule' + + __slots__ = ('_parent_commit', '_url', '_branch', '_name', '__weakref__') + _cache_attrs = ('path', '_url', '_branch') + + def __init__(self, repo, binsha, mode=None, path=None, name = None, parent_commit=None, url=None, branch=None): + """Initialize this instance with its attributes. We only document the ones + that differ from ``IndexObject`` + :param repo: Our parent repository + :param binsha: binary sha referring to a commit in the remote repository, see url parameter + :param parent_commit: see set_parent_commit() + :param url: The url to the remote repository which is the submodule + :param branch: Head instance to checkout when cloning the remote repository""" + super(Submodule, self).__init__(repo, binsha, mode, path) + self.size = 0 + if parent_commit is not None: + self._parent_commit = parent_commit + if url is not None: + self._url = url + if branch is not None: + assert isinstance(branch, git.Head) + self._branch = branch + if name is not None: + self._name = name + + def _set_cache_(self, attr): + if attr == '_parent_commit': + # set a default value, which is the root tree of the current head + self._parent_commit = self.repo.commit() + elif attr in ('path', '_url', '_branch'): + reader = self.config_reader() + # default submodule values + self.path = reader.get_value('path') + self._url = reader.get_value('url') + # git-python extension values - optional + self._branch = mkhead(self.repo, reader.get_value(self.k_head_option, self.k_head_default)) + elif attr == '_name': + raise AttributeError("Cannot retrieve the name of a submodule if it was not set initially") + else: + super(Submodule, self)._set_cache_(attr) + # END handle attribute name + + def _get_intermediate_items(self, item): + """:return: all the submodules of our module repository""" + try: + return type(self).list_items(item.module()) + except InvalidGitRepositoryError: + return list() + # END handle intermeditate items + + def __eq__(self, other): + """Compare with another submodule""" + # we may only compare by name as this should be the ID they are hashed with + # Otherwise this type wouldn't be hashable + # return self.path == other.path and self.url == other.url and super(Submodule, self).__eq__(other) + return self._name == other._name + + def __ne__(self, other): + """Compare with another submodule for inequality""" + return not (self == other) + + def __hash__(self): + """Hash this instance using its logical id, not the sha""" + return hash(self._name) + + def __str__(self): + return self._name + + def __repr__(self): + return "git.%s(name=%s, path=%s, url=%s, branch=%s)" % (type(self).__name__, self._name, self.path, self.url, self.branch) + + @classmethod + def _config_parser(cls, repo, parent_commit, read_only): + """:return: Config Parser constrained to our submodule in read or write mode + :raise IOError: If the .gitmodules file cannot be found, either locally or in the repository + at the given parent commit. Otherwise the exception would be delayed until the first + access of the config parser""" + parent_matches_head = repo.head.commit == parent_commit + if not repo.bare and parent_matches_head: + fp_module = cls.k_modules_file + fp_module_path = os.path.join(repo.working_tree_dir, fp_module) + if not os.path.isfile(fp_module_path): + raise IOError("%s file was not accessible" % fp_module_path) + # END handle existance + fp_module = fp_module_path + else: + try: + fp_module = cls._sio_modules(parent_commit) + except KeyError: + raise IOError("Could not find %s file in the tree of parent commit %s" % (cls.k_modules_file, parent_commit)) + # END handle exceptions + # END handle non-bare working tree + + if not read_only and (repo.bare or not parent_matches_head): + raise ValueError("Cannot write blobs of 'historical' submodule configurations") + # END handle writes of historical submodules + + return SubmoduleConfigParser(fp_module, read_only = read_only) + + def _clear_cache(self): + # clear the possibly changed values + for name in self._cache_attrs: + try: + delattr(self, name) + except AttributeError: + pass + # END try attr deletion + # END for each name to delete + + @classmethod + def _sio_modules(cls, parent_commit): + """:return: Configuration file as StringIO - we only access it through the respective blob's data""" + sio = StringIO(parent_commit.tree[cls.k_modules_file].data_stream.read()) + sio.name = cls.k_modules_file + return sio + + def _config_parser_constrained(self, read_only): + """:return: Config Parser constrained to our submodule in read or write mode""" + parser = self._config_parser(self.repo, self._parent_commit, read_only) + parser.set_submodule(self) + return SectionConstraint(parser, sm_section(self.name)) + + #{ Edit Interface + + @classmethod + def add(cls, repo, name, path, url=None, branch=None, no_checkout=False): + """Add a new submodule to the given repository. This will alter the index + as well as the .gitmodules file, but will not create a new commit. + If the submodule already exists, no matter if the configuration differs + from the one provided, the existing submodule will be returned. + :param repo: Repository instance which should receive the submodule + :param name: The name/identifier for the submodule + :param path: repository-relative or absolute path at which the submodule + should be located + It will be created as required during the repository initialization. + :param url: git-clone compatible URL, see git-clone reference for more information + If None, the repository is assumed to exist, and the url of the first + remote is taken instead. This is useful if you want to make an existing + repository a submodule of anotherone. + :param branch: branch at which the submodule should (later) be checked out. + The given branch must exist in the remote repository, and will be checked + out locally as a tracking branch. + It will only be written into the configuration if it not None, which is + when the checked out branch will be the one the remote HEAD pointed to. + The result you get in these situation is somewhat fuzzy, and it is recommended + to specify at least 'master' here + :param no_checkout: if True, and if the repository has to be cloned manually, + no checkout will be performed + :return: The newly created submodule instance + :note: works atomically, such that no change will be done if the repository + update fails for instance""" + if repo.bare: + raise InvalidGitRepositoryError("Cannot add submodules to bare repositories") + # END handle bare repos + + path = to_native_path_linux(path) + if path.endswith('/'): + path = path[:-1] + # END handle trailing slash + + # INSTANTIATE INTERMEDIATE SM + sm = cls(repo, cls.NULL_BIN_SHA, cls.k_default_mode, path, name) + if sm.exists(): + # reretrieve submodule from tree + return repo.head.commit.tree[path] + # END handle existing + + br = mkhead(repo, branch or cls.k_head_default) + has_module = sm.module_exists() + branch_is_default = branch is None + if has_module and url is not None: + if url not in [r.url for r in sm.module().remotes]: + raise ValueError("Specified URL '%s' does not match any remote url of the repository at '%s'" % (url, sm.abspath)) + # END check url + # END verify urls match + + mrepo = None + if url is None: + if not has_module: + raise ValueError("A URL was not given and existing repository did not exsit at %s" % path) + # END check url + mrepo = sm.module() + urls = [r.url for r in mrepo.remotes] + if not urls: + raise ValueError("Didn't find any remote url in repository at %s" % sm.abspath) + # END verify we have url + url = urls[0] + else: + # clone new repo + kwargs = {'n' : no_checkout} + if not branch_is_default: + kwargs['b'] = str(br) + # END setup checkout-branch + mrepo = git.Repo.clone_from(url, path, **kwargs) + # END verify url + + # update configuration and index + index = sm.repo.index + writer = sm.config_writer(index=index, write=False) + writer.set_value('url', url) + writer.set_value('path', path) + + sm._url = url + if not branch_is_default: + # store full path + writer.set_value(cls.k_head_option, br.path) + sm._branch = br.path + # END handle path + del(writer) + + # NOTE: Have to write the repo config file as well, otherwise + # the default implementation will be offended and not update the repository + # Maybe this is a good way to assure it doesn't get into our way, but + # we want to stay backwards compatible too ... . Its so redundant ! + repo.config_writer().set_value(sm_section(sm.name), 'url', url) + + # we deliberatly assume that our head matches our index ! + pcommit = repo.head.commit + sm._parent_commit = pcommit + sm.binsha = mrepo.head.commit.binsha + index.add([sm], write=True) + + return sm + + def update(self, recursive=False, init=True, to_latest_revision=False): + """Update the repository of this submodule to point to the checkout + we point at with the binsha of this instance. + :param recursive: if True, we will operate recursively and update child- + modules as well. + :param init: if True, the module repository will be cloned into place if necessary + :param to_latest_revision: if True, the submodule's sha will be ignored during checkout. + Instead, the remote will be fetched, and the local tracking branch updated. + This only works if we have a local tracking branch, which is the case + if the remote repository had a master branch, or of the 'branch' option + was specified for this submodule and the branch existed remotely + :note: does nothing in bare repositories + :note: method is definitely not atomic if recurisve is True + :return: self""" + if self.repo.bare: + return self + #END pass in bare mode + + + # ASSURE REPO IS PRESENT AND UPTODATE + ##################################### + try: + mrepo = self.module() + for remote in mrepo.remotes: + remote.fetch() + #END fetch new data + except InvalidGitRepositoryError: + if not init: + return self + # END early abort if init is not allowed + import git + + # there is no git-repository yet - but delete empty paths + module_path = join_path_native(self.repo.working_tree_dir, self.path) + if os.path.isdir(module_path): + try: + os.rmdir(module_path) + except OSError: + raise OSError("Module directory at %r does already exist and is non-empty" % module_path) + # END handle OSError + # END handle directory removal + + # don't check it out at first - nonetheless it will create a local + # branch according to the remote-HEAD if possible + mrepo = git.Repo.clone_from(self.url, module_path, n=True) + + # see whether we have a valid branch to checkout + try: + # find a remote which has our branch - we try to be flexible + remote_branch = find_first_remote_branch(mrepo.remotes, self.branch) + local_branch = self.branch + if not local_branch.is_valid(): + # Setup a tracking configuration - branch doesn't need to + # exist to do that + local_branch.set_tracking_branch(remote_branch) + #END handle local branch + + # have a valid branch, but no checkout - make sure we can figure + # that out by marking the commit with a null_sha + # have to write it directly as .commit = NULLSHA tries to resolve the sha + # This will bring the branch into existance + refpath = join_path_native(mrepo.git_dir, local_branch.path) + refdir = os.path.dirname(refpath) + if not os.path.isdir(refdir): + os.makedirs(refdir) + #END handle directory + open(refpath, 'w').write(self.NULL_HEX_SHA) + # END initial checkout + branch creation + + # make sure HEAD is not detached + mrepo.head.ref = local_branch + except IndexError: + print >> sys.stderr, "Warning: Failed to checkout tracking branch %s" % self.branch + #END handle tracking branch + #END handle initalization + + + # DETERMINE SHAS TO CHECKOUT + ############################ + binsha = self.binsha + hexsha = self.hexsha + is_detached = mrepo.head.is_detached + if to_latest_revision: + msg_base = "Cannot update to latest revision in repository at %r as " % mrepo.working_dir + if not is_detached: + rref = mrepo.head.ref.tracking_branch() + if rref is not None: + rcommit = rref.commit + binsha = rcommit.binsha + hexsha = rcommit.hexsha + else: + print >> sys.stderr, "%s a tracking branch was not set for local branch '%s'" % (msg_base, mrepo.head.ref) + # END handle remote ref + else: + print >> sys.stderr, "%s there was no local tracking branch" % msg_base + # END handle detached head + # END handle to_latest_revision option + + # update the working tree + if mrepo.head.commit.binsha != binsha: + if is_detached: + mrepo.git.checkout(hexsha) + else: + # TODO: allow to specify a rebase, merge, or reset + # TODO: Warn if the hexsha forces the tracking branch off the remote + # branch - this should be prevented when setting the branch option + mrepo.head.reset(hexsha, index=True, working_tree=True) + # END handle checkout + # END update to new commit only if needed + + # HANDLE RECURSION + ################## + if recursive: + for submodule in self.iter_items(self.module()): + submodule.update(recursive, init, to_latest_revision) + # END handle recursive update + # END for each submodule + + return self + + @unbare_repo + def move(self, module_path, configuration=True, module=True): + """Move the submodule to a another module path. This involves physically moving + the repository at our current path, changing the configuration, as well as + adjusting our index entry accordingly. + :param module_path: the path to which to move our module, given as + repository-relative path. Intermediate directories will be created + accordingly. If the path already exists, it must be empty. + Trailling (back)slashes are removed automatically + :param configuration: if True, the configuration will be adjusted to let + the submodule point to the given path. + :param module: if True, the repository managed by this submodule + will be moved, not the configuration. This will effectively + leave your repository in an inconsistent state unless the configuration + and index already point to the target location. + :return: self + :raise ValueError: if the module path existed and was not empty, or was a file + :note: Currently the method is not atomic, and it could leave the repository + in an inconsistent state if a sub-step fails for some reason + """ + if module + configuration < 1: + raise ValueError("You must specify to move at least the module or the configuration of the submodule") + #END handle input + + module_path = to_native_path_linux(module_path) + if module_path.endswith('/'): + module_path = module_path[:-1] + # END handle trailing slash + + # VERIFY DESTINATION + if module_path == self.path: + return self + #END handle no change + + dest_path = join_path_native(self.repo.working_tree_dir, module_path) + if os.path.isfile(dest_path): + raise ValueError("Cannot move repository onto a file: %s" % dest_path) + # END handle target files + + index = self.repo.index + tekey = index.entry_key(module_path, 0) + # if the target item already exists, fail + if configuration and tekey in index.entries: + raise ValueError("Index entry for target path did alredy exist") + #END handle index key already there + + # remove existing destination + if module: + if os.path.exists(dest_path): + if len(os.listdir(dest_path)): + raise ValueError("Destination module directory was not empty") + #END handle non-emptyness + + if os.path.islink(dest_path): + os.remove(dest_path) + else: + os.rmdir(dest_path) + #END handle link + else: + # recreate parent directories + # NOTE: renames() does that now + pass + #END handle existance + # END handle module + + # move the module into place if possible + cur_path = self.abspath + renamed_module = False + if module and os.path.exists(cur_path): + os.renames(cur_path, dest_path) + renamed_module = True + #END move physical module + + + # rename the index entry - have to manipulate the index directly as + # git-mv cannot be used on submodules ... yeah + try: + if configuration: + try: + ekey = index.entry_key(self.path, 0) + entry = index.entries[ekey] + del(index.entries[ekey]) + nentry = git.IndexEntry(entry[:3]+(module_path,)+entry[4:]) + index.entries[tekey] = nentry + except KeyError: + raise InvalidGitRepositoryError("Submodule's entry at %r did not exist" % (self.path)) + #END handle submodule doesn't exist + + # update configuration + writer = self.config_writer(index=index) # auto-write + writer.set_value('path', module_path) + self.path = module_path + del(writer) + # END handle configuration flag + except Exception: + if renamed_module: + os.renames(dest_path, cur_path) + # END undo module renaming + raise + #END handle undo rename + + return self + + @unbare_repo + def remove(self, module=True, force=False, configuration=True, dry_run=False): + """Remove this submodule from the repository. This will remove our entry + from the .gitmodules file and the entry in the .git/config file. + :param module: If True, the module we point to will be deleted + as well. If the module is currently on a commit which is not part + of any branch in the remote, if the currently checked out branch + is ahead of its tracking branch, if you have modifications in the + working tree, or untracked files, + In case the removal of the repository fails for these reasons, the + submodule status will not have been altered. + If this submodule has child-modules on its own, these will be deleted + prior to touching the own module. + :param force: Enforces the deletion of the module even though it contains + modifications. This basically enforces a brute-force file system based + deletion. + :param configuration: if True, the submodule is deleted from the configuration, + otherwise it isn't. Although this should be enabled most of the times, + this flag enables you to safely delete the repository of your submodule. + :param dry_run: if True, we will not actually do anything, but throw the errors + we would usually throw + :return: self + :note: doesn't work in bare repositories + :raise InvalidGitRepositoryError: thrown if the repository cannot be deleted + :raise OSError: if directories or files could not be removed""" + if not (module + configuration): + raise ValueError("Need to specify to delete at least the module, or the configuration") + # END handle params + + # DELETE MODULE REPOSITORY + ########################## + if module and self.module_exists(): + if force: + # take the fast lane and just delete everything in our module path + # TODO: If we run into permission problems, we have a highly inconsistent + # state. Delete the .git folders last, start with the submodules first + mp = self.abspath + method = None + if os.path.islink(mp): + method = os.remove + elif os.path.isdir(mp): + method = shutil.rmtree + elif os.path.exists(mp): + raise AssertionError("Cannot forcibly delete repository as it was neither a link, nor a directory") + #END handle brutal deletion + if not dry_run: + assert method + method(mp) + #END apply deletion method + else: + # verify we may delete our module + mod = self.module() + if mod.is_dirty(untracked_files=True): + raise InvalidGitRepositoryError("Cannot delete module at %s with any modifications, unless force is specified" % mod.working_tree_dir) + # END check for dirt + + # figure out whether we have new commits compared to the remotes + # NOTE: If the user pulled all the time, the remote heads might + # not have been updated, so commits coming from the remote look + # as if they come from us. But we stay strictly read-only and + # don't fetch beforhand. + for remote in mod.remotes: + num_branches_with_new_commits = 0 + rrefs = remote.refs + for rref in rrefs: + num_branches_with_new_commits = len(mod.git.cherry(rref)) != 0 + # END for each remote ref + # not a single remote branch contained all our commits + if num_branches_with_new_commits == len(rrefs): + raise InvalidGitRepositoryError("Cannot delete module at %s as there are new commits" % mod.working_tree_dir) + # END handle new commits + # END for each remote + + # gently remove all submodule repositories + for sm in self.children(): + sm.remove(module=True, force=False, configuration=False, dry_run=dry_run) + # END for each child-submodule + + # finally delete our own submodule + if not dry_run: + shutil.rmtree(mod.working_tree_dir) + # END delete tree if possible + # END handle force + # END handle module deletion + + # DELETE CONFIGURATION + ###################### + if configuration and not dry_run: + # first the index-entry + index = self.repo.index + try: + del(index.entries[index.entry_key(self.path, 0)]) + except KeyError: + pass + #END delete entry + index.write() + + # now git config - need the config intact, otherwise we can't query + # inforamtion anymore + self.repo.config_writer().remove_section(sm_section(self.name)) + self.config_writer().remove_section() + # END delete configuration + + return self + + def set_parent_commit(self, commit, check=True): + """Set this instance to use the given commit whose tree is supposed to + contain the .gitmodules blob. + :param commit: Commit'ish reference pointing at the root_tree + :param check: if True, relatively expensive checks will be performed to verify + validity of the submodule. + :raise ValueError: if the commit's tree didn't contain the .gitmodules blob. + :raise ValueError: if the parent commit didn't store this submodule under the + current path + :return: self""" + pcommit = self.repo.commit(commit) + pctree = pcommit.tree + if self.k_modules_file not in pctree: + raise ValueError("Tree of commit %s did not contain the %s file" % (commit, self.k_modules_file)) + # END handle exceptions + + prev_pc = self._parent_commit + self._parent_commit = pcommit + + if check: + parser = self._config_parser(self.repo, self._parent_commit, read_only=True) + if not parser.has_section(sm_section(self.name)): + self._parent_commit = prev_pc + raise ValueError("Submodule at path %r did not exist in parent commit %s" % (self.path, commit)) + # END handle submodule did not exist + # END handle checking mode + + # update our sha, it could have changed + self.binsha = pctree[self.path].binsha + + self._clear_cache() + + return self + + @unbare_repo + def config_writer(self, index=None, write=True): + """:return: a config writer instance allowing you to read and write the data + belonging to this submodule into the .gitmodules file. + + :param index: if not None, an IndexFile instance which should be written. + defaults to the index of the Submodule's parent repository. + :param write: if True, the index will be written each time a configuration + value changes. + :note: the parameters allow for a more efficient writing of the index, + as you can pass in a modified index on your own, prevent automatic writing, + and write yourself once the whole operation is complete + :raise ValueError: if trying to get a writer on a parent_commit which does not + match the current head commit + :raise IOError: If the .gitmodules file/blob could not be read""" + writer = self._config_parser_constrained(read_only=False) + if index is not None: + writer.config._index = index + writer.config._auto_write = write + return writer + + #} END edit interface + + #{ Query Interface + + @unbare_repo + def module(self): + """:return: Repo instance initialized from the repository at our submodule path + :raise InvalidGitRepositoryError: if a repository was not available. This could + also mean that it was not yet initialized""" + # late import to workaround circular dependencies + module_path = self.abspath + try: + repo = git.Repo(module_path) + if repo != self.repo: + return repo + # END handle repo uninitialized + except (InvalidGitRepositoryError, NoSuchPathError): + raise InvalidGitRepositoryError("No valid repository at %s" % self.path) + else: + raise InvalidGitRepositoryError("Repository at %r was not yet checked out" % module_path) + # END handle exceptions + + def module_exists(self): + """:return: True if our module exists and is a valid git repository. See module() method""" + try: + self.module() + return True + except Exception: + return False + # END handle exception + + def exists(self): + """:return: True if the submodule exists, False otherwise. Please note that + a submodule may exist (in the .gitmodules file) even though its module + doesn't exist""" + # keep attributes for later, and restore them if we have no valid data + # this way we do not actually alter the state of the object + loc = locals() + for attr in self._cache_attrs: + if hasattr(self, attr): + loc[attr] = getattr(self, attr) + # END if we have the attribute cache + #END for each attr + self._clear_cache() + + try: + try: + self.path + return True + except Exception: + return False + # END handle exceptions + finally: + for attr in self._cache_attrs: + if attr in loc: + setattr(self, attr, loc[attr]) + # END if we have a cache + # END reapply each attribute + # END handle object state consistency + + @property + def branch(self): + """:return: The branch instance that we are to checkout""" + return self._branch + + @property + def url(self): + """:return: The url to the repository which our module-repository refers to""" + return self._url + + @property + def parent_commit(self): + """:return: Commit instance with the tree containing the .gitmodules file + :note: will always point to the current head's commit if it was not set explicitly""" + return self._parent_commit + + @property + def name(self): + """:return: The name of this submodule. It is used to identify it within the + .gitmodules file. + :note: by default, the name is the path at which to find the submodule, but + in git-python it should be a unique identifier similar to the identifiers + used for remotes, which allows to change the path of the submodule + easily + """ + return self._name + + def config_reader(self): + """:return: ConfigReader instance which allows you to qurey the configuration values + of this submodule, as provided by the .gitmodules file + :note: The config reader will actually read the data directly from the repository + and thus does not need nor care about your working tree. + :note: Should be cached by the caller and only kept as long as needed + :raise IOError: If the .gitmodules file/blob could not be read""" + return self._config_parser_constrained(read_only=True) + + def children(self): + """:return: IterableList(Submodule, ...) an iterable list of submodules instances + which are children of this submodule + :raise InvalidGitRepositoryError: if the submodule is not checked-out""" + return self._get_intermediate_items(self) + + #} END query interface + + #{ Iterable Interface + + @classmethod + def iter_items(cls, repo, parent_commit='HEAD'): + """:return: iterator yielding Submodule instances available in the given repository""" + pc = repo.commit(parent_commit) # parent commit instance + try: + parser = cls._config_parser(repo, pc, read_only=True) + except IOError: + raise StopIteration + # END handle empty iterator + + rt = pc.tree # root tree + + for sms in parser.sections(): + n = sm_name(sms) + p = parser.get_value(sms, 'path') + u = parser.get_value(sms, 'url') + b = cls.k_head_default + if parser.has_option(sms, cls.k_head_option): + b = parser.get_value(sms, cls.k_head_option) + # END handle optional information + + # get the binsha + index = repo.index + try: + sm = rt[p] + except KeyError: + # try the index, maybe it was just added + try: + entry = index.entries[index.entry_key(p, 0)] + sm = cls(repo, entry.binsha, entry.mode, entry.path) + except KeyError: + raise InvalidGitRepositoryError("Gitmodule path %r did not exist in revision of parent commit %s" % (p, parent_commit)) + # END handle keyerror + # END handle critical error + + # fill in remaining info - saves time as it doesn't have to be parsed again + sm._name = n + sm._parent_commit = pc + sm._branch = mkhead(repo, b) + sm._url = u + + yield sm + # END for each section + + #} END iterable interface + diff --git a/lib/git/objects/submodule/root.py b/lib/git/objects/submodule/root.py new file mode 100644 index 00000000..2e02e7de --- /dev/null +++ b/lib/git/objects/submodule/root.py @@ -0,0 +1,259 @@ +from base import Submodule +from git.exc import InvalidGitRepositoryError +import git + +import sys + +__all__ = ["RootModule"] + + +class RootModule(Submodule): + """A (virtual) Root of all submodules in the given repository. It can be used + to more easily traverse all submodules of the master repository""" + + __slots__ = tuple() + + k_root_name = '__ROOT__' + + def __init__(self, repo): + # repo, binsha, mode=None, path=None, name = None, parent_commit=None, url=None, ref=None) + super(RootModule, self).__init__( + repo, + binsha = self.NULL_BIN_SHA, + mode = self.k_default_mode, + path = '', + name = self.k_root_name, + parent_commit = repo.head.commit, + url = '', + branch = mkhead(repo, self.k_head_default) + ) + + + def _clear_cache(self): + """May not do anything""" + pass + + #{ Interface + + def update(self, previous_commit=None, recursive=True, force_remove=False, init=True, to_latest_revision=False): + """Update the submodules of this repository to the current HEAD commit. + This method behaves smartly by determining changes of the path of a submodules + repository, next to changes to the to-be-checked-out commit or the branch to be + checked out. This works if the submodules ID does not change. + Additionally it will detect addition and removal of submodules, which will be handled + gracefully. + + :param previous_commit: If set to a commit'ish, the commit we should use + as the previous commit the HEAD pointed to before it was set to the commit it points to now. + If None, it defaults to ORIG_HEAD otherwise, or the parent of the current + commit if it is not given + :param recursive: if True, the children of submodules will be updated as well + using the same technique + :param force_remove: If submodules have been deleted, they will be forcibly removed. + Otherwise the update may fail if a submodule's repository cannot be deleted as + changes have been made to it (see Submodule.update() for more information) + :param init: If we encounter a new module which would need to be initialized, then do it. + :param to_latest_revision: If True, instead of checking out the revision pointed to + by this submodule's sha, the checked out tracking branch will be merged with the + newest remote branch fetched from the repository's origin""" + if self.repo.bare: + raise InvalidGitRepositoryError("Cannot update submodules in bare repositories") + # END handle bare + + repo = self.repo + + # HANDLE COMMITS + ################## + cur_commit = repo.head.commit + if previous_commit is None: + symref = repo.head.orig_head() + try: + previous_commit = symref.commit + except Exception: + pcommits = cur_commit.parents + if pcommits: + previous_commit = pcommits[0] + else: + # in this special case, we just diff against ourselve, which + # means exactly no change + previous_commit = cur_commit + # END handle initial commit + # END no ORIG_HEAD + else: + previous_commit = repo.commit(previous_commit) # obtain commit object + # END handle previous commit + + + psms = self.list_items(repo, parent_commit=previous_commit) + sms = self.list_items(self.module()) + spsms = set(psms) + ssms = set(sms) + + # HANDLE REMOVALS + ################### + for rsm in (spsms - ssms): + # fake it into thinking its at the current commit to allow deletion + # of previous module. Trigger the cache to be updated before that + #rsm.url + rsm._parent_commit = repo.head.commit + rsm.remove(configuration=False, module=True, force=force_remove) + # END for each removed submodule + + # HANDLE PATH RENAMES + ##################### + # url changes + branch changes + for csm in (spsms & ssms): + psm = psms[csm.name] + sm = sms[csm.name] + + if sm.path != psm.path and psm.module_exists(): + # move the module to the new path + psm.move(sm.path, module=True, configuration=False) + # END handle path changes + + if sm.module_exists(): + # handle url change + if sm.url != psm.url: + # Add the new remote, remove the old one + # This way, if the url just changes, the commits will not + # have to be re-retrieved + nn = '__new_origin__' + smm = sm.module() + rmts = smm.remotes + + # don't do anything if we already have the url we search in place + if len([r for r in rmts if r.url == sm.url]) == 0: + + + assert nn not in [r.name for r in rmts] + smr = smm.create_remote(nn, sm.url) + smr.fetch() + + # If we have a tracking branch, it should be available + # in the new remote as well. + if len([r for r in smr.refs if r.remote_head == sm.branch.name]) == 0: + raise ValueError("Submodule branch named %r was not available in new submodule remote at %r" % (sm.branch.name, sm.url)) + # END head is not detached + + # now delete the changed one + rmt_for_deletion = None + for remote in rmts: + if remote.url == psm.url: + rmt_for_deletion = remote + break + # END if urls match + # END for each remote + + # if we didn't find a matching remote, but have exactly one, + # we can safely use this one + if rmt_for_deletion is None: + if len(rmts) == 1: + rmt_for_deletion = rmts[0] + else: + # if we have not found any remote with the original url + # we may not have a name. This is a special case, + # and its okay to fail here + # Alternatively we could just generate a unique name and leave all + # existing ones in place + raise InvalidGitRepositoryError("Couldn't find original remote-repo at url %r" % psm.url) + #END handle one single remote + # END handle check we found a remote + + orig_name = rmt_for_deletion.name + smm.delete_remote(rmt_for_deletion) + # NOTE: Currently we leave tags from the deleted remotes + # as well as separate tracking branches in the possibly totally + # changed repository ( someone could have changed the url to + # another project ). At some point, one might want to clean + # it up, but the danger is high to remove stuff the user + # has added explicitly + + # rename the new remote back to what it was + smr.rename(orig_name) + + # early on, we verified that the our current tracking branch + # exists in the remote. Now we have to assure that the + # sha we point to is still contained in the new remote + # tracking branch. + smsha = sm.binsha + found = False + rref = smr.refs[self.branch.name] + for c in rref.commit.traverse(): + if c.binsha == smsha: + found = True + break + # END traverse all commits in search for sha + # END for each commit + + if not found: + # adjust our internal binsha to use the one of the remote + # this way, it will be checked out in the next step + # This will change the submodule relative to us, so + # the user will be able to commit the change easily + print >> sys.stderr, "WARNING: Current sha %s was not contained in the tracking branch at the new remote, setting it the the remote's tracking branch" % sm.hexsha + sm.binsha = rref.commit.binsha + #END reset binsha + + #NOTE: All checkout is performed by the base implementation of update + + # END skip remote handling if new url already exists in module + # END handle url + + if sm.branch != psm.branch: + # finally, create a new tracking branch which tracks the + # new remote branch + smm = sm.module() + smmr = smm.remotes + try: + tbr = git.Head.create(smm, sm.branch.name) + except git.GitCommandError, e: + if e.status != 128: + raise + #END handle something unexpected + + # ... or reuse the existing one + tbr = git.Head(smm, git.Head.to_full_path(sm.branch.name)) + #END assure tracking branch exists + + tbr.set_tracking_branch(find_first_remote_branch(smmr, sm.branch)) + # figure out whether the previous tracking branch contains + # new commits compared to the other one, if not we can + # delete it. + try: + tbr = find_first_remote_branch(smmr, psm.branch) + if len(smm.git.cherry(tbr, psm.branch)) == 0: + psm.branch.delete(smm, psm.branch) + #END delete original tracking branch if there are no changes + except InvalidGitRepositoryError: + # ignore it if the previous branch couldn't be found in the + # current remotes, this just means we can't handle it + pass + # END exception handling + + #NOTE: All checkout is done in the base implementation of update + + #END handle branch + #END handle + # END for each common submodule + + # FINALLY UPDATE ALL ACTUAL SUBMODULES + ###################################### + for sm in sms: + # update the submodule using the default method + sm.update(recursive=True, init=init, to_latest_revision=to_latest_revision) + + # update recursively depth first - question is which inconsitent + # state will be better in case it fails somewhere. Defective branch + # or defective depth. The RootSubmodule type will never process itself, + # which was done in the previous expression + if recursive: + type(self)(sm.module()).update(recursive=True, force_remove=force_remove, + init=init, to_latest_revision=to_latest_revision) + #END handle recursive + # END for each submodule to update + + def module(self): + """:return: the actual repository containing the submodules""" + return self.repo + #} END interface +#} END classes diff --git a/lib/git/objects/submodule/util.py b/lib/git/objects/submodule/util.py new file mode 100644 index 00000000..ab5e345a --- /dev/null +++ b/lib/git/objects/submodule/util.py @@ -0,0 +1,101 @@ +import git +from git.exc import InvalidGitRepositoryError +from git.config import GitConfigParser +from StringIO import StringIO +import weakref + +__all__ = ( 'sm_section', 'sm_name', 'mkhead', 'unbare_repo', 'find_first_remote_branch', + 'SubmoduleConfigParser') + +#{ Utilities + +def sm_section(name): + """:return: section title used in .gitmodules configuration file""" + return 'submodule "%s"' % name + +def sm_name(section): + """:return: name of the submodule as parsed from the section name""" + section = section.strip() + return section[11:-1] + +def mkhead(repo, path): + """:return: New branch/head instance""" + return git.Head(repo, git.Head.to_full_path(path)) + +def unbare_repo(func): + """Methods with this decorator raise InvalidGitRepositoryError if they + encounter a bare repository""" + def wrapper(self, *args, **kwargs): + if self.repo.bare: + raise InvalidGitRepositoryError("Method '%s' cannot operate on bare repositories" % func.__name__) + #END bare method + return func(self, *args, **kwargs) + # END wrapper + wrapper.__name__ = func.__name__ + return wrapper + +def find_first_remote_branch(remotes, branch): + """Find the remote branch matching the name of the given branch or raise InvalidGitRepositoryError""" + for remote in remotes: + try: + return remote.refs[branch.name] + except IndexError: + continue + # END exception handling + #END for remote + raise InvalidGitRepositoryError("Didn't find remote branch %r in any of the given remotes", branch) + +#} END utilities + + +#{ Classes + +class SubmoduleConfigParser(GitConfigParser): + """ + Catches calls to _write, and updates the .gitmodules blob in the index + with the new data, if we have written into a stream. Otherwise it will + add the local file to the index to make it correspond with the working tree. + Additionally, the cache must be cleared + + Please note that no mutating method will work in bare mode + """ + + def __init__(self, *args, **kwargs): + self._smref = None + self._index = None + self._auto_write = True + super(SubmoduleConfigParser, self).__init__(*args, **kwargs) + + #{ Interface + def set_submodule(self, submodule): + """Set this instance's submodule. It must be called before + the first write operation begins""" + self._smref = weakref.ref(submodule) + + def flush_to_index(self): + """Flush changes in our configuration file to the index""" + assert self._smref is not None + # should always have a file here + assert not isinstance(self._file_or_files, StringIO) + + sm = self._smref() + if sm is not None: + index = self._index + if index is None: + index = sm.repo.index + # END handle index + index.add([sm.k_modules_file], write=self._auto_write) + sm._clear_cache() + # END handle weakref + + #} END interface + + #{ Overridden Methods + def write(self): + rval = super(SubmoduleConfigParser, self).write() + self.flush_to_index() + return rval + # END overridden methods + + +#} END classes -- cgit v1.2.1 From 9519f186ce757cdba217f222c95c20033d00f91d Mon Sep 17 00:00:00 2001 From: Sebastian Thiel Date: Thu, 18 Nov 2010 23:10:23 +0100 Subject: Fixed all imports, refactoring appears to be complete --- lib/git/objects/__init__.py | 11 +++++++++-- lib/git/objects/submodule/__init__.py | 2 -- lib/git/objects/submodule/base.py | 36 ++++++++++++++++++++++++++++------- lib/git/objects/submodule/root.py | 4 ++++ lib/git/objects/tree.py | 2 +- 5 files changed, 43 insertions(+), 12 deletions(-) (limited to 'lib/git') diff --git a/lib/git/objects/__init__.py b/lib/git/objects/__init__.py index 85c7e38c..e8e0ef39 100644 --- a/lib/git/objects/__init__.py +++ b/lib/git/objects/__init__.py @@ -3,11 +3,18 @@ Import all submodules main classes into the package space """ import inspect from base import * +# Fix import dependency - add IndexObject to the util module, so that it can be +# imported by the submodule.base +import submodule.util +submodule.util.IndexObject = IndexObject +from submodule.base import * +from submodule.root import * + +# must come after submodule was made available from tag import * from blob import * -from tree import * from commit import * -from submodule import * +from tree import * from util import Actor __all__ = [ name for name, obj in locals().items() diff --git a/lib/git/objects/submodule/__init__.py b/lib/git/objects/submodule/__init__.py index 24663658..8b137891 100644 --- a/lib/git/objects/submodule/__init__.py +++ b/lib/git/objects/submodule/__init__.py @@ -1,3 +1 @@ -from base import * -from root import * diff --git a/lib/git/objects/submodule/base.py b/lib/git/objects/submodule/base.py index 6cdc57a0..347af58e 100644 --- a/lib/git/objects/submodule/base.py +++ b/lib/git/objects/submodule/base.py @@ -1,10 +1,24 @@ -import git.objects.base -from util import * +import util +from util import ( + mkhead, + sm_name, + sm_section, + unbare_repo, + SubmoduleConfigParser, + find_first_remote_branch + ) from git.objects.util import Traversable from StringIO import StringIO # need a dict to set bloody .name field -from git.util import Iterable, join_path_native, to_native_path_linux +from git.util import ( + Iterable, + join_path_native, + to_native_path_linux + ) from git.config import SectionConstraint -from git.exc import InvalidGitRepositoryError, NoSuchPathError +from git.exc import ( + InvalidGitRepositoryError, + NoSuchPathError + ) import stat import git @@ -13,11 +27,13 @@ import sys import shutil -__all__ = ("Submodule", "RootModule") +__all__ = ["Submodule"] - -class Submodule(git.objects.base.IndexObject, Iterable, Traversable): +# IndexObject comes via util module, its a 'hacky' fix thanks to pythons import +# mechanism which cause plenty of trouble of the only reason for packages and +# modules is refactoring - subpackages shoudn't depend on parent packages +class Submodule(util.IndexObject, Iterable, Traversable): """Implements access to a git submodule. They are special in that their sha represents a commit in the submodule's repository which is to be checked out at the path of this instance. @@ -41,6 +57,7 @@ class Submodule(git.objects.base.IndexObject, Iterable, Traversable): def __init__(self, repo, binsha, mode=None, path=None, name = None, parent_commit=None, url=None, branch=None): """Initialize this instance with its attributes. We only document the ones that differ from ``IndexObject`` + :param repo: Our parent repository :param binsha: binary sha referring to a commit in the remote repository, see url parameter :param parent_commit: see set_parent_commit() @@ -163,6 +180,7 @@ class Submodule(git.objects.base.IndexObject, Iterable, Traversable): as well as the .gitmodules file, but will not create a new commit. If the submodule already exists, no matter if the configuration differs from the one provided, the existing submodule will be returned. + :param repo: Repository instance which should receive the submodule :param name: The name/identifier for the submodule :param path: repository-relative or absolute path at which the submodule @@ -260,6 +278,7 @@ class Submodule(git.objects.base.IndexObject, Iterable, Traversable): def update(self, recursive=False, init=True, to_latest_revision=False): """Update the repository of this submodule to point to the checkout we point at with the binsha of this instance. + :param recursive: if True, we will operate recursively and update child- modules as well. :param init: if True, the module repository will be cloned into place if necessary @@ -382,6 +401,7 @@ class Submodule(git.objects.base.IndexObject, Iterable, Traversable): """Move the submodule to a another module path. This involves physically moving the repository at our current path, changing the configuration, as well as adjusting our index entry accordingly. + :param module_path: the path to which to move our module, given as repository-relative path. Intermediate directories will be created accordingly. If the path already exists, it must be empty. @@ -484,6 +504,7 @@ class Submodule(git.objects.base.IndexObject, Iterable, Traversable): def remove(self, module=True, force=False, configuration=True, dry_run=False): """Remove this submodule from the repository. This will remove our entry from the .gitmodules file and the entry in the .git/config file. + :param module: If True, the module we point to will be deleted as well. If the module is currently on a commit which is not part of any branch in the remote, if the currently checked out branch @@ -588,6 +609,7 @@ class Submodule(git.objects.base.IndexObject, Iterable, Traversable): def set_parent_commit(self, commit, check=True): """Set this instance to use the given commit whose tree is supposed to contain the .gitmodules blob. + :param commit: Commit'ish reference pointing at the root_tree :param check: if True, relatively expensive checks will be performed to verify validity of the submodule. diff --git a/lib/git/objects/submodule/root.py b/lib/git/objects/submodule/root.py index 2e02e7de..82b8b271 100644 --- a/lib/git/objects/submodule/root.py +++ b/lib/git/objects/submodule/root.py @@ -1,4 +1,8 @@ from base import Submodule +from util import ( + mkhead, + find_first_remote_branch + ) from git.exc import InvalidGitRepositoryError import git diff --git a/lib/git/objects/tree.py b/lib/git/objects/tree.py index 68c1ef2d..67431686 100644 --- a/lib/git/objects/tree.py +++ b/lib/git/objects/tree.py @@ -7,7 +7,7 @@ import util from base import IndexObject from git.util import join_path from blob import Blob -from submodule import Submodule +from submodule.base import Submodule import git.diff as diff from fun import ( -- cgit v1.2.1