# (c) 2005 Ian Bicking and contributors; written for Paste (http://pythonpaste.org) # Licensed under the MIT license: http://www.opensource.org/licenses/mit-license.php """ WSGI applications that parse the URL and dispatch to on-disk resources """ import os import six import sys import imp import mimetypes try: import pkg_resources except ImportError: pkg_resources = None from paste import request from paste import fileapp from paste.util import import_string from paste import httpexceptions from .httpheaders import ETAG from paste.util import converters class NoDefault(object): pass __all__ = ['URLParser', 'StaticURLParser', 'PkgResourcesParser'] class URLParser(object): """ WSGI middleware Application dispatching, based on URL. An instance of `URLParser` is an application that loads and delegates to other applications. It looks for files in its directory that match the first part of PATH_INFO; these may have an extension, but are not required to have one, in which case the available files are searched to find the appropriate file. If it is ambiguous, a 404 is returned and an error logged. By default there is a constructor for .py files that loads the module, and looks for an attribute ``application``, which is a ready application object, or an attribute that matches the module name, which is a factory for building applications, and is called with no arguments. URLParser will also look in __init__.py for special overrides. These overrides are: ``urlparser_hook(environ)`` This can modify the environment. Its return value is ignored, and it cannot be used to change the response in any way. You *can* use this, for example, to manipulate SCRIPT_NAME/PATH_INFO (try to keep them consistent with the original URL -- but consuming PATH_INFO and moving that to SCRIPT_NAME is ok). ``urlparser_wrap(environ, start_response, app)``: After URLParser finds the application, it calls this function (if present). If this function doesn't call ``app(environ, start_response)`` then the application won't be called at all! This can be used to allocate resources (with ``try:finally:``) or otherwise filter the output of the application. ``not_found_hook(environ, start_response)``: If no file can be found (*in this directory*) to match the request, then this WSGI application will be called. You can use this to change the URL and pass the request back to URLParser again, or on to some other application. This doesn't catch all ``404 Not Found`` responses, just missing files. ``application(environ, start_response)``: This basically overrides URLParser completely, and the given application is used for all requests. ``urlparser_wrap`` and ``urlparser_hook`` are still called, but the filesystem isn't searched in any way. """ parsers_by_directory = {} # This is lazily initialized init_module = NoDefault global_constructors = {} def __init__(self, global_conf, directory, base_python_name, index_names=NoDefault, hide_extensions=NoDefault, ignore_extensions=NoDefault, constructors=None, **constructor_conf): """ Create a URLParser object that looks at `directory`. `base_python_name` is the package that this directory represents, thus any Python modules in this directory will be given names under this package. """ if global_conf: import warnings warnings.warn( 'The global_conf argument to URLParser is deprecated; ' 'either pass in None or {}, or use make_url_parser', DeprecationWarning) else: global_conf = {} if os.path.sep != '/': directory = directory.replace(os.path.sep, '/') self.directory = directory self.base_python_name = base_python_name # This logic here should be deprecated since it is in # make_url_parser if index_names is NoDefault: index_names = global_conf.get( 'index_names', ('index', 'Index', 'main', 'Main')) self.index_names = converters.aslist(index_names) if hide_extensions is NoDefault: hide_extensions = global_conf.get( 'hide_extensions', ('.pyc', '.bak', '.py~', '.pyo')) self.hide_extensions = converters.aslist(hide_extensions) if ignore_extensions is NoDefault: ignore_extensions = global_conf.get( 'ignore_extensions', ()) self.ignore_extensions = converters.aslist(ignore_extensions) self.constructors = self.global_constructors.copy() if constructors: self.constructors.update(constructors) # @@: Should we also check the global options for constructors? for name, value in constructor_conf.items(): if not name.startswith('constructor '): raise ValueError( "Only extra configuration keys allowed are " "'constructor .ext = import_expr'; you gave %r " "(=%r)" % (name, value)) ext = name[len('constructor '):].strip() if isinstance(value, (str, unicode)): value = import_string.eval_import(value) self.constructors[ext] = value def __call__(self, environ, start_response): environ['paste.urlparser.base_python_name'] = self.base_python_name if self.init_module is NoDefault: self.init_module = self.find_init_module(environ) path_info = environ.get('PATH_INFO', '') if not path_info: return self.add_slash(environ, start_response) if (self.init_module and getattr(self.init_module, 'urlparser_hook', None)): self.init_module.urlparser_hook(environ) orig_path_info = environ['PATH_INFO'] orig_script_name = environ['SCRIPT_NAME'] application, filename = self.find_application(environ) if not application: if (self.init_module and getattr(self.init_module, 'not_found_hook', None) and environ.get('paste.urlparser.not_found_parser') is not self): not_found_hook = self.init_module.not_found_hook environ['paste.urlparser.not_found_parser'] = self environ['PATH_INFO'] = orig_path_info environ['SCRIPT_NAME'] = orig_script_name return not_found_hook(environ, start_response) if filename is None: name, rest_of_path = request.path_info_split(environ['PATH_INFO']) if not name: name = 'one of %s' % ', '.join( self.index_names or ['(no index_names defined)']) return self.not_found( environ, start_response, 'Tried to load %s from directory %s' % (name, self.directory)) else: environ['wsgi.errors'].write( 'Found resource %s, but could not construct application\n' % filename) return self.not_found( environ, start_response, 'Tried to load %s from directory %s' % (filename, self.directory)) if (self.init_module and getattr(self.init_module, 'urlparser_wrap', None)): return self.init_module.urlparser_wrap( environ, start_response, application) else: return application(environ, start_response) def find_application(self, environ): if (self.init_module and getattr(self.init_module, 'application', None) and not environ.get('paste.urlparser.init_application') == environ['SCRIPT_NAME']): environ['paste.urlparser.init_application'] = environ['SCRIPT_NAME'] return self.init_module.application, None name, rest_of_path = request.path_info_split(environ['PATH_INFO']) environ['PATH_INFO'] = rest_of_path if name is not None: environ['SCRIPT_NAME'] = environ.get('SCRIPT_NAME', '') + '/' + name if not name: names = self.index_names for index_name in names: filename = self.find_file(environ, index_name) if filename: break else: # None of the index files found filename = None else: filename = self.find_file(environ, name) if filename is None: return None, filename else: return self.get_application(environ, filename), filename def not_found(self, environ, start_response, debug_message=None): exc = httpexceptions.HTTPNotFound( 'The resource at %s could not be found' % request.construct_url(environ), comment=debug_message) return exc.wsgi_application(environ, start_response) def add_slash(self, environ, start_response): """ This happens when you try to get to a directory without a trailing / """ url = request.construct_url(environ, with_query_string=False) url += '/' if environ.get('QUERY_STRING'): url += '?' + environ['QUERY_STRING'] exc = httpexceptions.HTTPMovedPermanently( 'The resource has moved to %s - you should be redirected ' 'automatically.' % url, headers=[('location', url)]) return exc.wsgi_application(environ, start_response) def find_file(self, environ, base_filename): possible = [] """Cache a few values to reduce function call overhead""" for filename in os.listdir(self.directory): base, ext = os.path.splitext(filename) full_filename = os.path.join(self.directory, filename) if (ext in self.hide_extensions or not base): continue if filename == base_filename: possible.append(full_filename) continue if ext in self.ignore_extensions: continue if base == base_filename: possible.append(full_filename) if not possible: #environ['wsgi.errors'].write( # 'No file found matching %r in %s\n' # % (base_filename, self.directory)) return None if len(possible) > 1: # If there is an exact match, this isn't 'ambiguous' # per se; it might mean foo.gif and foo.gif.back for # instance if full_filename in possible: return full_filename else: environ['wsgi.errors'].write( 'Ambiguous URL: %s; matches files %s\n' % (request.construct_url(environ), ', '.join(possible))) return None return possible[0] def get_application(self, environ, filename): if os.path.isdir(filename): t = 'dir' else: t = os.path.splitext(filename)[1] constructor = self.constructors.get(t, self.constructors.get('*')) if constructor is None: #environ['wsgi.errors'].write( # 'No constructor found for %s\n' % t) return constructor app = constructor(self, environ, filename) if app is None: #environ['wsgi.errors'].write( # 'Constructor %s return None for %s\n' % # (constructor, filename)) pass return app def register_constructor(cls, extension, constructor): """ Register a function as a constructor. Registered constructors apply to all instances of `URLParser`. The extension should have a leading ``.``, or the special extensions ``dir`` (for directories) and ``*`` (a catch-all). `constructor` must be a callable that takes two arguments: ``environ`` and ``filename``, and returns a WSGI application. """ d = cls.global_constructors assert extension not in d, ( "A constructor already exists for the extension %r (%r) " "when attemption to register constructor %r" % (extension, d[extension], constructor)) d[extension] = constructor register_constructor = classmethod(register_constructor) def get_parser(self, directory, base_python_name): """ Get a parser for the given directory, or create one if necessary. This way parsers can be cached and reused. # @@: settings are inherited from the first caller """ try: return self.parsers_by_directory[(directory, base_python_name)] except KeyError: parser = self.__class__( {}, directory, base_python_name, index_names=self.index_names, hide_extensions=self.hide_extensions, ignore_extensions=self.ignore_extensions, constructors=self.constructors) self.parsers_by_directory[(directory, base_python_name)] = parser return parser def find_init_module(self, environ): filename = os.path.join(self.directory, '__init__.py') if not os.path.exists(filename): return None return load_module(environ, filename) def __repr__(self): return '<%s directory=%r; module=%s at %s>' % ( self.__class__.__name__, self.directory, self.base_python_name, hex(abs(id(self)))) def make_directory(parser, environ, filename): base_python_name = environ['paste.urlparser.base_python_name'] if base_python_name: base_python_name += "." + os.path.basename(filename) else: base_python_name = os.path.basename(filename) return parser.get_parser(filename, base_python_name) URLParser.register_constructor('dir', make_directory) def make_unknown(parser, environ, filename): return fileapp.FileApp(filename) URLParser.register_constructor('*', make_unknown) def load_module(environ, filename): base_python_name = environ['paste.urlparser.base_python_name'] module_name = os.path.splitext(os.path.basename(filename))[0] if base_python_name: module_name = base_python_name + '.' + module_name return load_module_from_name(environ, filename, module_name, environ['wsgi.errors']) def load_module_from_name(environ, filename, module_name, errors): if module_name in sys.modules: return sys.modules[module_name] init_filename = os.path.join(os.path.dirname(filename), '__init__.py') if not os.path.exists(init_filename): try: f = open(init_filename, 'w') except (OSError, IOError) as e: errors.write( 'Cannot write __init__.py file into directory %s (%s)\n' % (os.path.dirname(filename), e)) return None f.write('#\n') f.close() fp = None if module_name in sys.modules: return sys.modules[module_name] if '.' in module_name: parent_name = '.'.join(module_name.split('.')[:-1]) base_name = module_name.split('.')[-1] parent = load_module_from_name(environ, os.path.dirname(filename), parent_name, errors) else: base_name = module_name fp = None try: fp, pathname, stuff = imp.find_module( base_name, [os.path.dirname(filename)]) module = imp.load_module(module_name, fp, pathname, stuff) finally: if fp is not None: fp.close() return module def make_py(parser, environ, filename): module = load_module(environ, filename) if not module: return None if hasattr(module, 'application') and module.application: return getattr(module.application, 'wsgi_application', module.application) base_name = module.__name__.split('.')[-1] if hasattr(module, base_name): obj = getattr(module, base_name) if hasattr(obj, 'wsgi_application'): return obj.wsgi_application else: # @@: Old behavior; should probably be deprecated eventually: return getattr(module, base_name)() environ['wsgi.errors'].write( "Cound not find application or %s in %s\n" % (base_name, module)) return None URLParser.register_constructor('.py', make_py) class StaticURLParser(object): """ Like ``URLParser`` but only serves static files. ``cache_max_age``: integer specifies Cache-Control max_age in seconds """ # @@: Should URLParser subclass from this? def __init__(self, directory, root_directory=None, cache_max_age=None): self.directory = self.normpath(directory) self.root_directory = self.normpath(root_directory or directory) self.cache_max_age = cache_max_age def normpath(path): return os.path.normcase(os.path.abspath(path)) normpath = staticmethod(normpath) def __call__(self, environ, start_response): path_info = environ.get('PATH_INFO', '') if not path_info: return self.add_slash(environ, start_response) if path_info == '/': # @@: This should obviously be configurable filename = 'index.html' else: filename = request.path_info_pop(environ) full = self.normpath(os.path.join(self.directory, filename)) if not full.startswith(self.root_directory): # Out of bounds return self.not_found(environ, start_response) if not os.path.exists(full): return self.not_found(environ, start_response) if os.path.isdir(full): # @@: Cache? return self.__class__(full, root_directory=self.root_directory, cache_max_age=self.cache_max_age)(environ, start_response) if environ.get('PATH_INFO') and environ.get('PATH_INFO') != '/': return self.error_extra_path(environ, start_response) if_none_match = environ.get('HTTP_IF_NONE_MATCH') if if_none_match: mytime = os.stat(full).st_mtime if str(mytime) == if_none_match: headers = [] ## FIXME: probably should be ## ETAG.update(headers, '"%s"' % mytime) ETAG.update(headers, mytime) start_response('304 Not Modified', headers) return [''] # empty body fa = self.make_app(full) if self.cache_max_age: fa.cache_control(max_age=self.cache_max_age) return fa(environ, start_response) def make_app(self, filename): return fileapp.FileApp(filename) def add_slash(self, environ, start_response): """ This happens when you try to get to a directory without a trailing / """ url = request.construct_url(environ, with_query_string=False) url += '/' if environ.get('QUERY_STRING'): url += '?' + environ['QUERY_STRING'] exc = httpexceptions.HTTPMovedPermanently( 'The resource has moved to %s - you should be redirected ' 'automatically.' % url, headers=[('location', url)]) return exc.wsgi_application(environ, start_response) def not_found(self, environ, start_response, debug_message=None): exc = httpexceptions.HTTPNotFound( 'The resource at %s could not be found' % request.construct_url(environ), comment='SCRIPT_NAME=%r; PATH_INFO=%r; looking in %r; debug: %s' % (environ.get('SCRIPT_NAME'), environ.get('PATH_INFO'), self.directory, debug_message or '(none)')) return exc.wsgi_application(environ, start_response) def error_extra_path(self, environ, start_response): exc = httpexceptions.HTTPNotFound( 'The trailing path %r is not allowed' % environ['PATH_INFO']) return exc.wsgi_application(environ, start_response) def __repr__(self): return '<%s %r>' % (self.__class__.__name__, self.directory) def make_static(global_conf, document_root, cache_max_age=None): """ Return a WSGI application that serves a directory (configured with document_root) cache_max_age - integer specifies CACHE_CONTROL max_age in seconds """ if cache_max_age is not None: cache_max_age = int(cache_max_age) return StaticURLParser( document_root, cache_max_age=cache_max_age) class PkgResourcesParser(StaticURLParser): def __init__(self, egg_or_spec, resource_name, manager=None, root_resource=None): if pkg_resources is None: raise NotImplementedError("This class requires pkg_resources.") if isinstance(egg_or_spec, (six.binary_type, six.text_type)): self.egg = pkg_resources.get_distribution(egg_or_spec) else: self.egg = egg_or_spec self.resource_name = resource_name if manager is None: manager = pkg_resources.ResourceManager() self.manager = manager if root_resource is None: root_resource = resource_name self.root_resource = os.path.normpath(root_resource) def __repr__(self): return '<%s for %s:%r>' % ( self.__class__.__name__, self.egg.project_name, self.resource_name) def __call__(self, environ, start_response): path_info = environ.get('PATH_INFO', '') if not path_info: return self.add_slash(environ, start_response) if path_info == '/': # @@: This should obviously be configurable filename = 'index.html' else: filename = request.path_info_pop(environ) resource = os.path.normcase(os.path.normpath( self.resource_name + '/' + filename)) if self.root_resource is not None and not resource.startswith(self.root_resource): # Out of bounds return self.not_found(environ, start_response) if not self.egg.has_resource(resource): return self.not_found(environ, start_response) if self.egg.resource_isdir(resource): # @@: Cache? child_root = self.root_resource is not None and self.root_resource or \ self.resource_name return self.__class__(self.egg, resource, self.manager, root_resource=child_root)(environ, start_response) if environ.get('PATH_INFO') and environ.get('PATH_INFO') != '/': return self.error_extra_path(environ, start_response) type, encoding = mimetypes.guess_type(resource) if not type: type = 'application/octet-stream' # @@: I don't know what to do with the encoding. try: file = self.egg.get_resource_stream(self.manager, resource) except (IOError, OSError) as e: exc = httpexceptions.HTTPForbidden( 'You are not permitted to view this file (%s)' % e) return exc.wsgi_application(environ, start_response) start_response('200 OK', [('content-type', type)]) return fileapp._FileIter(file) def not_found(self, environ, start_response, debug_message=None): exc = httpexceptions.HTTPNotFound( 'The resource at %s could not be found' % request.construct_url(environ), comment='SCRIPT_NAME=%r; PATH_INFO=%r; looking in egg:%s#%r; debug: %s' % (environ.get('SCRIPT_NAME'), environ.get('PATH_INFO'), self.egg, self.resource_name, debug_message or '(none)')) return exc.wsgi_application(environ, start_response) def make_pkg_resources(global_conf, egg, resource_name=''): """ A static file parser that loads data from an egg using ``pkg_resources``. Takes a configuration value ``egg``, which is an egg spec, and a base ``resource_name`` (default empty string) which is the path in the egg that this starts at. """ if pkg_resources is None: raise NotImplementedError("This function requires pkg_resources.") return PkgResourcesParser(egg, resource_name) def make_url_parser(global_conf, directory, base_python_name, index_names=None, hide_extensions=None, ignore_extensions=None, **constructor_conf): """ Create a URLParser application that looks in ``directory``, which should be the directory for the Python package named in ``base_python_name``. ``index_names`` are used when viewing the directory (like ``'index'`` for ``'index.html'``). ``hide_extensions`` are extensions that are not viewable (like ``'.pyc'``) and ``ignore_extensions`` are viewable but only if an explicit extension is given. """ if index_names is None: index_names = global_conf.get( 'index_names', ('index', 'Index', 'main', 'Main')) index_names = converters.aslist(index_names) if hide_extensions is None: hide_extensions = global_conf.get( 'hide_extensions', ('.pyc', 'bak', 'py~')) hide_extensions = converters.aslist(hide_extensions) if ignore_extensions is None: ignore_extensions = global_conf.get( 'ignore_extensions', ()) ignore_extensions = converters.aslist(ignore_extensions) # There's no real way to set constructors currently... return URLParser({}, directory, base_python_name, index_names=index_names, hide_extensions=hide_extensions, ignore_extensions=ignore_extensions, **constructor_conf)