Fixes #422 - Allow ability to lazy-load external modules (#430)

3 files changed, 72 insertions, 43 deletions

diff --git a/src/tablib/core.py b/src/tablib/core.py
index 24a3471..9eaeb61 100644
--- a/src/tablib/core.py
+++ b/src/tablib/core.py
@@ -236,11 +236,11 @@ class Dataset:
     # Internals
     # ---------
 
-    def _get_in_format(self, fmt, **kwargs):
-        return fmt.export_set(self, **kwargs)
+    def _get_in_format(self, fmt_key, **kwargs):
+        return registry.get_format(fmt_key).export_set(self, **kwargs)
 
-    def _set_in_format(self, fmt, *args, **kwargs):
-        return fmt.import_set(self, *args, **kwargs)
+    def _set_in_format(self, fmt_key, *args, **kwargs):
+        return registry.get_format(fmt_key).import_set(self, *args, **kwargs)
 
     def _validate(self, row=None, col=None, safety=False):
         """Assures size of every row in dataset is of proper proportions."""
diff --git a/src/tablib/formats/__init__.py b/src/tablib/formats/__init__.py
index 9398607..5c46008 100644
--- a/src/tablib/formats/__init__.py
+++ b/src/tablib/formats/__init__.py
@@ -2,23 +2,14 @@
 """
 from collections import OrderedDict
 from functools import partialmethod
+from importlib import import_module
 from importlib.util import find_spec
 
 from tablib.exceptions import UnsupportedFormat
 
 from ._csv import CSVFormat
-from ._dbf import DBFFormat
-from ._df import DataFrameFormat
-from ._html import HTMLFormat
-from ._jira import JIRAFormat
 from ._json import JSONFormat
-from ._latex import LATEXFormat
-from ._ods import ODSFormat
-from ._rst import ReSTFormat
 from ._tsv import TSVFormat
-from ._xls import XLSFormat
-from ._xlsx import XLSXFormat
-from ._yaml import YAMLFormat
 
 uninstalled_format_messages = {
     'df': (
@@ -48,68 +39,105 @@ uninstalled_format_messages = {
 }
 
 
+def load_format_class(dotted_path):
+    try:
+        module_path, class_name = dotted_path.rsplit('.', 1)
+        return getattr(import_module(module_path), class_name)
+    except (ValueError, AttributeError) as err:
+        raise ImportError("Unable to load format class '{}' ({})".format(dotted_path, err))
+
+
+class FormatDescriptorBase:
+    def __init__(self, key, format_or_path):
+        self.key = key
+        self._format_path = None
+        if isinstance(format_or_path, str):
+            self._format = None
+            self._format_path = format_or_path
+        else:
+            self._format = format_or_path
+
+    def ensure_format_loaded(self):
+        if self._format is None:
+            self._format = load_format_class(self._format_path)
+
+
+class ImportExportBookDescriptor(FormatDescriptorBase):
+    def __get__(self, obj, cls, **kwargs):
+        self.ensure_format_loaded()
+        return self._format.export_book(obj, **kwargs)
+
+    def __set__(self, obj, val):
+        self.ensure_format_loaded()
+        return self._format.import_book(obj, val)
+
+
+class ImportExportSetDescriptor(FormatDescriptorBase):
+    def __get__(self, obj, cls, **kwargs):
+        self.ensure_format_loaded()
+        return self._format.export_set(obj, **kwargs)
+
+    def __set__(self, obj, val):
+        self.ensure_format_loaded()
+        return self._format.import_set(obj, val)
+
+
 class Registry:
     _formats = OrderedDict()
 
-    def register(self, key, format_):
+    def register(self, key, format_or_path):
         from tablib.core import Databook, Dataset
 
         # Create Databook.<format> read or read/write properties 
-        try:
-            setattr(Databook, format_.title, property(format_.export_book, format_.import_book))
-        except AttributeError:
-            try:
-                setattr(Databook, format_.title, property(format_.export_book))
-            except AttributeError:
-                pass
+        setattr(Databook, key, ImportExportBookDescriptor(key, format_or_path))
 
         # Create Dataset.<format> read or read/write properties,
         # and Dataset.get_<format>/set_<format> methods.
+        setattr(Dataset, key, ImportExportSetDescriptor(key, format_or_path))
         try:
-            try:
-                setattr(Dataset, format_.title, property(format_.export_set, format_.import_set))
-                setattr(Dataset, 'get_%s' % format_.title, partialmethod(Dataset._get_in_format, format_))
-                setattr(Dataset, 'set_%s' % format_.title, partialmethod(Dataset._set_in_format, format_))
-            except AttributeError:
-                setattr(Dataset, format_.title, property(format_.export_set))
-                setattr(Dataset, 'get_%s' % format_.title, partialmethod(Dataset._get_in_format, format_))
-
+            setattr(Dataset, 'get_%s' % key, partialmethod(Dataset._get_in_format, key))
+            setattr(Dataset, 'set_%s' % key, partialmethod(Dataset._set_in_format, key))
         except AttributeError:
-            raise Exception("Your format class should minimally implement the export_set interface.")
+            setattr(Dataset, 'get_%s' % key, partialmethod(Dataset._get_in_format, key))
 
-        self._formats[key] = format_
+        self._formats[key] = format_or_path
 
     def register_builtins(self):
         # Registration ordering matters for autodetection.
         self.register('json', JSONFormat())
         # xlsx before as xls (xlrd) can also read xlsx
         if find_spec('openpyxl'):
-            self.register('xlsx', XLSXFormat())
+            self.register('xlsx', 'tablib.formats._xlsx.XLSXFormat')
         if find_spec('xlrd') and find_spec('xlwt'):
-            self.register('xls', XLSFormat())
+            self.register('xls', 'tablib.formats._xls.XLSFormat')
         if find_spec('yaml'):
-            self.register('yaml', YAMLFormat())
+            self.register('yaml', 'tablib.formats._yaml.YAMLFormat')
         self.register('csv', CSVFormat())
         self.register('tsv', TSVFormat())
         if find_spec('odf'):
-            self.register('ods', ODSFormat())
-        self.register('dbf', DBFFormat())
+            self.register('ods', 'tablib.formats._ods.ODSFormat')
+        self.register('dbf', 'tablib.formats._dbf.DBFFormat')
         if find_spec('MarkupPy'):
-            self.register('html', HTMLFormat())
-        self.register('jira', JIRAFormat())
-        self.register('latex', LATEXFormat())
+            self.register('html', 'tablib.formats._html.HTMLFormat')
+        self.register('jira', 'tablib.formats._jira.JIRAFormat')
+        self.register('latex', 'tablib.formats._latex.LATEXFormat')
         if find_spec('pandas'):
-            self.register('df', DataFrameFormat())
-        self.register('rst', ReSTFormat())
+            self.register('df', 'tablib.formats._df.DataFrameFormat')
+        self.register('rst', 'tablib.formats._rst.ReSTFormat')
 
     def formats(self):
-        yield from self._formats.values()
+        for key, frm in self._formats.items():
+            if isinstance(frm, str):
+                self._formats[key] = load_format_class(frm)
+            yield self._formats[key]
 
     def get_format(self, key):
         if key not in self._formats:
             if key in uninstalled_format_messages:
                 raise UnsupportedFormat(uninstalled_format_messages[key])
             raise UnsupportedFormat("Tablib has no format '%s' or it is not registered." % key)
+        if isinstance(self._formats[key], str):
+            self._formats[key] = load_format_class(self._formats[key])
         return self._formats[key]
 
 
diff --git a/src/tablib/formats/_csv.py b/src/tablib/formats/_csv.py
index 5454bd9..cb209fc 100644
--- a/src/tablib/formats/_csv.py
+++ b/src/tablib/formats/_csv.py
@@ -48,6 +48,7 @@ class CSVFormat:
             elif row:
                 dset.append(row)
 
+    @classmethod
     def detect(cls, stream, delimiter=None):
         """Returns True if given stream is valid CSV."""
         try: