Merge pull request #2181 from andrew-vant/recursive_epm

Allow directory entries in EXTRA_PATH_METADATA

docs/settings.rst

@@ -783,7 +783,8 @@ Metadata
 
    Extra metadata dictionaries keyed by relative path. Relative paths require
    correct OS-specific directory separators (i.e. / in UNIX and \\ in Windows)
-   unlike some other Pelican file settings.
+   unlike some other Pelican file settings. Paths to a directory apply to all
+   files under it. The most-specific path wins conflicts.
 
 Not all metadata needs to be :ref:`embedded in source file itself
 <internal_metadata>`. For example, blog posts are often named following a

pelican/readers.py

@@ -698,8 +698,20 @@ def path_metadata(full_path, source_path, settings=None):
         if settings.get('DEFAULT_DATE', None) == 'fs':
             metadata['date'] = SafeDatetime.fromtimestamp(
                 os.stat(full_path).st_mtime)
-        metadata.update(settings.get('EXTRA_PATH_METADATA', {}).get(
+
-            source_path, {}))
+        # Apply EXTRA_PATH_METADATA for the source path and the paths of any
+        # parent directories. Sorting EPM first ensures that the most specific
+        # path wins conflicts.
+
+        epm = settings.get('EXTRA_PATH_METADATA', {})
+        for path, meta in sorted(epm.items()):
+            # Enforce a trailing slash when checking for parent directories.
+            # This prevents false positives when one file or directory's name
+            # is a prefix of another's.
+            dirpath = os.path.join(path, '')
+            if source_path == path or source_path.startswith(dirpath):
+                metadata.update(meta)
+
     return metadata
 
 

pelican/tests/test_readers.py

@@ -273,6 +273,38 @@ class RstReaderTest(ReaderTest):
         }
         self.assertDictHasSubset(page.metadata, expected)
 
+    def test_article_extra_path_metadata_recurse(self):
+        parent = "TestCategory"
+        notparent = "TestCategory/article"
+        path = "TestCategory/article_without_category.rst"
+
+        epm = {
+            parent: {'epmr_inherit': parent,
+                     'epmr_override': parent, },
+            notparent: {'epmr_bogus': notparent},
+            path:   {'epmr_override': path, },
+            }
+        expected_metadata = {
+            'epmr_inherit': parent,
+            'epmr_override': path,
+            }
+
+        page = self.read_file(path=path, EXTRA_PATH_METADATA=epm)
+        self.assertDictHasSubset(page.metadata, expected_metadata)
+
+        # Make sure vars aren't getting "inherited" by mistake...
+        path = "article.rst"
+        page = self.read_file(path=path, EXTRA_PATH_METADATA=epm)
+        for k in expected_metadata.keys():
+            self.assertNotIn(k, page.metadata)
+
+        # Same, but for edge cases where one file's name is a prefix of
+        # another.
+        path = "TestCategory/article_without_category.rst"
+        page = self.read_file(path=path, EXTRA_PATH_METADATA=epm)
+        for k in epm[notparent].keys():
+            self.assertNotIn(k, page.metadata)
+
     def test_typogrify(self):
         # if nothing is specified in the settings, the content should be
         # unmodified