ensure correct casing when opening articles to check identifiers

tests/conftest.py

@@ -13,17 +13,17 @@
 import tests.visual
 
 from wikitools import console
-from wikitools.file_utils import exists_case_insensitive
+from wikitools.file_utils import file_tree
 
 
 sys.path.append(os.path.join(os.path.dirname(__file__), ".."))
 
 
 def clear_function_cache():
-    # exists_case_insensitive does some internal caching of the directory tree, making it stateful
+    # exists_case_insensitive and get_canonical_path_casing cache all directory paths
     # during normal execution, the current working directory never changes, but tests use a new temporary directory for each test case
-    if hasattr(exists_case_insensitive, 'all_article_paths_lowercased'):
-        delattr(exists_case_insensitive, 'all_article_paths_lowercased')
+    if hasattr(file_tree, 'cache'):
+        delattr(file_tree, 'cache')
 
 
 @pytest.fixture(scope='function')
@@ -92,6 +92,7 @@ def get_visual_tests():
 
 
 def run_visual_test(tests, test_index, case_index):
+    clear_function_cache()
     test = tests[test_index]
     print(f"({test_index + 1}/{len(tests)})", console.red(test.name), "-", test.description)
     test_case = test.cases[case_index]

wikitools/file_utils.py

@@ -23,6 +23,27 @@ def normalised(path: str) -> str:
     return normalised
 
 
+"""
+Returns a dictionary of file and directory paths, with a lowercased path for the key and the original casing for the value.
+Useful for looking up file paths case-insensitively on case-sensitive file systems.
+"""
+def file_tree():
+    # this cache would only become invalid when the current working directory changes, which only happens in tests and not during normal execution
+    if not hasattr(file_tree, "cache"):
+        tree = {normalised(article_path.lower()): normalised(article_path) for article_path in itertools.chain(list_all_dirs(["."]), list_all_files(["."]))}
+        setattr(file_tree, "cache", tree)
+    return getattr(file_tree, "cache")
+
+
+def get_canonical_path_casing(path: pathlib.Path) -> pathlib.Path:
+    """
+    Converts a file/directory path into a path with the correct casing.
+    The path must exist (throws KeyError otherwise)
+    """
+
+    return pathlib.Path(file_tree()[normalised(os.path.relpath(path.as_posix()).lower())])
+
+
 def exists_case_sensitive(path: pathlib.Path) -> bool:
     """
     Case-sensitive file/directory existence check
@@ -49,13 +70,8 @@ def exists_case_insensitive(path: pathlib.Path) -> bool:
         return path.exists()
     else:
         # case-insensitive directory/file existence checking isn't trivial in case-sensitive file systems because os-provided existence checks can't be relied upon
-        # this cache would only become invalid when the current working directory changes, which only happens in tests and not during normal execution
-        if not hasattr(exists_case_insensitive, 'all_article_paths_lowercased'):
-            article_set = set(normalised(article_path.lower()) for article_path in itertools.chain(list_all_dirs(["."]), list_all_files(["."])))
-            setattr(exists_case_insensitive, 'all_article_paths_lowercased', article_set)
-        all_article_paths_lowercased = getattr(exists_case_insensitive, 'all_article_paths_lowercased')
 
-        return normalised(os.path.relpath(path.as_posix()).lower()) in all_article_paths_lowercased
+        return normalised(os.path.relpath(path.as_posix()).lower()) in file_tree()
 
 
 def is_newspost(path: str) -> bool:

wikitools/link_checker.py

@@ -6,6 +6,7 @@
 from wikitools import redirect_parser, reference_parser, errors, link_parser, article_parser
 from wikitools import console
 from wikitools.file_utils import exists_case_sensitive, exists_case_insensitive
+from wikitools.file_utils import get_canonical_path_casing
 from wikitools.file_utils import is_article
 
 
@@ -186,17 +187,20 @@ def check_link(
     if not parsed_location.fragment:
         return None
 
+    target_path = repo_path.path
+    if os.name != 'nt' and not case_sensitive:
+        target_path = get_canonical_path_casing(target_path)
+
     # link to a section
     match repo_path.path_type:
         case PathType.GITHUB:
             # github links can either be directories or files
             # but section links are only relevant for markdown files
             if repo_path.path.suffix == ".md":
-                target_file = repo_path.path
-                raw_path = repo_path.path.as_posix()
+                raw_path = target_path.as_posix()
                 if raw_path not in all_articles:
                     # this is safe to do since the caller iterates over a copy of all_articles -> we can modify it as we wish
-                    all_articles[raw_path] = article_parser.parse(target_file)
+                    all_articles[raw_path] = article_parser.parse(target_path)
 
                 target_article = all_articles[raw_path]
 
@@ -209,22 +213,22 @@ def check_link(
                     return errors.MissingIdentifierError(link, raw_path, parsed_location.fragment, False, translation_outdated)
         case PathType.NEWS:
             # always a file path
-            raw_path = repo_path.path.as_posix()
+            raw_path = target_path.as_posix()
             if raw_path not in all_articles:
-                all_articles[raw_path] = article_parser.parse(repo_path.path)
+                all_articles[raw_path] = article_parser.parse(target_path)
             target_article = all_articles[raw_path]
 
             if parsed_location.fragment not in target_article.identifiers:
                 return errors.MissingIdentifierError(link, raw_path, parsed_location.fragment, False, False)
         case PathType.WIKI:
             # directory -> need to find the target article; it could be a translation
             # XXX(TicClick): this part assumes there is always an English version of the article in a folder
-            target_file = repo_path.path / article.filename
+            target_file = target_path / article.filename
             translation = target_file # verified to be the case later
             no_translation_available = article.filename != 'en.md' and not target_file.is_file()
 
             if no_translation_available:
-                target_file = repo_path.path / 'en.md'
+                target_file = target_path / 'en.md'
 
             raw_path = target_file.as_posix()
             if raw_path not in all_articles: