filter out URLs before finding hashtags

obsidiantools/_constants.py

@@ -1,6 +1,7 @@
 # WIKILINKS AND EMBEDDED FILES: regex that includes any aliases
 # group 0 captures embedded link; group 1 is everything inside [[]]
 WIKILINK_REGEX = r'(!)?\[{2}([^\]\]]+)\]{2}'
+URL_REGEX = r"(?i)\b((?:https?://|www\d{0,3}[.]|[a-z0-9.\-]+[.][a-z]{2,4}/)(?:[^\s()<>]+|\(([^\s()<>]+|(\([^\s()<>]+\)))*\))+(?:\(([^\s()<>]+|(\([^\s()<>]+\)))*\)|[^\s`!()\[\]{};:'\".,<>?«»“”‘’]))"
 
 # TAGS
 TAG_INCLUDE_NESTED_REGEX = r'(?<!\()(?<!\\)#{1}([A-z]+[0-9_\-]*[A-Z0-9]?[^\s]+(?![^\[\[]*\]\]))\/?'

obsidiantools/md_utils.py

@@ -5,6 +5,7 @@
 import markdown
 import frontmatter
 from ._constants import (WIKILINK_REGEX,
+                         URL_REGEX,
                          TAG_MAIN_ONLY_REGEX, TAG_INCLUDE_NESTED_REGEX,
                          WIKILINK_AS_STRING_REGEX,
                          EMBEDDED_FILE_LINK_AS_STRING_REGEX,
@@ -241,6 +242,8 @@ def get_tags(filepath: Path, *, show_nested: bool = False) -> list[str]:
         str_transform_func=_transform_md_file_string_for_tag_parsing)
     # remove wikilinks so that '#' headers are not caught:
     src_txt = _remove_wikilinks_from_source_text(src_txt)
+    # remove URLs so that '#' in URLs are not caught
+    src_txt = _remove_URLs_from_source_text(src_txt)
     tags = _get_tags_from_source_text(src_txt, show_nested=show_nested)
     return tags
 
@@ -444,6 +447,10 @@ def _remove_wikilinks_from_source_text(src_txt: str) -> str:
     return re.sub(WIKILINK_REGEX, '', src_txt)
 
 
+def _remove_URLs_from_source_text(src_txt: str) -> str:
+    return re.sub(URL_REGEX, '', src_txt)
+
+
 def _transform_md_file_string_for_tag_parsing(txt: str) -> str:
     return txt.replace('\\#', '')
 

tests/test_md_utils.py

@@ -15,6 +15,7 @@
                                     get_front_matter,
                                     get_tags,
                                     _remove_wikilinks_from_source_text,
+                                    _remove_URLs_from_source_text,
                                     _replace_wikilinks_with_their_text,
                                     _replace_md_links_with_their_text,
                                     get_readable_text_from_md_file)
@@ -288,8 +289,16 @@ def test_latex():
     assert actual_latex_list == expected_latex_list
 
 
-def test_remove_wikilinks(txt_wikilink_extraction_stub):
+def test_remove_wikilinks(txt_md_link_extraction_stub):
     out_str = _remove_wikilinks_from_source_text(
+        txt_md_link_extraction_stub)
+
+    expected_str = '[Obsidian.md homepage]()\n[Github homepage]()\n\n[Github homepage]()\n'
+    assert out_str == expected_str
+
+
+def test_remove_URLs(txt_wikilink_extraction_stub):
+    out_str = _remove_URLs_from_source_text(
         txt_wikilink_extraction_stub)
 
     expected_str = "\n" * 6

tests/vault-stub/Sussudio.md

@@ -32,6 +32,7 @@ However these shouldn't be recognised as tags:
 - #1985
 - [[American Psycho (film)#Patrick Bateman]]
 - \#hash_char_not_tag
+- docs.google.com/document/d/12345/edit#heading
 
 ```python
 # #code_comment_not_tag