Fix link-patterns replacing matches inside inline HTML markup

lib/markdown2.py

@@ -3861,11 +3861,27 @@ class LinkPatterns(Extra):
     def run(self, text: str):
         link_from_hash = {}
         for regex, repl in self.options:
+            # Spans of inline HTML markup in the current text: tags,
+            # comments, processing instructions and `<...>` auto-link
+            # tokens (as recognized by the same tokenizer that
+            # `_escape_special_chars` uses). Matches inside them must
+            # not be linkified: that is markup, not prose, and replacing
+            # inside it corrupts attributes and destroys auto-links
+            # (e.g. `<word@example.com>`, `<img alt="word">`).
+            markup_spans = [
+                m.span(1)
+                for m in self.md._sorta_html_tokenize_re.finditer(text)
+                if self.md._is_unescaped_re.match(m.group(1))
+            ]
             replacements = []
             for match in regex.finditer(text):
                 if any(self.md._match_overlaps_substr(text, match, h) for h in link_from_hash):
                     continue
 
+                # Do not match inside inline HTML markup.
+                if any(s <= match.start() and match.end() <= e for s, e in markup_spans):
+                    continue
+
                 if callable(repl):
                     href = repl(match)
                 else:

test/tm-cases/link_patterns_email_autolink.html

@@ -0,0 +1 @@
+<p><a href="mailto:foo@example.com">foo@example.com</a> is an auto-linked email, but plain <a href="/w/foo">foo</a> is still linkified.</p>

test/tm-cases/link_patterns_email_autolink.opts

@@ -0,0 +1,4 @@
+{"extras": ["link-patterns"],
+ "link_patterns": [
+    (re.compile(r"\b(foo)\b"), r"/w/\1"),
+ ]}

test/tm-cases/link_patterns_email_autolink.tags

@@ -0,0 +1 @@
+link_patterns

test/tm-cases/link_patterns_email_autolink.text

@@ -0,0 +1 @@
+<foo@example.com> is an auto-linked email, but plain foo is still linkified.

test/tm-cases/link_patterns_html_tags.html

@@ -0,0 +1,7 @@
+<p>Inline HTML attributes must not be linkified:</p>
+
+<p><span title="foo bar">baz</span></p>
+
+<p><img src="/qux.png" alt="foo"></p>
+
+<p>But plain <a href="/w/foo">foo</a> in prose is linkified.</p>

test/tm-cases/link_patterns_html_tags.opts

@@ -0,0 +1,4 @@
+{"extras": ["link-patterns"],
+ "link_patterns": [
+    (re.compile(r"\b(foo)\b"), r"/w/\1"),
+ ]}

test/tm-cases/link_patterns_html_tags.tags

@@ -0,0 +1 @@
+link_patterns

test/tm-cases/link_patterns_html_tags.text

@@ -0,0 +1,7 @@
+Inline HTML attributes must not be linkified:
+
+<span title="foo bar">baz</span>
+
+<img src="/qux.png" alt="foo">
+
+But plain foo in prose is linkified.