🐛 Keep backslash escapes and entities in image alt text

CHANGELOG.md

@@ -5,6 +5,7 @@
 * ✨ Add `--enable-tables` to the CLI for file, standard input and interactive parsing in [#422](https://github.com/executablebooks/markdown-it-py/pull/422)
 * 🐛 Fix CLI interactive mode joining input lines with an extra newline, which split every line into its own paragraph and broke hard line breaks, in [#172](https://github.com/executablebooks/markdown-it-py/issues/172)
 * 🐛 Fix trimming and splitting with the Python whitespace set instead of the CommonMark one, which dropped U+001C–U+001F and U+0085 from paragraphs, headings, table cells and fence info strings and let distinct reference labels resolve each other, in [#418](https://github.com/executablebooks/markdown-it-py/pull/418), thanks to [@Nexory](https://github.com/Nexory)
+* 🐛 Fix image `alt` text dropping backslash escapes and entities, by also joining `text_special` tokens inside an image (ported from markdown-it 15.0.0), in [#445](https://github.com/executablebooks/markdown-it-py/issues/445)
 * 📚 Document the Python renderer constructor contract.
 * 🔧 Drop the optional dependency on the `commonmark` package and its benchmark, since the upstream package is unmaintained, in [#401](https://github.com/executablebooks/markdown-it-py/pull/401)
 

markdown_it/rules_core/text_join.py

@@ -19,35 +19,39 @@ def text_join(state: StateCore) -> None:
         if inline_token.type != "inline":
             continue
 
-        # convert text_special to text and join all adjacent text nodes
-        new_tokens: list[Token] = []
         children = inline_token.children or []
-        i = 0
-        while i < len(children):
-            child_token = children[i]
-            if child_token.type == "text_special":
-                child_token.type = "text"
-            if (
-                child_token.type == "text"
-                and new_tokens
-                and new_tokens[-1].type == "text"
-            ):
-                # Collapse a run of adjacent text nodes in a single join, instead
-                # of pairwise `a + b` concatenation. The pairwise form is O(L*k)
-                # in the size of the run because each step rebuilds the growing
-                # prefix; "".join is O(L).
-                parts = [new_tokens[-1].content, child_token.content]
+        for child_token in children:
+            # image `alt` is parsed into its own token tree
+            if child_token.children:
+                child_token.children = _join_text_tokens(child_token.children)
+        inline_token.children = _join_text_tokens(children)
+
+
+def _join_text_tokens(tokens: list[Token]) -> list[Token]:
+    """Convert `text_special` to `text` and join all adjacent text tokens"""
+    new_tokens: list[Token] = []
+    i = 0
+    while i < len(tokens):
+        token = tokens[i]
+        if token.type == "text_special":
+            token.type = "text"
+        if token.type == "text" and new_tokens and new_tokens[-1].type == "text":
+            # Collapse a run of adjacent text nodes in a single join, instead
+            # of pairwise `a + b` concatenation. The pairwise form is O(L*k)
+            # in the size of the run because each step rebuilds the growing
+            # prefix; "".join is O(L).
+            parts = [new_tokens[-1].content, token.content]
+            i += 1
+            while i < len(tokens):
+                next_token = tokens[i]
+                if next_token.type == "text_special":
+                    next_token.type = "text"
+                if next_token.type != "text":
+                    break
+                parts.append(next_token.content)
                 i += 1
-                while i < len(children):
-                    next_token = children[i]
-                    if next_token.type == "text_special":
-                        next_token.type = "text"
-                    if next_token.type != "text":
-                        break
-                    parts.append(next_token.content)
-                    i += 1
-                new_tokens[-1].content = "".join(parts)
-            else:
-                new_tokens.append(child_token)
-                i += 1
-        inline_token.children = new_tokens
+            new_tokens[-1].content = "".join(parts)
+        else:
+            new_tokens.append(token)
+            i += 1
+    return new_tokens

tests/test_port/fixtures/issue-fixes.md

@@ -54,3 +54,22 @@ Fix parsing of incorrect numeric character references
 <p><a href="&amp;#X22y;"></a> &amp;#X22y;
 <a href="&amp;#35y;"></a> &amp;#35y;</p>
 .
+
+#445 Image alt keeps backslash escapes and entities
+.
+![a \* b](/u)
+
+![a &amp; b](/u)
+
+![a &lt; b](/u)
+
+![a \* *b* &amp; c](/u)
+
+![C:\Python26](/u)
+.
+<p><img src="/u" alt="a * b" /></p>
+<p><img src="/u" alt="a &amp; b" /></p>
+<p><img src="/u" alt="a &lt; b" /></p>
+<p><img src="/u" alt="a * b &amp; c" /></p>
+<p><img src="/u" alt="C:\Python26" /></p>
+.