Render Read/Write of any Markdown file as Markdown, not Pygments

claude_code_log/html/tool_formatters.py

@@ -23,6 +23,7 @@
 
 from .utils import (
     escape_html,
+    is_markdown_path,
     is_memory_path,
     render_collapsible_code,
     render_async_result_body,
@@ -391,6 +392,26 @@ def format_read_input(read_input: ReadInput) -> str:  # noqa: ARG001
 # Parsing (parse_read_output, parse_edit_output) is now in factories/tool_factory.py
 
 
+def _is_full_read(output: ReadOutput) -> bool:
+    """True when a Read result covers the file from line 1 with no truncation.
+
+    A full read is safe to render as Markdown; a partial slice (``start_line``
+    > 1 or truncated content) can split a code fence and is kept as Pygments
+    source (issue #232). The text-only parser fallback can't recover
+    ``total_lines`` and reports ``is_truncated=False`` with ``start_line=1``
+    for a whole-file read, so it correctly reads as full.
+
+    Residual edge: a *truncated* read that started at line 1, parsed via the
+    text-only fallback (no structured ``toolUseResult.file``), forces
+    ``is_truncated=False`` and so reads as full — rendering as Markdown even
+    though the tail was cut, which may garble a fence straddling the cut.
+    This is narrow (old transcripts only; modern ones carry the structured
+    metadata that classifies it correctly) and cosmetic (escaping still
+    applies, so no XSS/crash).
+    """
+    return output.start_line == 1 and not output.is_truncated
+
+
 def format_read_output(output: ReadOutput) -> str:
     """Format Read tool result as HTML with syntax highlighting.
 
@@ -411,12 +432,24 @@ def format_read_output(output: ReadOutput) -> str:
     # Auto-memory files are Markdown (MEMORY.md + topic .md), so render a
     # recalled-memory body as rendered Markdown rather than syntax-highlighted
     # source — using the project's usual collapsible-markdown helper (#192).
+    # Memory bodies render as Markdown unconditionally (even partial reads):
+    # memory files are small and read whole, and #192 pinned this behavior.
     if is_memory_path(output.file_path):
         # Escape HTML: memory files are untrusted content — raw <script>/HTML
         # must render as text, not live DOM when the transcript is opened.
         body = render_user_markdown_collapsible(output.content, "read-tool-result")
         return resolve_memory_body_links(body, output.file_path) + suffix_html
 
+    # Any other Markdown file: render the body as Markdown too (#232) — but
+    # only for a *full* read. A partial read (offset/limit) can begin or end
+    # mid-fence, which would render as garbled Markdown; a line-numbered
+    # source view is both safe and more useful for a slice, so partial reads
+    # keep Pygments. The escaping helper is used because file content is
+    # untrusted regardless of whether it's a memory file.
+    if is_markdown_path(output.file_path) and _is_full_read(output):
+        body = render_user_markdown_collapsible(output.content, "read-tool-result")
+        return body + suffix_html
+
     return render_file_content_collapsible(
         output.content,
         output.file_path,
@@ -651,13 +684,21 @@ def format_write_input(write_input: WriteInput) -> str:
     Note: File path is now shown in the header, so we skip it here.
     """
     # Memory files are Markdown — render a written memory body as rendered
-    # Markdown rather than highlighted source (#192).
+    # Markdown rather than highlighted source (#192), with memory link
+    # resolution applied.
     if is_memory_path(write_input.file_path):
         # Escape HTML (untrusted memory content) — see format_read_output.
         body = render_user_markdown_collapsible(
             write_input.content, "write-tool-content"
         )
         return resolve_memory_body_links(body, write_input.file_path)
+    # Any other Markdown file: render the written body as Markdown too (#232).
+    # A Write always carries the file's full content, so there's no partial
+    # slice to worry about — unlike Read. No memory link resolution applies.
+    if is_markdown_path(write_input.file_path):
+        return render_user_markdown_collapsible(
+            write_input.content, "write-tool-content"
+        )
     return render_file_content_collapsible(
         write_input.content, write_input.file_path, "write-tool-content"
     )

claude_code_log/html/utils.py

@@ -93,6 +93,23 @@ def is_memory_path(file_path: Optional[str]) -> bool:
     )
 
 
+# -- Markdown-file detection --------------------------------------------------
+# A file whose own format is Markdown. Read/Write of such a file renders the
+# body as Markdown rather than Pygments-highlighted source (issue #232).
+# Auto-memory files (always ``.md``) are a subset: ``is_memory_path`` ⊂
+# ``is_markdown_path``. The memory specialization that survives is the 🧠 title
+# (and memory's relative-link resolution + always-Markdown body); for any other
+# ``.md`` the generalization keys on this predicate instead.
+_MARKDOWN_EXTS = (".md", ".markdown")
+
+
+def is_markdown_path(file_path: Optional[str]) -> bool:
+    """True if ``file_path`` names a Markdown file (``.md`` / ``.markdown``)."""
+    return bool(file_path) and _normalize_sep(file_path).lower().endswith(
+        _MARKDOWN_EXTS
+    )
+
+
 def is_memory_tool(tool_name: Optional[str], file_path: Optional[str]) -> bool:
     """True if a tool call/result is an auto-memory interaction.
 

test/test_markdown_file_rendering.py

@@ -0,0 +1,142 @@
+"""Render Read/Write of any Markdown file as Markdown, not Pygments (issue #232).
+
+A follow-up to the auto-memory work (#192): a fully-contained Markdown file
+should render the usual way (rendered Markdown) instead of syntax-highlighted
+source. This applies to every ``.md`` file, not just memory files.
+
+- **Write** always carries the whole file → always rendered as Markdown.
+- **Read** renders as Markdown only for a *full* read; a partial slice
+  (offset/limit) can split a code fence, so partial reads keep Pygments.
+- Memory keeps its extra specialization (🧠 title, relative-link resolution,
+  and always-Markdown body even when partial) — see test_memory_rendering.py.
+"""
+
+import re
+from typing import Optional
+
+from claude_code_log.html.tool_formatters import (
+    format_read_output,
+    format_write_input,
+)
+from claude_code_log.html.utils import is_markdown_path
+from claude_code_log.models import ReadOutput, WriteInput
+
+
+MD = "/home/u/proj/docs/guide.md"
+MD_CAPS = "/home/u/proj/README.MARKDOWN"
+PY = "/home/u/proj/src/app.py"
+MEM = "/home/u/.claude/projects/-home-u-proj/memory/MEMORY.md"
+
+MD_BODY = "# Guide\n\nUse `just ci` before pushing.\n"
+
+
+def _read(
+    file_path: str, content: str, *, start_line: int = 1, total: Optional[int] = None
+):
+    n = len(content.splitlines())
+    total = n if total is None else total
+    return ReadOutput(
+        file_path=file_path,
+        content=content,
+        start_line=start_line,
+        num_lines=n,
+        total_lines=total,
+        is_truncated=n < total,
+    )
+
+
+# ----------------------------- is_markdown_path ------------------------------
+
+
+class TestIsMarkdownPath:
+    def test_md_and_markdown_extensions(self):
+        assert is_markdown_path(MD)
+        assert is_markdown_path("/x/y.markdown")
+
+    def test_case_insensitive(self):
+        assert is_markdown_path(MD_CAPS)
+        assert is_markdown_path("/x/Notes.Md")
+
+    def test_windows_separators(self):
+        assert is_markdown_path(r"C:\Users\u\docs\guide.md")
+
+    def test_memory_paths_are_markdown(self):
+        # Memory files are a subset: is_memory_path ⊂ is_markdown_path.
+        assert is_markdown_path(MEM)
+
+    def test_non_markdown_and_none(self):
+        assert not is_markdown_path(PY)
+        assert not is_markdown_path("/x/data.json")
+        assert not is_markdown_path("/x/MD")  # bare, no extension
+        assert not is_markdown_path(None)
+        assert not is_markdown_path("")
+
+
+# ----------------------------- Read rendering --------------------------------
+# The full/partial/truncated split (the ``_is_full_read`` predicate) is pinned
+# behaviorally through ``format_read_output`` rather than by importing the
+# private helper — keeps the test off the private symbol so it stays clean if
+# ``test/`` ever joins the pyright include scope (cf. #216's _PARAMS_TABLE_MAX_
+# DEPTH reportPrivateUsage).
+
+
+class TestReadMarkdownRendering:
+    def test_full_md_read_rendered_as_markdown(self):
+        # Whole file from line 1, not truncated → full read → Markdown.
+        html = format_read_output(_read(MD, MD_BODY))
+        assert re.search(r'class="read-tool-result markdown"', html)
+        assert re.search(r"<h1[^>]*>Guide", html)
+        assert "<code>just ci</code>" in html
+
+    def test_partial_md_read_keeps_pygments(self):
+        # A slice (start_line > 1) could land mid-fence → keep highlighted source.
+        html = format_read_output(_read(MD, MD_BODY, start_line=10, total=999))
+        assert 'class="read-tool-result markdown"' not in html
+        assert "<h1" not in html
+
+    def test_truncated_md_read_keeps_pygments(self):
+        # Starts at line 1 but truncated (num_lines < total_lines) → not full.
+        html = format_read_output(_read(MD, MD_BODY, total=999))
+        assert 'class="read-tool-result markdown"' not in html
+
+    def test_non_markdown_read_keeps_pygments(self):
+        html = format_read_output(_read(PY, "x = 1\n"))
+        assert 'class="read-tool-result markdown"' not in html
+
+    def test_markdown_body_escapes_raw_html(self):
+        # File content is untrusted regardless of being a memory file.
+        html = format_read_output(_read(MD, "# T\n\n<script>alert(1)</script>\n"))
+        assert "&lt;script&gt;" in html
+        assert "<script>alert(1)</script>" not in html
+
+    def test_general_md_read_has_no_memory_link_resolution(self):
+        # Relative links in a non-memory .md stay as-authored (no file:// rewrite).
+        html = format_read_output(_read(MD, "[peer](peer.md)\n"))
+        assert 'href="peer.md"' in html
+        assert "file://" not in html
+
+
+# ----------------------------- Write rendering -------------------------------
+
+
+class TestWriteMarkdownRendering:
+    def test_md_write_rendered_as_markdown(self):
+        html = format_write_input(WriteInput(file_path=MD, content=MD_BODY))
+        assert re.search(r'class="write-tool-content markdown"', html)
+        assert re.search(r"<h1[^>]*>Guide", html)
+
+    def test_non_markdown_write_keeps_pygments(self):
+        html = format_write_input(WriteInput(file_path=PY, content="x = 1\n"))
+        assert 'class="write-tool-content markdown"' not in html
+
+    def test_md_write_escapes_raw_html(self):
+        html = format_write_input(
+            WriteInput(file_path=MD, content="# T\n\n<script>alert(1)</script>")
+        )
+        assert "&lt;script&gt;" in html
+        assert "<script>alert(1)</script>" not in html
+
+    def test_general_md_write_has_no_memory_link_resolution(self):
+        html = format_write_input(WriteInput(file_path=MD, content="[peer](peer.md)\n"))
+        assert 'href="peer.md"' in html
+        assert "file://" not in html