Regenerate a --no-cache directory when a source .jsonl grows (#254)

[cboos]

Fixes #254.

Problem

The single-file source-mtime freshness check added in #221 was scoped to
input_path.is_file(). A directory run without a cache (e.g.
--no-cache, where cache_manager is None) falls through to the same
no-cache fallback branch, but with no file mtime to compare and no DB tracking
per-source mtimes, should_regenerate rode is_outdated() (embedded tool
version) alone. So a session .jsonl that grew between runs with an unchanged
tool version was wrongly reported "current" → stale combined_transcripts.html.

The cached directory path is unaffected — it tracks source mtimes via
is_html_stale.

Fix

Extend the no-cache fallback's freshness check to directory sources: regenerate
when the newest source .jsonl in the directory is newer than the output —
the directory analogue of the single-file input.mtime > output.mtime,
persistence-free (the output's own mtime is the basis).

• Non-recursive glob("*.jsonl"), matching how directory mode discovers
  sessions.
• Shares the cache's filesystem-mtime granularity limit; also, a
  <stem>/subagents/ transcript growing without its top-level parent being
  touched isn't caught (rare — the parent session records the spawning
  tool_use and usually grows too). Noted in the code comment.
• Feeds only the combined regeneration decision, so the RegenerationReport
  signals stay accurate ("Successfully combined" prints because it truly did).

Test

test_grown_session_regenerates_without_cache — grow a session in a directory,
run --no-cache twice, assert the 2nd run regenerates (new content present),
coarse-FS-safe via os.utime. Verified RED before (served stale) and
GREEN after; also confirmed with a real CLI repro.

🤖 Generated with Claude Code

Summary by CodeRabbit

• Bug Fixes
  • Improved refresh detection when cache is disabled, so updated source files are more reliably reflected in generated output.
  • Directory-based inputs now correctly trigger regeneration when any newer source file is present.
• Tests
  • Added coverage for stale-output handling with directory inputs to verify refreshed content is regenerated as expected.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: 6b810f07-baa2-41a7-b61b-af6fe00f621d

📥 Commits

Reviewing files that changed from the base of the PR and between fc80008 and 03033fa.

📒 Files selected for processing (2)

• claude_code_log/converter.py
• test/test_output_explicit.py

---

📝 Walkthrough

Walkthrough

The fallback freshness check in convert_jsonl_to was updated to support directory inputs by comparing output mtime against the newest top-level *.jsonl file's mtime, in addition to the existing single-file comparison. A new test validates regeneration under --no-cache when a directory's session file grows.

Changes

Directory Staleness Fix

Layer / File(s)	Summary
Freshness check for directory inputs and regression test claude_code_log/converter.py, test/test_output_explicit.py	source_is_newer fallback logic now handles directory inputs by comparing output mtime to the newest top-level .jsonl mtime, feeding should_regenerate; a new test asserts --no-cache regenerates combined HTML when a directory's session grows.

Estimated code review effort: 2 (Simple) | ~10 minutes

Sequence Diagram(s)

sequenceDiagram
    participant User
    participant CLI
    participant Converter as convert_jsonl_to
    participant FS as Filesystem

    User->>CLI: run with --no-cache on directory
    CLI->>Converter: convert_jsonl_to(directory)
    Converter->>FS: get output mtime
    Converter->>FS: get max mtime of top-level *.jsonl files
    Converter->>Converter: compute source_is_newer
    Converter->>Converter: compute should_regenerate
    alt should_regenerate is True
        Converter->>FS: regenerate combined_transcripts.html
    else
        Converter-->>CLI: skip regeneration
    end

Loading

Related Issues: #254

Suggested labels: bug, converter

Suggested reviewers: daaain

🐰 A session grew, the cache stayed blind,
Now directories no longer leave old html behind,
mtimes compared, a freshness true,
Stale pages vanish, replaced anew,
Hop hop, the fix is neat and kind.

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly names the no-cache directory regeneration fix and the relevant issue, matching the main change.
Linked Issues check	✅ Passed	The code and test directly address #254 by regenerating when a directory's newest .jsonl is newer than the output.
Out of Scope Changes check	✅ Passed	The diff stays focused on directory no-cache freshness logic and its test, with no obvious unrelated changes.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

✨ Finishing Touches

📝 Generate docstrings

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch dev/nocache-dir-staleness

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}