Skip to content

prompt: record latex-raw-string-docstrings as shipped (#491) - #256

Merged
Jammy2211 merged 1 commit into
mainfrom
claude/latex-raw-string-docstrings-9h4ine
Aug 20, 2026
Merged

prompt: record latex-raw-string-docstrings as shipped (#491)#256
Jammy2211 merged 1 commit into
mainfrom
claude/latex-raw-string-docstrings-9h4ine

Conversation

@Jammy2211

Copy link
Copy Markdown
Collaborator

Final lifecycle step: active/complete/2026/08/, active.md section removed, complete/index.md and the dashboard regenerated. lifecycle.py check and orphans both clean.

All six PRs merged — HowToFit#47, HowToGalaxy#70, HowToLens#73, autofit_workspace#145, autogalaxy_workspace#218, autolens_workspace#492. 41 files, 180 literals, 131 silent corruptions repaired.

The record leads with the findings worth carrying forward rather than the tally:

  • The diff-empty gate cannot catch a wrong edit here. The generator reads source text, not runtime values, so a corrupted value regenerates byte-identically. The check that does catch it — comparing every changed literal's runtime value HEAD vs worktree, asserting the r prefix only ever removes corruption — caught two real errors mid-task: a print("\nInfo:") newline about to be destroyed, and 12 already-escaped $\\chi^2$ about to be doubled into LaTeX line breaks. Build that check first.
  • \\ is always deliberate, whatever the surrounding context — a math-context rule alone lets these through, since they sit inside $…$.
  • The warning sweep is interpreter-dependent — a SyntaxWarning-only sweep returns a vacuous zero on 3.11.
  • Run the generator before editing and confirm a no-op, or pre-existing drift is indistinguishable from damage you caused.
  • Preserve line endings — a write_text round-trip on a CRLF file turns a 1-character edit into a 549-line diff.

Also recorded: the one deliberate residue (autolens_workspace/scripts/group/likelihood_function.py, double-backslash convention, needs a convention decision), and the open question on HowToLens#73 — merged with smoke red, with a control experiment proving main is red too, and PyAutoArray#453 suspected but explicitly not proven.

Follow-up already shipped: PyAutoBrain#245 adds a hygiene escapes mode so this debt is caught continuously.

Note: the designated branch's previous PR (#255) was merged, so per the branch policy this restarts the branch from main and carries only the new commit.

🤖 Generated with Claude Code

https://claude.ai/code/session_01MTjtx5mdituitiyQYFGn2E


Generated by Claude Code

All six PRs merged: HowToFit#47, HowToGalaxy#70, HowToLens#73,
autofit_workspace#145, autogalaxy_workspace#218, autolens_workspace#492.
41 files, 180 literals, 131 silent corruptions repaired.

The record leads with the trap worth carrying forward: the prompt's diff-empty
gate CANNOT catch a wrong edit here, because the generator reads source text
rather than runtime values, so a corrupted value regenerates byte-identically.
The check that does catch it — comparing every changed literal's runtime value
HEAD vs worktree — caught two real errors mid-task.

Also recorded: the Python 3.11-vs-3.12 warning-class trap, that '\\' is always
deliberate whatever the surrounding math context, the CRLF-preservation trap,
the one deliberate residue, and the open question on HowToLens#73's red smoke
(proven to be red on main too; PyAutoArray#453 suspected but NOT proven).

active/ -> complete/2026/08/; active.md section removed; complete/index.md and
the dashboard regenerated.
@Jammy2211
Jammy2211 merged commit b99fdd6 into main Aug 20, 2026
2 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants