comments: cut the development log from two tools and two more language layers - #686
Merged
Conversation
…e layers Comments only — all four verified code-identical to main by transpiling both revisions with removeComments and comparing the emit. xhosa 289 → 164 (2.63:1) vietnamese 239 → 144 (4.12:1) defects 412 → 314 (1.44:1) wiki-health 404 → 301 (1.41:1) The two language layers cut like the earlier ones: corpus tabulations, "WHAT WAS BROKEN" transcripts, decode evidence and policy-reversal narrative go; the ordering couplings, guards and invariants stay. The two tools/ files land at ~1.0:1 rather than the ~0.5:1 the language files reach, and that is the honest floor for them. Their bulk is threshold calibration — the de/ceb/ka numbers ARE the justification for every constant in wiki-health, and the regex-guard reasoning in defects is what stops the next person widening a lookbehind that is load-bearing. What did come out was the drift history, the issue references, and three doc blocks that had drifted away from the declaration they document. Also moved (comments only, no declarations reordered): defects.ts had the SIGN_CASES doc sitting above CITED_WORDS and the dropsIn doc above allOccurrencesInMarkup; wiki-health.ts carried two copies of its verdict header. Each now sits on what it describes. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_012Lc3WnUgogC7okV7n53vjr
xhosa/normalize.ts carried five "step 12" cross-references, and four of them were wrong on main: they name the step that spells a decimal out (`14.7` → `14 7`), which is step 15, not the degree rule at step 12. A reader following them lands on the wrong coupling. Corrected to 15; the fifth, which really does mean the degree rule, is unchanged. defects.ts: my paraphrase called hi's one true negative "the era-range shape's neighbour", which is not what it is — it is an astronomical magnitude that happens to share the shape. Stated plainly. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_012Lc3WnUgogC7okV7n53vjr
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Sixth batch. Comments only; all four files are code-identical to
main, verified by transpiling both revisions withremoveCommentsand comparing the emitted code.xhosa/normalize.tsvietnamese/normalize.tstools/normalization/defects.tstools/normalization/wiki-health.tsThe language layers
Same cut as the earlier batches: corpus tabulations, the "WHAT WAS BROKEN" IPA transcripts, decode evidence, and the policy-reversal narrative around Xhosa's plus sign all go. What stays, marked ⚠:
na-is a bound morpheme; ⟨c⟩ is a click, which is why no scale name is emitted; and the sign capture must be letter-guarded, because the concord hyphen looks exactly like a minus.. , : ; ? !are clause marks, so this file may only ever CONSUME them, never emit one;bằngis a homograph trap where frequency picks the instrumental preposition over the equality word.The tools files
These land at ~1.0:1, not the ~0.5:1 the language files reach, and I think that is the honest floor rather than timidity. Their bulk is threshold calibration: the de/ceb/ka numbers are the justification for every constant in
wiki-health, and the regex-guard reasoning indefectsis what stops the next person widening a lookbehind that is load-bearing. Deleting either would leave a magic number and an unexplained guard.What did come out: the drift history, the issue references, and the "first version got it wrong" narratives.
Three doc blocks had drifted away from what they document and are now moved (comments only — no declaration was reordered):
defects.tshad theSIGN_CASESdoc sitting aboveCITED_WORDS, and thedropsIndoc aboveallOccurrencesInMarkupwiki-health.tscarried two copies of its verdict header, one of them orphaned aboveHistogramChecks
npm run typecheckclean🤖 Generated with Claude Code
https://claude.ai/code/session_012Lc3WnUgogC7okV7n53vjr