manifests: the Adlam and Roman-Balochi script tables leave the .ts - #753
Merged
Conversation
They were named like IPA classes (VOWELS, R_VOWEL) but scan SPELLING, which is why pointing them at core/ipa.ts broke Quechua. #748 reverted them with a comment; this puts them where #746 says they belong, with the trap recorded beside the data: quechua spellingVowels — the -yuq/-niyuq suffix reads the last LETTER; ⟨y⟩ is absent on purpose, being a consonant letter in Quechua but the IPA vowel luo spellingVowels — the gi-/g- elision reads the first letter of the next number word balochi romanVowels — the ROMAN scan, distinct from the Arabic-script vowelLetters beside it fula latinVowels — the letters the Adlam lengthener doubles in adlamToLatin's output quechua gains a manifest.ts for the single parse (quechua.ts imports numbers.ts, so a back-import would cycle) — the abkhaz/georgian pattern. luo and fula already had one; balochi.ts already had DEF. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_012Lc3WnUgogC7okV7n53vjr
The last two whole transliteration tables still living in code — noted in #746, skipped in #748 because moving three sets out of a file whose surrounding table stayed would have been incoherent. balochi gains a "roman" block: Balochi is written in BOTH scripts and the Roman half is a second full g2p (phonemic, where the Arabic half is defective). Vowel letters, long/short, consonants, and the retroflex/postalveolar diacritic maps. The DIACRITIC LOGIC stays in balochi.ts, which is the point of the split. fula gains an "adlam" block: 34 letters plus the combining marks. Two things worth the reader's time are now recorded there — that Adlam marks length and gemination with MARKS rather than doubling (so the transliteration doubles what it just emitted rather than mapping them), and that the marks are written as SURROGATE PAIRS because they are astral and JSON has no \u{…} form, and invisible besides. And the move surfaced a triplicate: Adlam digits were folded in THREE places — a lookup table in fulaAdlam.ts, foldAdlamDigits in numbers.ts, and core/unicode.ts's foldNativeDigits, which has carried the Adlam base since #745. Both copies now defer to the shared one. Pure relocation: 998 generated Adlam/Roman words phonemized against main, 0 differ. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_012Lc3WnUgogC7okV7n53vjr
foldNativeDigits returns a NON-digit unchanged, so ADLAM[ch] ?? foldNativeDigits(ch) was never undefined — which killed the pass-through branch below it and, worse, set lastBase (the gemination target) from characters that are not Adlam letters. A stray ’ or an Arabic-Indic digit before U+1E946 doubled itself instead of the last real letter. It also folded every other script's digits inside an Adlam token. Guarded on the Adlam digit range, which is what the DIGITS lookup it replaced covered. Verified against main on 1292 generated words that now include stray and foreign characters precisely so this path is exercised — the earlier 998-word run used only well-formed Adlam and missed it. Also split balochi's stacked header: the first paragraph still described the flat romanVowels key from #751 rather than the six-table roman object that landed, so it now sits on roman.vowelLetters where it is true. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_012Lc3WnUgogC7okV7n53vjr
#751 squash-merged, so its commit replayed here as a duplicate carrying the ORPHANED header layout that the follow-up fix had already corrected on main. Taking main's version outright; this branch now contains only its own six files. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_012Lc3WnUgogC7okV7n53vjr
christopherthompson81
force-pushed
the
script-tables
branch
from
August 8, 2026 05:16
e5b2f5c to
a701cde
Compare
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.
The last two whole transliteration tables still living in code — noted in #746, and skipped in #748 because pulling three sets out of a file whose surrounding table stayed behind would have been incoherent.
Stacked on #751 (both languages have manifest changes there); merge that first.
balochi — a
romanblockBalochi is written in both scripts, and the Roman half is a second full g2p: phonemic, where the Arabic half is defective and leans on the lexicon. Its vowel letters, long/short maps, consonants, and the retroflex/postalveolar diacritic maps all move. The DIACRITIC LOGIC — which mark reaches for which table, and that a macron may arrive combining or precomposed — stays in
balochi.ts, which is the point of the split.One thing the move let the data state plainly: ⟨e o⟩ are long in both columns, because Balochi has no short /e o/. That was a trailing
// e,o have no short counterparton a const.fula — an
adlamblock34 letters plus the combining marks. Two things now recorded where a reader will find them:
\u{…}form — and they are invisible combining marks, so a literal would be a blank in the file. Verified they round-trip: all 34 letters and every mark come back at the right codepoint.A triplicate the move surfaced
Adlam digits were being folded in three places: a lookup table in
fulaAdlam.ts,foldAdlamDigitsinnumbers.ts, andcore/unicode.ts'sfoldNativeDigits— which has carried the Adlam base since #745. Both local copies now defer to the shared one.Verification
Pure relocation: 998 generated Adlam and Roman-Balochi words phonemized against
main, 0 differ. Spot-checked end to end —𞤀𞤣𞤤𞤢𞤥 → adlam → ˈadlam(uppercase folds),𞤢𞥅𞤤𞤢 → aala → ˈaːla(lengthener),𞥑𞥒𞥓 → 123 →number words.npm run cigreen: 228 files / 3136 tests, typecheck, fence.Also includes a fix to #751's quechua entry, which orphaned the digraphs header from its table — the same defect as #746 and #750, caught by the same script.
🤖 Generated with Claude Code
https://claude.ai/code/session_012Lc3WnUgogC7okV7n53vjr