Repository navigation
ja: rebuild unicode-full.yaml readings and take the good terms from MathPlayer - #929
Merged
Merged
Conversation
Adding the preference is a stall tactic while I wait for feedback from ICEB about whether BANA rules are obsolete. Introduce UEB_G1_BANA_Style (default true) to choose BANA 2019 vs ICEB GTM 1.7.3/1.7.5 grade-1 indicator placement in ueb_cleanup. Implement GTM 1.7 per-symbols-sequence counting with gtm_1_7_mode, and add matching ICEB tests. Extend handle_contractions for technical vocabulary: UEB 10.8 final-letter groupsigns, dot-5 through/part, dis/con prefixes, cc/ea exception words, and a ful-pattern fix. Add ueb_technical_word_contractions (311 words, goldens cross-checked with brailletranslators.com). Wire the preference through prefs, fuzz, and docs.
…art of a 'word' (where 'word' has a specific UEB meaning)
…ds new tests and changes a few tests. Also added more contractions.
Fix intervals german
Add German ClearSpeak rule and tests for pause intent
…German Add German localization for currency symbol placement (with test)
…Log) and corresponding tests
…g-for-German Add support for principal logarithm in German
…ction definitions
…bersOrdinalFractionalPluralOnes improve German fractions
…-subset German ClearSpeak: subset & superset
…ical-states de: fix chemical state names
Bumps [libfuzzer-sys](https://github.com/rust-fuzz/libfuzzer) from 0.4.12 to 0.4.13. - [Changelog](https://github.com/rust-fuzz/libfuzzer/blob/main/CHANGELOG.md) - [Commits](rust-fuzz/libfuzzer@0.4.12...0.4.13) --- updated-dependencies: - dependency-name: libfuzzer-sys dependency-version: 0.4.13 dependency-type: direct:production update-type: version-update:semver-patch ... Signed-off-by: dependabot[bot] <support@github.com> Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
The entity regex in set_mathml() bounded the entity name at 10 letters:
&([a-zA-Z]{2,10});
but 2127 of the names in src/entities.in are longer, up to
"CounterClockwiseContourIntegral" (31 letters). Those names were never
substituted, so they reached the XML parser unresolved and set_mathml()
rejected the whole expression with "Tag/attribute not found" instead of
translating it.
MathML produced by MathJax uses ⁢ (14 letters) for
invisible multiplication, which is why the regression was reported by
JAWS users as soon as 0.7.7 shipped (issue daisy#896).
Widen the bound to 32 letters, matching the {1,32} bound already used
for namespace declarations and element prefixes. Unknown names are still
rejected by the parser, so the DoS guard is unaffected.
Add tests/entities.rs covering names at the old limit, one letter over
it, the longest name in the table, numeric character references, and
that unknown names are still rejected.
…a-stripe row alternation" (daisy#927)
* "Enhance rule coverage UI with hover effects, focus styling, and zebra-stripe row alternation" * Update `phf` dependencies to version 0.14.0 in `fuzz/Cargo.lock`.
unicode-full.yaml: every reading was compared against the English text and the MathPlayer/DeepL candidates from daisy#715, and rewritten with one glossary (Yamaguchi et al. 1996 for relations, brackets and arrows; JIS-style names for shapes). Structure and comments are unchanged; values stay t: since they were not checked by a human translator one by one. unicode.yaml: keep the current file, take カッコ/カッコ閉じ, 向き arrows and プラス オア マイナス; × "By" and ∦ follow the same readings. definitions.yaml and the menclose arrows are kept in step. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01RH9oPBRrM4zTh2hHPEQiwi
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01RH9oPBRrM4zTh2hHPEQiwi
… words The bracket words are 3 bytes longer, which is enough for the length-based auto pause to insert a short pause after the opening bracket. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01RH9oPBRrM4zTh2hHPEQiwi
Contributor
Author
|
The two Clippy failures are in src/canonicalize.rs ( |
Collaborator
|
yes I think that makes sense. after that, this PR is ready to go.
|
Brings in the Clippy fix in src/canonicalize.rs and the shorter fuzz runs from main.
Linux library size: 0.59 MiB (+2.34%)
Built with default features, Rust 1.96.0, and Ubuntu 24.04. Workflow run. |
… verbosity) main swapped the two delimiters and updated every other language. ja still had the old form, so "function=デル:勾配" was read as the name デル with glue 勾配. Also add FunctionApplicationWord (オブ, the same word unicode.yaml uses for U+2061).
moritz-gross
approved these changes
Oct 27, 2026
main now turns a∶b into the function intent ratio(a, b), but ja only had an infix form. The header comment still described the old IntentMappings syntax; it now matches en.
Contributor
Author
|
Merged main into this branch. ja needed a few follow-ups: the |
Collaborator
|
Great that this is done. I want to make sure I understand the status of the Japanese translation: is it ready to be merged into the the main branch? |
Contributor
Author
|
Yes, I think it's ready for main. Navigation is fully reviewed (every string but one is |
Collaborator
|
Great. I'll merge it in. |
NSoiffer
added a commit
that referenced
this pull request
Oct 28, 2026
* add AI seeding for Japanese, created by gpt-5.6 sol
* ja: read fractions in Japanese word order (#720)
* ja: read fractions in Japanese word order
A fraction of two numbers is read denominator-first in Japanese: 21/22 is
"22 分の 21". The seeded rules emitted numerator, 分の, denominator, so the
tests were asserting that 21/22 is spoken as 22/21. That is a meaning
change, not a wording preference, so the rule and its test move together.
Reference: 山口雄仁・川根深・澤崎陽彦, 日本語による数式読み上げ法の基本構成に
ついて, 日本数学教育学会誌 78(9), 239-247 (1996), item (4): when numerator and
denominator are both plain numbers, use ordinary Japanese word order
(denominator 分の numerator); otherwise keep the written order and say
分数 A オーバー B 分数終了.
Following that split:
- common-fraction (both mn, no decimal separator) now says 分母 分の 分子.
The English range limits (numerator < 20, denominator 2..10) existed
because English only has ordinals for small numbers; Japanese has no
separate ordinal form to build, so the pattern is uniform and ToOrdinal
is no longer used here.
- the remaining fraction rules keep the written order and now say オーバー
rather than 分の, which would otherwise claim the opposite order.
- ClearSpeak's general fraction said 分子 との 分数 ... デノミネーター; it now
uses the same 分数 ... オーバー ... 分数終了 frame.
- 分数終わり -> 分数終了, matching the terminology of the reference (and the
根号終了 / 上付き終了 markers that will follow).
Also in this file: per-fraction said パーカー, which is a hooded sweatshirt.
"5 meters per second" is 5 メートル 毎 秒.
* ja: keep one definition of "a fraction of two numbers"
The previous commit widened common-fraction, but ClearSpeak copies that
predicate in five other places: the EndFrac test inside fraction-over-simple
(twice, for numerator and denominator) and the four nested-* exponent rules
that ask whether the base is a common fraction. Left alone, 3/128 would count
as a common fraction where it is spoken and not count where those rules look,
so the two would disagree.
The copies now use the same test: both children are mn without a decimal
separator. The English range limits (numerator < 20, denominator 2..10) are
gone for the same reason as before -- they bound what English has ordinals
for, and Japanese builds no ordinal form.
No wording changes here; the phrasing of the exponent rules themselves is a
separate change.
* ja: read exponents in Japanese (#721)
* ja: read exponents as (base) no (exponent) jou
* ja: drop the negative-exponent test from this change
* ja: cover the superscript frame for a complex exponent
* ja: read a leading minus as マイナス (#722)
A unary + or - was spoken 正の / 負の. Those name a kind of number -- 負の数
is "the negative numbers" as a category -- and are not how -5 is read aloud.
Japanese says マイナス 5, with the same word used for the binary operator;
the reference's worked example reads -b as マイナス b, and its symbol table
gives ± as プラス・オア・マイナス.
Three pairs in ClearSpeak_Rules.yaml (the two PosNegSqRoot rules and
negative_and_positive) and the prefix forms of "plus" and "minus" in
definitions.yaml.
Not touched: SharedRules/general.yaml, where 正の / 負の are correct -- there
they modify 整数 in "the set of all positive integers".
* ja: read brackets and absolute value in Japanese
The seeded wording transliterated the English cues and kept the English word
order, which puts the closing cue before the group it closes.
Japanese names the shape when the group opens and postposes the closing cue
(Yamaguchi, Kawane and Sawazaki 1996, item 8):
( 括弧 / 丸括弧 ) 括弧閉じ / 丸括弧閉じ
[ 角括弧 ] 角括弧閉じ
{ 中括弧 } 中括弧閉じ
The Terse/Verbose split is kept where en has one: 丸 (the shape) drops out in
Terse the same way "paren" does in English.
Absolute value: 「絶対値 の x」 is ungrammatical because の attaches the other
way round in Japanese, so "the absolute value of x" would have to be 「x の絶対値」.
The cited reading keeps the written order and simply drops the preposition, so
|x| is 絶対値 x. With AbsEnd the closing cue moves after the contents and the
transliteration リリース ("release", from "end") becomes 閉じ.
audit-translations ja: Rule differences unchanged at 22, untranslated 3681 to 3680.
* ja: match the test expectations to the actual output
* ja: read an n-th root as n 乗根 (#723)
* ja: use the Japanese geometry terms (#724)
* ja: use the Japanese geometry terms
* ja: a ray passes through its second point, it does not end there
Verbose said 半直線 A から B まで, which claims the ray stops at B. A ray has A as its endpoint and continues through B, so read it as A を始点として B を通る. A segment does end at its second point, so から...まで stays there; a test now pins the two apart.
* ja: read large operators in Japanese word order
* ja: match the product test to the actual output
* ja: use the Japanese mathematical terms in definitions.yaml
* ja: use 行列 and かける for matrices everywhere
* ja: match the matrix test to the actual output
* ja: use 集合 for sets and the standard names for the number sets
* ja: finish the number-set sweep (rational, integers) and test all five
* ja: use the Japanese names for the elementary functions and the parts of a complex number
* ja: read intervals with a verb on each endpoint
* ja: mark the interval rule audit-ignore
* ja: name the accents and fix two symbols read as unrelated words
* ja: a tilde is チルダ (チルド is the loanword for chilled)
* ja: name the relations, the ellipses and the geometric translation
* ja: use the Japanese names for the symbols that state a relation
* ja: finish sweeping unicode.yaml (Greek letters, set membership, arrows)
* ja: set membership is 属さない, not 所属しない (belonging to an organisation)
* ja: read the function-application of as オブ (#731)
* ja: read the function-application of as オブ
* ja: the invisible function-apply character also reads オブ
* ja: read the remaining function applications with オブ too
These are the sites the first two commits could not reach while #730 was
open — the square root in ClearSpeak_Rules, SimpleSpeak_Rules,
SharedRules/general and overview — plus three where the connector was fused
into the preceding word, so grepping for a standalone one missed them: ∆ and
√ in unicode.yaml, and the square-root-of prefix in definitions.yaml. en
reads "laplacian of" and "square root of" as two words in those same places.
The exponent readings keep の: 3 の 2 乗 is ordinary Japanese and correct, as
are 分の, 正の and 負の.
SharedRules/calculus.yaml is brought in line too, but it is dormant — neither
ja nor en includes it, only fi, es, vi, sv, id and zh-tw do. Its four
operators reach speech through the function= entries in definitions.yaml
instead, and that is the path the new test covers. Two of the four had lost
the connector in this file altogether while the two beside them kept it.
Left alone on purpose: 順列の in SharedRules/general.yaml. en reads
"permutations of" there, but that rule already reverses its two arguments, so
the right Japanese wording is a decision rather than a swap.
* ja: fix the navigation announcements, four of which said the opposite
navigate.yaml is what a blind reader hears on every move, and the seeded
Japanese had not been checked against en. Four announcements had lost the
negation: "no previous row" and "no next row" were 前の行 and 次の行, so at
the top or bottom edge of a table the reader is told "the previous row"
and moves on believing it exists. They now read 前の行はありません and
次の行はありません, and the two column messages beside them are given the
same shape (前の列はありません; 次の列はありません was already right).
Twelve more strings had been replaced by unrelated Japanese words that
happen to start the same way. "inside" was インスタグラム (Instagram),
"read" was フィードバック (feedback), "describe" was コンテンツ (content),
"current" was 最近の投稿 (recent posts), "right" was アクセス (access),
"out" was インフォメーション (information), "move up" was サインアップ
(sign up), "move down" was シフトダウン (shift down), "pre-superscript"
was 事前原稿 (advance manuscript), "enhanced" was 強化強化 (doubled), and
"all of the way" was read as すべての方法 ("by every method"). "inside of
nothing more" was 何もない内部 and now says これ以上内側はありません.
Terms are made consistent: "column" was コラム (a newspaper column) in
eight places and 列 in one, now 列 everywhere; "character" was キャラクター
(a fictional character), now 文字; "part" was パーツ (machine parts), now
部分; "table" was テーブル (furniture), now 表; and "math" was 数学 (the
academic subject) where the file means the expression, now 数式.
One rule is reordered rather than reworded. At the right edge of an
expression the announcement is composed as cannot + move + right, which in
Japanese has to end with the verb; it now emits 右に + 移動 + できません,
matching 右に移動できません that two other rules in this same file already
say as one string.
Per AGENTS.md only these 58 verified lines are promoted from t: to T:.
The rest of the file stays t: — in particular the say-command prefix is
still the English 'zoom'/'move'/'read'/'describe' spoken through
x: "$Prefix" (ru, nb and fr translate it), and getting Japanese word order
there means putting the direction before the verb, so that is its own
change. There are no ja navigation tests yet either; both follow.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016JCoREgn1pJzcbdnUx4iuh
* ja: read nPk permutations with the school phrase, in display order
The pochhammer speech rule (reached from the nPk P-notation intents) copied
the en argument reversal and glued 順列の between the arguments, so 5 P 2 was
spoken as 2 順列の 5 — word salad in Japanese. en says "2 permutations of 5"
after Wikipedia's "k-permutations of n"; nb and sv rebuilt the phrase
entirely ("antalet permutationer av 2 element ur 5"). Japanese has a
standard school phrase that reads the notation in display order, so the
reversal is simply not needed: 5 個から 2 個取る順列, "permutations taking
2 out of 5".
The rule is marked audit-ignore the way nb's is, since the x:/T: sequence no
longer mirrors en. A speech test covers the msubsup form in ClearSpeak and
SimpleSpeak. audit-translations: untranslated text 3674 -> 3673, rule
differences 28 -> 28, missing/extra rules 0 -> 0.
This is the item #731's description left alone on purpose (順列の — "the
right Japanese wording is a decision rather than a swap").
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016JCoREgn1pJzcbdnUx4iuh
* ja: fix the rule vocabulary outside navigate.yaml
Five rule files still carried seed mistranslations of the same kind #746
cleared from navigate.yaml: doubled words, furniture and equipment where
mathematics was meant, and English word order glued between Japanese
tokens.
ClearSpeak multi-line labels: case was ケースケース, step was ステップステップ,
equation was 設備 (equipment) and line was ライン; now ケース, ステップ, 式
and 行, in both the count rule and the per-line rule.
SharedRules/default.yaml: "table with 3 rows and 4 columns" was テーブルと 3 行
および 4 コラム (the noun first, as in English, and コラム is a newspaper
column); it is now 3 行 および 4 列 の表. The menclose line notation was
ライン + 左から/アクセス/トップトップ/ボトム ("line" + "from the left" /
"access" / "top top" / "bottom"); it is now 左に/右に/上に/下に + 線, side first
and the noun last. "with 2 prescripts" was 付き 2 プレスクリプト; it is now
2 個の前置き添字 付き, and the overflow branch's 終了原稿 ("end manuscript",
the same seed error family as 事前原稿 in navigate) is 前置き添字終了, with
"and alternating prescripts" as および交互の前置き添字. These three are line
reorders inside a rule, no lines added or removed.
overview.yaml: "the 2 by 3 table" ended in テーブル (furniture); it now ends in
の表, the way 行列 is already read. definitions.yaml: "change in" (Δ) was 変更点
("modification point"), now 変化量. SharedRules/calculus.yaml: ラ・プラハシアン
was the en TTS respelling "LahPlahsian" copied as katakana; definitions.yaml
already says ラプラシアン, so it does now too.
Per AGENTS.md the 22 verified lines are promoted t: -> T:. audit-translations:
untranslated text 3674 -> 3652, rule differences 28 -> 28, missing/extra
rules 0 -> 0. Two tests cover the multi-line case label and the menclose
side line.
Left alone on purpose: "such that" (そのようなこと, 4 sites), round
(ラウンド値) and fenced-group (フェンスグループ) need a wording decision
rather than a swap.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016JCoREgn1pJzcbdnUx4iuh
* ja: finish the vocabulary in the same rules (postscripts, enclosure, matrix column)
Follow-up inside the same commit's scope, found by re-reading the rules the
first pass touched.
The postscripts branch is the mirror of the prescripts one and had the same
two faults: 付き N 投稿原稿 read "with N postscripts" in English order and
投稿原稿 is a manuscript submitted to a magazine; the overflow branch said
スクリプトの変更と変更 ("script change and change") and エンドスクリプト. They
now read N 個の後置き添字 付き, および交互の添字 and 添字終了, matching the
prescripts wording.
menclose's own enclosure markers were transliterations: エンクロージャー and
エンドエンクロージャ. The marker is emitted before the content, so a verb
phrase does not work; it is 囲み ... 囲み終了, the marker/end-marker shape this
language already uses (上付き ... 上付き終了). The side pause moved with the
noun, so the announcement is 左に 線, ... rather than 左に, 線 ....
The matrix column rule still said コラム after the table rule was fixed; it is
now 列 like the rest.
audit-translations: untranslated text 3674 -> 3645, rule differences 28 -> 28,
missing/extra rules 0 -> 0.
Left alone on purpose: ロングディビジョンシンボル ("long division symbol")
needs a wording decision, as do such-that, round and fenced-group.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016JCoREgn1pJzcbdnUx4iuh
* ja: unify the laplacian reading in unicode-full.yaml too
Caught in review. unicode-full.yaml was already inconsistent with itself:
the verbose branch said ラプラシアン の while the two terse branches still
said ラ・プラハシアン, the en TTS respelling "LahPlahsian" copied as katakana.
All three now say ラプラシアン, matching definitions.yaml, unicode.yaml and
SharedRules/calculus.yaml.
This file is otherwise being left for the MathPlayer-derived seed; these
three lines are only the reading of one word already decided elsewhere, so
a reseed can overwrite them harmlessly.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016JCoREgn1pJzcbdnUx4iuh
* ja: sweep the same multi-line labels in the SimpleSpeak path
#747 has landed, so SharedRules/general.yaml is free to touch. It carries the
SimpleSpeak copy of the multi-line labels this branch already fixed in
ClearSpeak_Rules.yaml, and leaving it would have made the two styles disagree:
ClearSpeak saying ケース and SimpleSpeak ケースケース for the same piecewise.
Both rules in general.yaml now read ケース, 式 and 行, and the test asserts the
SimpleSpeak path as well as ClearSpeak.
The rest of general.yaml (vector, end, sub, the chemistry terms) is a separate
batch, not this one.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016JCoREgn1pJzcbdnUx4iuh
* ja: assert the SimpleSpeak multi-line labels too
The SimpleSpeak path classifies this table as a system of equations rather
than cases, so it reads 2 式 / 式 1 / 式 2 through SharedRules/general.yaml
while ClearSpeak with MultiLineLabel=Case reads ケース. Both are now asserted.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016JCoREgn1pJzcbdnUx4iuh
* Fix Japanese vocabulary in SharedRules (general.yaml, default.yaml)
Machine-translation artifacts in the shared rules, where the wrong word is
spoken every time the rule fires. Every occurrence of each word in
Rules/Languages/ja/SharedRules/ is fixed, so the two files stay consistent.
general.yaml
- vector -> コンテンツ ("content"), 7 places, including the announcement of
every column/row vector. definitions.yaml already reads vector as ベクトル.
- end (of matrix/vector/determinant) -> リリース ("release"), 5 places. The
file already says 行列終了 for the combined "end matrix", so the branches
now read 行列終了 / ベクトル終了 / 行列式終了 and the separate "end" token
is emptied, the way the file already empties the English "the".
- by 1 column -> 1列ずつ ("one column at a time"), 2 places. The row rule
beside it says かける, so the column rule now reads "3 かける 1 列".
- N (the letter, in N squared) -> ネクタイ ("necktie"). The four sibling
letters read シー キュー アール ゼット; definitions.yaml reads ℕ as エヌ.
- Chemical states: liquid -> 液体液体 (doubled), gas -> ガスレンジ ("gas
cooker"), aqueous -> アキュース (not a word). Now 液体 / 気体 / 水溶液,
matching 固体 already there.
- Bonds: quadruple bond -> 四倍の結束 ("quadruple solidarity") -> 四重結合.
The three siblings were transliterations, so all four now use the standard
Japanese terms 単結合 / 二重結合 / 三重結合 / 四重結合.
Both files
- sub -> サブサブ (the word "sub" doubled) and super -> スーパー (which in
Japanese is a supermarket). The verbose branch beside them already says
下付き文字 / 上付き文字, so the short branches are now 下付き / 上付き.
- The matching end markers were English word order or transliterations
(終了サブスクリプト, エンドサブ, エンドスーパー, 終了スーパースクリプト).
Japanese puts the noun first, as ClearSpeak_Rules.yaml already does with
上付き終了 / 根号終了 / 分数終了, so they are now 下付き終了 /
下付き文字終了 / 上付き終了 / 上付き文字終了.
- end (of a fraction, spoken to blind readers) -> リリース in default.yaml,
now 分数終了 to match ClearSpeak_Rules.yaml.
Lines whose reading is verified are promoted from t: to T: per AGENTS.md.
audit-translations ja: untranslated 3577 -> 3493 (-84, the number of lines
promoted); missing rules 0, extra rules 0, rule differences 28 unchanged.
Adds ja tests for the number-set letters, the vector/matrix/determinant end
markers, the short sub/superscript readings, and the chemical states and
bonds.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016JCoREgn1pJzcbdnUx4iuh
* Translate the Japanese navigation command prefixes
`say-command` still set the four prefixes to the English words, so a Japanese
synthesiser read them out as English: "move 右", "zoom イン", "read 右",
"describe 右". They are now 移動 / ズーム / 読み上げ / 説明.
`$CommandOffset` (from ec36e057) already makes this safe: the offset is the
length of the English `$NavCommand` stem, not of the spoken prefix, so the
suffix test still matches. ja already carried the correct offsets (5/5/5/9).
ru, nb, fr, hu and sv translate their prefixes the same way.
Word order is the other half. Japanese puts the target before the verb, and
the particle belongs to the verb, so a new `$Particle` is set beside the
prefix (に for move, を for read and describe) and emitted inside each
direction branch: 右 に 移動, 右 を 読み上げ. Keeping the particle inside the
branch means a command whose suffix matches nothing still falls back to the
bare verb, the way it does in en.
The Zoom commands keep the English order: ズーム + イン is the ordinary
loanword, and the concatenation that produces ズームインを最大にしました
requires the prefix to come first. The two suffix sets are disjoint (Zoom only
produces In/InAll/Out/OutAll, the others only Next/Previous/Current/LineStart/
LineEnd), so splitting the branch loses nothing.
Two other rules in the same file spoke the same words in the other order and
would have contradicted this, so they move too:
- `current` (ReadCurrent / DescribeCurrent) said 読み上げ 現在; it now says
現在 を読み上げ.
- `move-next-no-auto-zoom-at-edge-math` said 右に for all three verbs, so
read and describe got the wrong particle; the direction and particle are now
part of each verb branch (右に移動 / 右を読み上げ / 右を説明 できません).
audit-translations ja: untranslated text 3577 -> 3574; missing rules 0, extra
rules 0. Rule differences 28 -> 30, both in navigate.yaml and both intended:
one "variable difference" for the added `$Particle`, one "structure
difference" for the reordered branch.
Adds tests/Languages/ja/navigate.rs, the first navigation tests for ja. They
assert the whole command prefix -- everything up to the first pause -- because
the description that follows comes from NavigationParts and is not what this
change touches.
Wiring it up needs one line in tests/languages.rs, which is outside
Rules/Languages/ja and tests/Languages/ja.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_016JCoREgn1pJzcbdnUx4iuh
* ja: read the remaining fractions denominator-first, and name the menclose marks
Two groups of fixes in the shared rule files.
1. Three fraction rules read the fraction backwards. Japanese puts the
denominator first: 3/4 is "4 分の 3", so "3 分の 4" is 4/3 -- a different
number, not just an odd phrasing. ClearSpeak and SimpleSpeak already swap
the operands (common-fraction), but the mfrac rules in SharedRules/default.yaml
(used by LiteralSpeak) and the fraction rule in overview.yaml (used by the
"describe" navigation command) did not.
- literal-simple and overview-default now say the denominator first.
- literal-default has explicit start/end markers, so it keeps the written
order and uses the borrowed オーバー: 分数 A オーバー B 分数終了. It also
opened with スタート, the English "start".
2. menclose notation names and two neighbouring rules were still the machine
translation:
- 上下矢印 for "up arrow" is an arrow with a head at each end, so uparrow,
updownarrow, leftrightarrow and both diagonal double ended arrows all
named the same thing.
- ダブルは対角矢印を終了 / 二重は対角矢印を終了 ("double is ends diagonal
arrow") are not sentences.
- 行動シンボル is an "action symbol"; the actuarial symbol is 保険数理記号.
- アラビア語の要素のシンボル reads "Arabic-language element symbol"; madrub
is the factorial symbol written in Arabic script.
- クロスアウト, ダウン矢印, ダウンの対角, サークル, ラウンドボックス, ボックス and
ロングディビジョンシンボル are transliterations of the English.
- 対角形 does not say which diagonal.
- 詳しくはこちら ("click here for details") was used for "above" in the mover
and munderover rules, opposite the 下 that is already there.
- ノルダム is Notre-Dame; a norm is ノルム.
- エントリー for a matrix entry is 成分, and a coordinate is 点, the word
geometry.yaml already uses.
Every line that was checked against the en rule beside it is promoted to T:.
Tests cover both fraction paths, the overview path, the menclose marks, above
and below, the norm and the matrix entry.
* ja: match the test expectations to the actual output
* ja: use the arrow and box names the ja files already use
* ja: drop the decimal point assertion; a plain mn is spoken as one token
* ja: use readings for the relations that work between the operands (#757)
* ja: use the source's readings for the comparison and element-of symbols
`unicode.yaml` puts the relation word between the two operands, so the word has
to work there. Several did not.
- `>` said より大きい. より needs its comparand in front of it, so "x は より大きい 5"
parses as "x is a bigger 5". It also did not match the 小なり already used for
`<` in the same file. Yamaguchi, Kawane & Sawazaki (1996) give the pair
小なり / 大なり, and 小なり・オア・イコール for `≤` (the ・ is only a word
separator and is not read), so `≥` follows as 大なり・オア・イコール.
- `≠` said 等しくない, a predicate with the same problem. The source gives
ノット・イコール, which pairs with the イコール already used for `=`.
- `∈` and `∊` said 要素の for "element of". の attaches backwards: 要素の A is
"A of an element". The source gives 要素オブ, the same borrowed オブ this
translation already uses for "of" elsewhere. The two rules also disagreed with
each other -- ∈ said 属する where ∊ said 要素の, for the same English.
- The ClearSpeak "In" option said 中へ, which is "into", a direction of motion.
The source's reading for "in" is the borrowed イン.
- `∾` said 最も肯定的な, "most affirmative"; en is "most positive" in the numeric
sense, so 最も正の.
Left alone: the negative set relations (∉ ⊄ ⊅) and the subset family
(⊂ ⊃ ⊆ ⊇). The source gives 部分集合オブ for ⊂ but nothing for the superset or
the negated forms, and I did not want to change half a family.
* ja: update the tests that pinned the old relation wording
* ja: sweep the same relations in definitions.yaml and cover the set branches
* ja: flatten the set-builder test expression
* ja: use the set-builder shape the en tests use
* ja: member-of said に属する where unicode.yaml says 元
* ja: make the rest of the set relations work between the operands
#757 left the negated membership and the subset family alone: the source gives
部分集合オブ for ⊂ but nothing for the superset or the negated forms, and I did
not want to change half a family. This does the whole family at once.
- ⊂ said 部分集合 and ⊆ said 部分集合または等しい. A bare noun, or a predicate
that ends the sentence, strands the set that follows it: "A は 部分集合 B"
reads as "A is a subset" and then a B. The source gives 部分集合オブ for ⊂ and
部分集合（オブ）・オア・イコール for ⊆, where the オア イコール is the one this
translation already uses for ≤ (the ・ is only a word separator, not read).
- ⊃ and ⊇ keep the 上位集合 the seed chose, since the source gives no superset
reading, and take the same オブ so they work in the same position.
- ∉ ⊄ ⊅ ended in ではない / でない, which closes the sentence in front of the set
the statement is about. The source gives ノット・イコール for ≠, so the negation
is that same ノット in front of whatever the positive form says.
- The ClearSpeak Member option said 元 on its own -- the same missing-preposition
shape 要素の had in #757 -- and outside a set the In option said に含まれる,
whose に attaches to what comes before it and so reverses the relation.
- definitions.yaml carries the same readings for intent, so the eight entries of
this family are swept with it.
Left alone: the Belongs option (に属する / 属する / 属さない). Its Japanese is a
verb phrase, so making it work between the operands is not a matter of adding
the borrowed オブ; it needs a noun the source does not give. It stays t: in both
polarities, and is the one branch of ∈/∊/∉ this commit does not promote.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XzNYYoiVRzKGnDEWTJnuHh
* ja: navigate.yaml stopped speaking English, and now uses Japanese word order
Three of the navigation announcements were still in the shape en gives them.
- The 2D move announcements spoke $Move2D as it stands, and that variable holds
an English literal: 'in', 'out of', 'end of', 'start of'. Zooming into x
squared said "ズーム イン; in 底; x", with the "in" read out in English. Those
four are the only values this file ever sets, so the two rules that emit it
now map them (に入る / から出る / の終わり / の始め) and say them after the part
they apply to, which is where Japanese wants them: "ズーム イン; 底 に入る; x".
- Undo named the verb first: 元に戻す ズームイン is "undo zoom in" word for word.
Japanese names what is being undone first, ズームイン を 元に戻す.
- The placemarker announcements had the same order -- 読み上げ プレースホルダー 3
for "read placeholder 3". They now name the placeholder and its number first
and then take the particle the matching move command already uses (を 読み上げ,
を 説明, に 移動), the pattern #752 set up. セット becomes 設定, the ordinary
Japanese for setting a value.
The four rules whose children are reordered carry `# audit-ignore`. The
divergence from en is the point of the change, and without the marker the rule
difference count goes 30 -> 34 and stops being usable as evidence that a PR did
not break the rules. The marker sits at the end of the `- name:` line: that is
the position that works today, and the one 170 of the 298 markers under
Rules/Languages already use. (#742 is deciding which position becomes canonical;
if it moves, these move with the rest.)
This also promotes 底 / 上付き文字 / 下付き文字 to T:, checked against en's base /
superscript / subscript. In en the `phrase(...)` hints on those lines and the
values they annotate do not agree with each other; ja follows the values, which
are what is actually spoken.
audit-translations ja: untranslated text 3447 -> 3436, which is the 8 lines this
changes plus those 3 promotions; rule differences 30 -> 30; missing, extra and
definition counts unchanged at 0.
Tests: four in navigate.rs, with a helper that compares the whole announcement
rather than the command prefix, since none of these three is a prefix.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XzNYYoiVRzKGnDEWTJnuHh
* ja: mark the reviewed navigation strings as verified
Lowercase `t:` means untranslated or unverified and uppercase `T:` means
translated and verified, so this is a review rather than bookkeeping. Each of
the twenty-three lines was read against the value the English source carries in
the same rule -- the value, not the `phrase(...)` hint, because those two can
disagree: move-cell-next says "no next column" under a hint that reads "no next
row", and the Japanese follows the value.
Nineteen lines were already right and only change case. Four were the odd one
out among their own siblings and are reworded as well:
最初の文字にズームイン -> 最初の文字までズームしました
the other two zoom reports in that same test are sentences
(ズームインを最大にしました, 文字までズームしました); this one was a bare
noun phrase and did not read as a report of where the zoom ended up.
現在のエントリを読む -> 現在のセルを読む
エントリ appears exactly once in the whole Japanese rule set. The rule
speaks 行 and 列 straight after it, and the sibling rule says 表, so セル is
the word these rules already use for the thing being read.
シンプル -> 簡易
the three navigation modes are spoken as <word> + モード, and 文字 and 拡張
are ordinary words while this one was a transliteration between them.
移動後の式の概要 -> 移動後に式の概要を読み上げる
the two states of one toggle: the other is the sentence
移動後に式を読み上げる, and this one did not say what would happen to it.
There is no Japanese appendix for navigation wording, unlike the ClearSpeak
terms, so those four are a judgement call and each one has a test.
audit-translations ja: untranslated 3415 -> 3392, which is exactly the number of
lines promoted; rule differences unchanged at 30; missing, extra and definition
counts unchanged at 0.
Tests: four in navigate.rs. Reaching the first-character branch needs Character
mode, so the helpers grew a mode argument; the sequence is the one
move_inside_leaves in src/navigate.rs uses -- ZoomInAll to the leaf, then one
more ZoomIn to step inside it.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XzNYYoiVRzKGnDEWTJnuHh
* ja: the phrases that were carried over word for word
Each of these was translated token by token, in English order, so the Japanese
carries the words and not the meaning. Where a phrase is split across several
`t:` entries the fix moves material between them, which is what the merged
navigation work (#752, #773) already did; the en `phrase(...)` comments are left
alone.
{x | x > 2}
集合 すべて x そのようなこと x は 大なり 2
-> 集合 すべての x ただし x は 大なり 2
そのようなこと is "such a thing": it translates the words of "such that" and
none of its work, which is to attach a condition to what was just named.
ただし is what Japanese mathematical prose uses. 集合 stays in front of the
variable rather than moving to the head-noun position Japanese would
normally give it, because in a speech stream it tells the listener what
they are about to hear, and the first argument can be long ({x ∈ ℤ : x > 5}).
P(A | B)
A 与えられた B -> A 条件は B
与えられた is the past participle "given", which in Japanese modifies the
noun after it, so B was being described rather than named as the condition.
a labelled row
行 1 ラベルを使って 第1式 -> 行 1 ラベルは 第1式
ラベルを使って is "using a label", which says the row does something with
one. The label is simply what the row is called.
f|_a^b
f 評価される b 同じ式が評価されるマイナス a
-> f 次の点で評価 b マイナス 同じ式を次の点で評価 a
評価される is the passive "is evaluated" with nothing to say where, and the
minus came last, so the subtraction arrived after the thing subtracted.
round() ラウンド値 ("round" in katakana + 値) -> 丸めた値
fenced-group フェンスグループ, an English term in katakana -> 括弧のまとまり
Each phrase is swept across every file that carries it, including the sibling
branches: unicode.yaml has two more `given`s, general.yaml a second
`evaluated at`, and definitions.yaml the intent mappings for such-that, given
and conditional-probability. The copy of `such that` in unicode-full.yaml is
left alone, since that file is waiting on a MathPlayer seed.
audit-translations ja: untranslated 3415 -> 3400, which is exactly the number of
rule lines promoted; rule differences unchanged at 30; missing, extra and
definition counts unchanged at 0. definitions.yaml entries carry no case marker,
so they do not move that count.
Tests: five new in ja.rs and one updated, since set_builder_member_symbol had
pinned the old reading. Reverting the readings and nothing else turns exactly
those six red.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01XzNYYoiVRzKGnDEWTJnuHh
* ja: rebuild unicode-full.yaml readings and take the good terms from MathPlayer (#929)
* Add UEB G1 BANA/GTM mode preference and expand Grade-2 contractions.
Adding the preference is a stall tactic while I wait for feedback from ICEB about whether BANA rules are obsolete.
Introduce UEB_G1_BANA_Style (default true) to choose BANA 2019 vs ICEB GTM
1.7.3/1.7.5 grade-1 indicator placement in ueb_cleanup. Implement GTM 1.7
per-symbols-sequence counting with gtm_1_7_mode, and add matching ICEB tests.
Extend handle_contractions for technical vocabulary: UEB 10.8 final-letter
groupsigns, dot-5 through/part, dis/con prefixes, cc/ea exception words, and
a ful-pattern fix. Add ueb_technical_word_contractions (311 words, goldens
cross-checked with brailletranslators.com). Wire the preference through
prefs, fuzz, and docs.
* updated to new BANA and ICEB spec updates. Five tests fail. Still working on them
* removed an option
* fix contractions involving 'be, con, dis' which can only be at the start of a 'word' (where 'word' has a specific UEB meaning)
* Updated to UEB to the latest guidance for grade 1 indicators. This adds new tests and changes a few tests.
Also added more contractions.
* update rc #
* fix new clippy warnings
* fix up Vietanmese tests
* fix str compile issue with no-unsafe feature
* add tests for intervals in German
* Refine German interval descriptions and update translation rules
* Add German ClearSpeak rule and tests for pause intent
* Add German localization for currency symbol placement and corresponding tests
* add another test for euro
* add another test for pound/pfund
* Update German ClearSpeak rules: Add support for principal logarithm (Log) and corresponding tests
* Refine German ClearSpeak fraction translations and update ordinal fraction definitions
* German ClearSpeak: subset & superset
* de: fix chemical state names
* de: fix chemical reaction arrow
* de: fix integer exponentiation
* de: fix exponentiation by arbitrary exponents
* de: add vector norm
* de: fix "transpose"
* replace lamda with lambda (#789)
* replace lamda with lambda
* fix line endings
* fix(pl): przywróć szablony sum i iloczynów po naprawie słowa łączącego (#783)
* Refactor/simplify prefs rs (#780)
* remove pointless cloning
* simplify logic flow of is_user_pref
* simplify path joins
* Add comment to clarify update condition
Added a comment to prevent unnecessary updates if the value hasn't changed.
---------
Co-authored-by: NSoiffer <NSoiffer@users.noreply.github.com>
* ci: add scheduled workflow to alert translators of English rule changes
* Add comment to cron schedule in translator-alert.yml
* de: fix "k-th root"
* de: fix "negativ positiv"
* de: fix RPM singular/plural
* Set CRLF line endings for .rs and .yaml files
* Add support for stdin, inline MathML, and default fallback
* Re-add addon info to user guide
Added back the info on how to use MathCAT as an addon to NVDA, which was accidentally removed.
* Add rule order checks and reporting for shared rule mismatches
* fix ruff
* uv run ruff format audit_translations/
* Refactor: Replace custom path utilities with `Path` API and update zip methods accordingly (#775)
* Refactor navigation state management: unify position and command stacks into a single history vector (#776)
* Refactor navigation state management: unify position and command stacks into a single history vector
* fix cargo clippy no-unsafe
* simplify `read_prefs_file` (#808)
* Refactor `read_prefs_file` to simplify error handling and optimize section processing
* use with_context(..) from anyhow-crate for error-handling
* Replace `unwrap()` with `let Some(...) else` constructs for safer option handling
* simplify destructuring syntax using tuples
* use elements.rs as chemistry database
* fix redundant call of pref_to_string (#815)
* clean up unused types in definitions and remove outdated docstring (#818)
* remove dead functions set_string_value and set_bool_value (#819)
* replace custom constructor with struct literals for TTSCommandRule (#821)
* Add GitHub Actions workflows to measure and report library size changes
* Refactor library size comment script to improve readability and error handling
* Update README with supported and in-progress speech languages list
* replace buggy Display implementation for Pronounce with Debug trait (#820)
* replace buggy Display implementation for Pronounce with Debug trait
* remove unused `eloquence` field and related code from `Pronounce` struct
* fix outdated syntax
* define ratio (U+2236)
* added/modified comments for functions that deal with function names and parts
* Swap IntentMappings so ":" is argument glue and "|" is verbosity -- this looks much better. All language `definitions.yaml` are updated.
Fix parsing of `IntentMappings`. Parsing now lives in one place, so ratio-style "name : glue" pairs, arity templates, and bracketing speech stay consistent across languages.
This all started as a fix to #709, but then I discovered that the format used in `IntentMappings` was parsed incorrectly in some cases (such as "ratio").
* shorten 'ratio' def after debugging
* fix clippy warnings (about comments)
* Normalize tracked files to match .gitattributes (#843)
* Create rule coverage functionality (#834) (#845)
* v1 of rule coverage report
* Add Python script to generate Markdown and HTML rule coverage reports
* clarify name of logfiles
* Switch rule coverage events to JSONL format
* switch to jinja2 templates
* Exclude Unicode mapping files from rule coverage report and improve report generation
* Exclude definition files from rule coverage report and update related documentation
* Add hit counts and test details to rule coverage report
Enhance the rule coverage report to include hit counts, test details, and hoverable tooltips for matched rules. Improve report formatting and coverage fraction calculation. Update related tests and documentation to reflect these changes.
* fix title
* Simplify rule coverage event recording
* Stream rule coverage events and trim unused report fields
* Suppress unused variable warning when `rule-coverage` feature is disabled
* Migrate `rule_coverage` implementation to a dedicated `rulecoverage` package, separating logic, templates, and CLI.
* uv run ruff format
* simplify dataclasses in Python
* remove "loaded" as statistic for rule coverage
* remove MD report and only keep HTML.
simplify rest accordingly.
* uv ruff format
* Rules/Languages/hu/definitions.yaml: add a new rule, based the english new definitions.yaml rule (the of word related)… (#842)
* Rules/Languages/hu/definitions.yaml: add a new rule, based the english new definitions rule
Signed-off-by: Attila Hammer <hammera@pickup.hu>
* Update tests, add navigation test and table separator test
Signed-off-by: Attila Hammer <hammera@pickup.hu>
* Rules/Languages/hu/definitions.yaml: added missign navigation related fixes with not merged to the main branch
Signed-off-by: Attila Hammer <hammera@pickup.hu>
* tests/Languages/hu/mtable.rs: added a not merged table separator tests related modification
Signed-off-by: Attila Hammer <hammera@pickup.hu>
* convert tests/Languages/hu/navigate.rs with crlf format
Signed-off-by: Attila Hammer <hammera@pickup.hu>
* Rules/Languages/hu/navigate.yaml: remove unneed in suffix after hungarian zoom part translation (the nagyítás word related test)
Signed-off-by: Attila Hammer <hammera@pickup.hu>
* tests/Languages/hu/navigate.rs: adapt the nagyítás (zoom in) related changes to the test file to get correct translation
Signed-off-by: Attila Hammer <hammera@pickup.hu>
* Rules/Languages/hu/navigate.yaml: fixed few translations, based with experiences
Signed-off-by: Attila Hammer <hammera@pickup.hu>
* Rules/Languages/hu/navigate.yaml: added audit-ignore with required places
Signed-off-by: Attila Hammer <hammera@pickup.hu>
* tests/Languages/hu/navigate.rs: add new navigation tests, based with last hungarian navigation translations
Signed-off-by: Attila Hammer <hammera@pickup.hu>
---------
Signed-off-by: Attila Hammer <hammera@pickup.hu>
* src/canonicalize.rs: apply clippy fixes with two places (#850)
Signed-off-by: Attila Hammer <hammera@pickup.hu>
* Fix intent detection for limits and norms (#840)
This commit improves intent recognition for several TeX-derived patterns: it ignores the invisible function-application character when matching limit expressions, treats TeX's gathered environments as lines, and recognizes subscripted norms from TeX4ht and canonicalized MathML shapes. These fixes keep speech and navigation output consistent for common display math cases.
* Improve translator alert workflow by including file links in notifications (#852)
* docs: create the CONTRIBUTING.md file for newcomers (#860)
* deps: upgrade zip to 9.x (#861)
* Add lang-* and braille-* features to choose which rules include-zip embeds (#816)
Every language and braille code has a feature, all on by default through rules-all, so existing builds embed exactly the same rules.zip. A build that turns off default features and names a subset embeds only those directories. en is always kept because the library falls back to it, a subset with no braille code keeps UEB so set_rules_dir still succeeds, and the staged prefs.yaml BrailleCode default is pointed at a kept code when Nemeth was pruned. build.rs fails if a Rules directory has no feature in lang-all or braille-all, so the lists cannot silently fall behind.
Co-authored-by: NSoiffer <NSoiffer@users.noreply.github.com>
* Simplify `is_leaf_any_name` (#829)
* simplify some Vector handling (#828)
* simplify some Vector handling
* undo changes in `is_structural`
* Refactor comma-check logic in `is_leaf_any_name` (#830)
* Remove `\mathmit` as that is non-standard. Fixes undefined italic `pi` in RC not in beta (although the `\mathmit` is in the beta).
Fixes #857
* remove duplicate entries
* Bring IntentMappings “of”/glue syntax from main onto release-candidate.
Use ":" for argument glue and "|" for verbosity, move the function-application
word into FunctionApplicationWord, and sync language definitions plus de/pl
tests and Polish unit-inflection rules so they match.
* fix clippy warnings
* Copy navigation MathML into the temporary document for no-unsafe.
`get_navigation_braille` wraps the focused node in a new `<math>` element that lives in a temporary Package. `copy_mathml` built that copy in the source document, and the result was then parented under the temporary `<math>`.
With the default sxd build an Element is a raw pointer, so that cross-document append still found the node. With --features no-unsafe an Element is only an index into its own document's arena. append_child looks that index up in the parent document, so the temporary document either panicked (index out of bounds) or attached the wrong node. navigation_braille_copies_focused_node_into_temporary_document failed only on the no-unsafe build.
`copy_mathml` and `copy_mathml_recursive` now take the destination Document and create every new element there, including the max-depth leaf. Names, attribute values, and text are copied as strings and re-interned in that document. `get_navigation_braille` passes the temporary document. The navigation intent copy passes `intent.document()`, so it stays in the same document as before.
Note: this is slightly more efficient because the temporary copy is in a different package which gets dropped/cleaned up instead of staying in the arena.
Fixes #827.
* Update the version of zip due to security concerns.
Update the MathCAT version #.
* Store in-memory file keys the same way they are looked up (#877)
* don't ignore changes to Cargo.lock (updated Rust guidance for libraries)
* update version # and zip version -- I thought I had already done this
* fix typos recently introduced into LaTeX unicode file
* Remove the tc: marker from LaTeX output (#876)
* bump version number again after some tweaks
* version # bump
* fixes after bumping zip version
* Add Windows testing matrix to CI workflow for broader platform coverage (#878)
* The rule for
```
"𝟎-𝟗": # 0x1d7ce - 0x1d7d7
```
was wrong, resulting in a infinite loop (that gets caught).
* bump version #
* The rule for
```
"𝟎-𝟗": # 0x1d7ce - 0x1d7d7
```
was wrong, resulting in a infinite loop (that gets caught).
* update to actual release. Need to bump number because of NVDA addon store not allowing rc suffix
* Bump anyio from 4.12.1 to 4.14.2 in /PythonScripts (#879)
Bumps [anyio](https://github.com/agronholm/anyio) from 4.12.1 to 4.14.2.
- [Release notes](https://github.com/agronholm/anyio/releases)
- [Commits](https://github.com/agronholm/anyio/compare/4.12.1...4.14.2)
---
updated-dependencies:
- dependency-name: anyio
dependency-version: 4.14.2
dependency-type: indirect
...
Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
* Bump h2 from 4.3.0 to 4.4.1 in /PythonScripts (#880)
Bumps [h2](https://github.com/python-hyper/h2) from 4.3.0 to 4.4.1.
- [Changelog](https://github.com/python-hyper/h2/blob/master/CHANGELOG.rst)
- [Commits](https://github.com/python-hyper/h2/compare/v4.3.0...v4.4.1)
---
updated-dependencies:
- dependency-name: h2
dependency-version: 4.4.1
dependency-type: indirect
...
Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
* Bump hpack from 4.1.0 to 4.2.0 in /PythonScripts (#882)
Bumps [hpack](https://github.com/python-hyper/hpack) from 4.1.0 to 4.2.0.
- [Changelog](https://github.com/python-hyper/hpack/blob/master/CHANGELOG.rst)
- [Commits](https://github.com/python-hyper/hpack/compare/v4.1.0...v4.2.0)
---
updated-dependencies:
- dependency-name: hpack
dependency-version: 4.2.0
dependency-type: indirect
...
Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
* Bump soupsieve from 2.8.3 to 2.9 in /PythonScripts (#883)
Bumps [soupsieve](https://github.com/facelessuser/soupsieve) from 2.8.3 to 2.9.
- [Release notes](https://github.com/facelessuser/soupsieve/releases)
- [Commits](https://github.com/facelessuser/soupsieve/compare/2.8.3...2.9)
---
updated-dependencies:
- dependency-name: soupsieve
dependency-version: '2.9'
dependency-type: indirect
...
Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
* Bump idna from 3.11 to 3.15 in /PythonScripts (#884)
Bumps [idna](https://github.com/kjd/idna) from 3.11 to 3.15.
- [Release notes](https://github.com/kjd/idna/releases)
- [Changelog](https://github.com/kjd/idna/blob/master/HISTORY.md)
- [Commits](https://github.com/kjd/idna/compare/v3.11...v3.15)
---
updated-dependencies:
- dependency-name: idna
dependency-version: '3.15'
dependency-type: indirect
...
Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
* Bump requests from 2.32.5 to 2.33.0 in /PythonScripts (#885)
Bumps [requests](https://github.com/psf/requests) from 2.32.5 to 2.33.0.
- [Release notes](https://github.com/psf/requests/releases)
- [Changelog](https://github.com/psf/requests/blob/main/HISTORY.md)
- [Commits](https://github.com/psf/requests/compare/v2.32.5...v2.33.0)
---
updated-dependencies:
- dependency-name: requests
dependency-version: 2.33.0
dependency-type: indirect
...
Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
* Bump pygments from 2.19.2 to 2.20.0 in /PythonScripts (#886)
Bumps [pygments](https://github.com/pygments/pygments) from 2.19.2 to 2.20.0.
- [Release notes](https://github.com/pygments/pygments/releases)
- [Changelog](https://github.com/pygments/pygments/blob/master/CHANGES)
- [Commits](https://github.com/pygments/pygments/compare/2.19.2...2.20.0)
---
updated-dependencies:
- dependency-name: pygments
dependency-version: 2.20.0
dependency-type: indirect
...
Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
* Bump pytest from 9.0.2 to 9.0.3 in /PythonScripts (#887)
Bumps [pytest](https://github.com/pytest-dev/pytest) from 9.0.2 to 9.0.3.
- [Release notes](https://github.com/pytest-dev/pytest/releases)
- [Changelog](https://github.com/pytest-dev/pytest/blob/main/CHANGELOG.rst)
- [Commits](https://github.com/pytest-dev/pytest/compare/9.0.2...9.0.3)
---
updated-dependencies:
- dependency-name: pytest
dependency-version: 9.0.3
dependency-type: direct:development
...
Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>
* Added explainer video (#888)
* Refine when and how much fuzzing happens on pushes. This should greatly shorten the runs when doing a push. The weekly long fuzzes should continue unchanged.
* bump crate "anyhow" to 1.0.103 (#868)
* **bump crate "dirs" to 7.0.0** (#869)
* Fix ASCIIMath bold digits translating from sans-serif digits (#874)
* fix wrong check for CheckRuleFiles which was backwards. Should help with speed a little when dealing with unicode-full.yaml
* This fixes a bug that is very unlikely for someone reading math, but sometimes showed up in running the tests in github. The failures were a test-suite pattern, not a normal speech pattern and involve reading a new unicode-full.yaml file.
The panic needs three things at once. The character has to miss the short Unicode table, its rule has to look up another character in the full table, and that inner lookup has to decide the file must be read again while the outer lookup still holds the table. Cyrillic capitals do the first two: А spells the lowercase letter, which also lives only in unicode-full.yaml. Everyday Latin math stays in the short table and never reaches this path.
The third step is the rare one. The full table is read once, then kept. With the default CheckRuleFiles value, Prefs, later lookups do not check the file's timestamp, so the inner lookup does not try to read it again. The CI runs hit that third step about 10% of the time because one test process switches language over and over, clears the file list between tests, and sometimes the next timestamp check disagrees. A screen reader does not do that between a capital and the lowercase letter it spells.
The fix is to stop a nested character lookup from reloading unicode-full.yaml while the outer lookup still holds the table.
replace_single_char inverted the CheckRuleFiles test used by read_files, so the default Prefs value re-stat'd the full table on every short-table miss. The rule is copied *only* for the lookup that reads the file; later lookups borrow it.
This adds a little extra code but avoids a copy (clone) for every unicode-full.yaml lookup.
* canonicalize double exclamation !! to speak as double factorial (#837)
* simplify definitions.rs (#826)
* simplify definitions.rs
* reintroduce manual tweaking of hashmap capacity
* refactor: simplify `is_simple` and `is_trivially_simple` logic (#825)
* refactor: simplify `is_simple` and `is_trivially_simple` logic
* Refactor is_simple function for better readability
Put logical connectors at end like rest of code.
Also align conditions
---------
Co-authored-by: NSoiffer <NSoiffer@users.noreply.github.com>
* Refactor YAML parsing functions (#835)
* refactor some more iterators in speech.rs (#867)
* Merge in 'ru' branch after several updates (#798)
* Ru - Improve wording of Russian docs (#661)
* update acknowledgment for Russian braille
* update version number because added Russian braille
* UEB has a rule for some pseudo-script chars. The match for that rule was overly broad, so I removed scripted letters and numbers. That rule made its way into other braille codes, so I removed them everywhere. All tests pass.
* Added new UEB chars. Many of these are accented letters. I added 'A' (for accented letters) so that the pattern for them looks like:
```
- "Ñ": [t: "CA⠘⠻L⠝"] # 0x00D1
```
Added a few tests.
* Translate German absolute-value ending
* Fix German unit plurals
* Fix German pluralization for "ruthe" and add regression test
* Correct German translations for "ruthe" to "rute" and update regression test
* Expand German unit definitions and add regression test for pluralization
* fix some symbols
* fix german set grammar
* typo in example
* Russian Braille fixes
This commit adds fixes to Russian Braille. Now the test with parentheses is passed. Also there are devision and multiplication signs have been modified according to the Russian braille specs.
* Update Russian Braille rules and mappings
Fix YAML replace syntax and extend Russian braille rules: add chemical-formula-operator-bond and prefix-geometry-ops rules, broaden function-name matching (include tg/ctg variants) and add separator markers. Update unicode mappings (degree sign, invisible characters, conditional ∪ mapping, parallel/arrow glyphs) and adjust several function/operator braille outputs. Add/expand tests to cover functions, geometry, matrices and chemical formulas to validate the changes.
* Add Russian language to docs
Add Russian (ru) to the list of supported speech languages and update Russian documentation and translation guidance.
- docs/User_guide_for_MathCAT_en.md, docs/index.md: add Russian (ru) to supported languages lists.
- docs/ru/User_guide_for_MathCAT_ru.md: include Russian in language options, reorder entries and apply minor wording fixes.
- docs/ru/helpers.md: clarify wording, update examples and test instructions to reference ru instead of fr, and improve translation guidance and examples (including test and cargo commands).
- docs/ru/index.md: reorder supported languages and polish phrasing.
These edits enable and document Russian support and improve clarity for translators and developers working on the Russian localization.
* Refactor Russian function braille mapping
Replace hardcoded function-name branches in Russian_Rules.yaml with a DefinitionValue lookup that pulls from a new RussianFunctionBraille mapping. Add RussianFunctionBraille entries to definitions.yaml (private-use codepoints U+E100–U+E10E) and map those codepoints to braille strings in unicode.yaml. Add a test covering tan/cot rendering. This centralizes function braille representations and simplifies the rule logic.
* Addition of Russian documentation
This commit adds Russian documentation for users, developers and translators.
* typo in example
* Add vector/bar/arc mover rules and tests
Introduce three new mover rules in Rules/Braille/Russian/Russian_Rules.yaml: vector-arrow, vector-bar, and geometry-arc. Each rule detects specific accent characters (arrows: → ⟶ ⇀ ⃗; bars: ¯ ‾ _ ￣; arcs: ⌒ ⏜ ︵) and emits the appropriate braille indicators, using a different form when the base has more than one character. Add corresponding tests in tests/braille/Russian/russian.rs to cover single- and multi-character movers and combined expressions.
* Add Russian braille math functions and limits
Extend Russian braille rules to support additional math functions and large operator limits. Added new rules for large-op-limits, operator/functions (including special handling for div), and div-specific replacements in Rules/Braille/Russian/*. Updated Russian definitions and unicode mappings to include sinh/cosh/tanh/coth, arg/sgn/det/rank/rg, Re/Im, grad/rot/div, gcd/lcm and related braille codes. Updated global Rules/definitions.yaml to recognize new function names. Fixed braille cleanup in src/braille.rs to normalize div encoding. Added unit tests exercising the new function mappings and large operator limits.
* Support system left brace in Russian braille
Introduce handling for system left-brace matrices in the Russian braille rules: add InSystemLeftBrace variable, a system-left-brace-opener rule and a system-left-brace-table rule to detect brace-enclosed mtable as a system. Update row-start logic to emit appropriate braille tokens when InSystemLeftBrace is set. Add tests exercising curly-brace matrices, piecewise function (brace with conditions), and vertical-bar matrices to verify the new behavior.
* Add signed-integer and nested-fraction rules
Add handling for nested fractions and signed integers inside script positions and adjust grouping behavior for scripts in Russian braille rules. Changes introduce a nested-fraction rule for mfrac and new signed-integer-number / signed-integer rules for msub/munder and msup/mover, plus a small grouping test (⠐) when scripts contain grouped mrow with '+' or a fraction. Also add a comprehensive regression test (script_grouping_regressions) covering many sup/sub cases to ensure correct braille output. Files updated: Rules/Braille/Russian/Russian_Rules.yaml and tests/braille/Russian/russian.rs.
* Enhance Russian Braille support with mathematical symbols and rules (#589)
* Refine matrix rule handling by adding support for augmented matrices; update unit tests to include 2x3 and 3x4 augmented matrix scenarios. (#555)
* remove logging statement
* removed uneeded eprintlin!()
* Alternative fix to PR #540.
Fixes #528.
Includes the tests in the PR.
* Add Russian translation
* Fix #585 -- SSML was being escaped.
I earlier added code to prevent a SSML injection attack by escaping HTML. It was done at the wrong point and escape legal SSML. I have moved the check and conversion down to the leaves.
I was very surprised there were no useful tests with SSML. All the main tests set `TTS=None`. I added a test that catches bad escaping. I also added some tests to check for leaf-based SSML attacks and also attribute-based ones.
* bump beta number
* The range for MATHEMATICAL SANS-SERIF DIGITs had MATHEMATICAL MONOSPACE DIGITs as the match part of translate, so if one of the sans-serif digits were used, you ended up in a loop (no match, to it returned itself).
* comment out some debug statements
* Add 02DA (˚) to list of circle-like chars -- I thought I had done this already, so it might in some other branch
* Add source language option to translation audit tool
Allow audit-translations to compare a target language against any source
language via --source, defaulting to English for existing workflows. Update
CLI output, docs, and tests for dynamic source/target labels, including
Swedish/Norwegian comparison coverage.
Normalize Rich ANSI output in tests so golden and string assertions remain
stable when terminal color is forced.
* fix ruff
* fix ruff
* fix ruff
* Core concept names additions and tests -- issue 381 (#592)
* initial addition of core concept names
* defnitions and tests, with a few failing
* minus, plus, volume fixes. Other defined, not appearing in site
* fix clippy warnings
* use "ruamel.yaml>=0.19.1" in uv.lock and pyproject.toml
* handle canonicalize failure in containerized environments (#574)
* add 1 week dependency cooldown for uv (#599)
* fix rules for digits in german (#596)
* update navigation rules for traditional chinese (#588)
* copy english version of navigate.yaml from NVDA2026.1 and translate
* 更新 definitions.yaml 與 navigate.yaml
* update unicode-full.yaml
* 區分完全/已完全放大/縮小
* 移至->移到, 表->表格, 數學式->數學, 退出->離開, 不能->無法, 當前->目前。
overview line 47, SimpleSpeak_Rules.yaml line 39, "根號"->""。mroot.rs 對應修改
* 移到上一格 -> 往上一格，移到下一格 -> 往下一格， 佔位符 -> 書籤，設定佔位符 -> 設定書籤，已是最上層 -> 不在別的式子裡面，上一層是 -> 在後面式子裡面，字元模式 開n 次方根 的讀法，基底 -> 基、底
* 向量 -> 射線
* 基 -> 基底，底 -> 基底
* Internally in definitions.yaml, MathCAT uses "nofix", but that's not
a MathML airity property. Remove it.
* don't say "indexed by", say "sub"
* add some terse and verbose forms of speech
* fix typo
* Added an html file with all the test exprs so that translators have examples to test.
* Fix remaining erros after `definitions.yaml` was updated -- needed to add `function:`
* fix clippy warning
* minor cleanup: remove ASCIIMath-fi reference
* Refactor braille.rs to a BrailleCode trait + registry
Replace the four scattered string-match dispatc…
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
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.
Follows up on #715.
I need to correct what I said there about the MathPlayer unicode-full diff. When I sorted it by the comments, about 1,780 of its Japanese entries were marked as DeepL translations and only about 120 carried MathPlayer's own Japanese, and the DeepL ones had errors of their own (₈ came out as "subscript at", ㉘ as "20 and at"). So rather than taking it as the base, I went through all 2,754 entries, compared each against the English, the current file and the MathPlayer/DeepL text, and wrote the readings with one glossary: Yamaguchi et al. (1996) for relations, brackets and arrows, and the usual Japanese names for shapes. The readings were drafted and cross-checked with Claude rather than checked by a person one by one, so they stay
t:. The structure and comments of the file are unchanged. I left out MathPlayer's 256 braille pattern entries because en/unicode-full.yaml doesn't have them.For unicode.yaml I kept the current file and took カッコ / カッコ閉じ and the 向き arrow names, both from Yamaguchi, plus プラス オア マイナス. × in the "By" setting now reads 掛ける, as the matrix dimension already does, and ∦ reads ノット 平行 like ノット イコール. ∑ is サム (Yamaguchi), matching プロダクト for ∏. definitions.yaml and the menclose arrows were changed to match.
audit-translations ja: untranslated text 3377 → 3353, which is exactly the lines I promoted toT:. Rule differences stay at 30, and there are no missing or extra rules. Nine test expectations changed to the new words, and one new test covers characters that only unicode-full.yaml knows (¥, ∅, ⊕, ∋, ≺). The bracket tests now have a comma after the opening bracket: the auto pause incompute_auto_pauseis based on the byte length of the neighbouring text, and カッコ is three bytes longer than 括弧.🤖 Generated with Claude Code