Add fast-check property suite; fix two escaping bugs it found - #113
Merged
Conversation
A translation containing a Unicode line or paragraph separator was written raw into the double-quoted value. The entry-line scan on the following sync then failed to match the line (JavaScript's dot excludes line terminators), classified it as unrecognized, and re-appended the key as a duplicate — so the sync after that could not parse the file at all. Literal-string values now also fall back to double quotes when a translation gains one of these characters. Found by the fast-check property suite; closes cli-0jkg. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
The value parser strips unescaped whitespace after the separator, so a translation beginning with a space silently lost it on the next read. The writer now escapes each leading space; leading tabs, trailing spaces, and newlines were already handled. Found by the fast-check property suite; closes cli-ekmg. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
…on utils Round-trip laws parameterized over all 11 format parsers: translated values survive reconstruct/extract, re-applying the same translations is a no-op, and identity syncs are fixed points. PO and XLIFF, whose extract reads the source slot while reconstruct writes the target, get idempotence and source-preservation laws instead. Preservation utilities and ICU parsing get identity and segment-stability laws. 200 seeded-random cases per law by default; FC_NUM_RUNS deepens a run, FC_SEED/FC_PATH replay a recorded counterexample. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
…nd validator translateBatch: chunks never exceed TRANSLATE_BATCH_SIZE, unique texts are each sent once, every input position (duplicates included) maps to its own translation, and the call count matches the chunk arithmetic. Message preprocess: expandPlurals/detectIcu/writebackPlurals preserve the positional invariants the pipeline documents — original texts stay an unchanged prefix, each plural slot points at its own appended value, and writeback touches exactly the slots that have results. Validator: total on arbitrary input, identity is never a structural error, and dropped or added placeholders are always flagged. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
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.
Summary
Adds a fast-check property-based test suite (
tests/property/) enforcing round-trip laws across all 11 format parsers plus the preservation, ICU, batch-splitting, message-preprocess, and validator layers — and fixes the two real bugs the suite found during evaluation: TOML file corruption from unescaped U+2028/U+2029 and silent leading-space loss in.properties.Changes Made
d0dfd35) — a translation containing a Unicode line/paragraph separator round-tripped fine on the first sync, silently appended duplicate keys on the second (JS.in the entry-line regex excludes line terminators, so the written line was re-classified as new), and made the third sync throw on parse.encodeTomlStringnow escapes both characters; literal-string values gaining one fall back to double quotes..propertiesleading-space loss (158486d) — the writer emitted leading spaces raw and the value parser strips unescaped whitespace after the separator, so a translation beginning with a space lost it on every subsequent read. The writer now escapes each leading space (\); the reader already handled the escape.tests/property/(64deabd) — parameterized laws over every parser in the registry: translated values survive reconstruct/extract intact (L1), re-applying the same translations is byte-identical (L2), and identity syncs are fixed points (L3). PO and XLIFF, whoseextractreads the source slot whilereconstructwrites the target, get idempotence + source-preservation laws instead. Plus identity/multiset laws for placeholder preservation, segment-stability laws for the ICU parser, and a directed pin of the current (un-normalized) cache-key behavior.5081f77) —translateBatchchunking (≤ 50 texts per call), dedup, duplicate-position index mapping, and call-count arithmetic against a fake client; positional invariants for theexpandPlurals → detectIcu → writebackPluralspipeline; validator totality and guaranteed detection of dropped/added placeholders.fast-checkas a devDependency — the only new dependency; properties run via plainfc.assertinsideit(), no jest integration layer.Test Coverage
tests/property/(24 fc properties × 200 seeded-random cases each ≈ 5,000 generated law checks per run) and 3 regression tests intests/unit/formats-roundtrip.test.tsshipped with the fixes.FC_NUM_RUNSdeepens a run (2000 ≈ 7s);FC_SEED/FC_PATHreplay a recorded counterexample exactly.35s (+1.6s vs. before this branch).Backward Compatibility
✅ Maintained: no API, CLI, or config surface changes.
✅ Same behavior: TOML/properties output is unchanged for all values that do not contain U+2028/U+2029 or leading spaces; for values that do, output changes from silently corrupting/lossy to correct (files written by the buggy path are read back correctly).
❌ Breaking changes: None.
Size: Medium
Two one-function fixes plus a self-contained new test directory; no production code paths restructured.
🤖 Generated with Claude Code