Skip to content

v2.3.0 — Linter-friendly file writes

Choose a tag to compare

View all tags
@BrunosCode BrunosCode released this 21 May 12:52
· 6 commits to main since this release
v2.3.0
755f52c

Added

  • Linter-friendly writes — File handlers (PHP, JSON, CSV) now compare the new content against what's already on disk before writing. If the result is identical, including key order, the write is skipped entirely. A file already formatted by the user's linter (Pint, PHP-CS-Fixer, Prettier, …) is left untouched as long as its translations and ordering match the new state.
  • Concerns\ComparesTranslations trait — strict-equality check (via json_encode) plus a leaf-level diff counter, used by all three file handlers.
  • Idempotency tests — every file handler now has a returns 0 when content is unchanged regression test.

Changed

  • Return semantics of put() / set() — the integer returned now represents the number of translations actually changed (added, modified, or removed), not the total count in the input collection. A no-op write returns 0.
  • sync() / import() / export() return type — changed from bool to false|int. false signals an actual failure (the source has no translations to read); any integer (including 0) signals success and reports how many translations were changed in the destination.
  • SyncCommand, ImportCommand, ExportCommand, and the deprecated bare Command — after the main success message, the command now prints either Already in sync. or N translation(s) changed..
  • SetCommand — when set() returns 0, the command checks whether the current value already matches the requested one. If it does, the command prints Translation already set! and returns SUCCESS (previously this case was reported as FAILURE).
  • MCP tools — set-translation-tool now also exposes the integer count field next to written; sync-translations-tool now exposes success (bool) and count (int).

Why

The previous behavior rewrote every file on every operation, even when nothing had changed. Combined with code formatters that normalize array syntax / indentation / quoting, this produced spurious diffs in git on every sync and forced the formatter to run again afterwards. Comparing parsed content (not raw bytes) makes the operation a true no-op when the resulting state matches.

Full Changelog: v2.2.0...v2.3.0

Assets 2
Loading