review
Per-locale summary: string path counts and English-identical counts vs the source locale.
i18nprune review
i18nprune review --target ja
i18nprune --json review--json emits a structured localeReview object on stdout.
Planned behaviour (machine + human)
The review command will grow toward a full locale metadata report: structured leaves (value, status, confidence, needsReview, source, updated-at style fields) when your locale JSON carries that metadata. It does not replace quality / validate for en.json parity — those stay separate concerns.
Human mode
- Print an applied options line block: effective
--target, filters,localesDir, scope — respect-q/-s(no duplicate noise).
Output routing (planned)
--csv→ CSV on stdout (exclusive with JSON).--json:- With
--format by-source→ grouped JSON by source. - With
--format by-status→ grouped JSON by status. - Otherwise → flat JSON.
- With
--format summary|by-status|by-sourcewithout--json/--csv→ human layout.
--json and --csv are mutually exclusive (error if both).
Human format variants (planned)
summary— compact stats per locale (default).by-status/by-source— path lists with caps.
List caps (planned)
--top N— parsed today as a positive integer (same error rules asmissing --top). Until per-path human lists ship, this flag is reserved but validated so scripts fail fast on typos (review: --top must be a positive integer).--full— reserved alongside--topfor unbounded path lists when that output exists.- Default sample when neither: a small fixed cap (e.g. 10 rows) for human lists, flat JSON, and CSV slices — unless
--fullor--topoverrides.
Filters and flags (planned)
| Flag | Role |
|---|---|
--format (summary / by-status / by-source) | Human layout; use --json / --csv for machine sinks (not format=json on the CLI). |
--json | Machine JSON on stdout (variant depends on --format). |
--csv (optional alias for typos) | Machine CSV on stdout. |
--all | All non-source locales (default scope where applicable). |
--target, multi-lang argv (planned) | Restrict which locale files. |
--full | Unbounded lists (subject to --top). |
--top N | Cap paths per section or flat rows. |
--needs-review | Filter rows that need review. |
--status (pending / translated) | Filter by status. |
--source <name> | e.g. provider or provenance label. |
--min-confidence / --max-confidence | Numeric filters. |
Current vs planned
- Today: per-locale string path counts + English-identical counts vs source;
review --top/review --fullare accepted and validated but do not change output until path lists are implemented. - Next: format switches, filters, CSV, applied-options banner, and richer stats when locale JSON includes review metadata.
Related
- JSON mode and long commands —
--json, prompts, progress - Roadmap —
--report-fileand review sequencing