Crate pockingbird 

pockingbird

Report-only library + CLI that finds duplicate translation keys across gettext .po catalogs.

Library entry point

The one call you need is scan — point it at a directory and a Config and get back a Report:

use std::path::Path;
use pockingbird::{scan, Config};

let report = scan(Path::new("./locales"), &Config::default())?;
println!("{}", pockingbird::report::to_json(&report.groups, report.total_keys));

Pipeline

walk → po → index → group → report

• walk — discover .po files under the configured roots.
• po — parse each catalog (polib) into keys and per-locale values; locale derives the locale id from the path.
• index — build the KeyId × locale → Cell matrix; normalize and fuzzy canonicalize cells per tier.
• group — tier-agnostic signature bucketing + leave-one-out over the canonical matrix → CandidateGroups.
• report — render the groups as text (colored) or json.

These stages are crate-internal; scan wires them together. config holds the TOML schema and defaults that parameterize every stage.

Re-exports

pub use config::Config;

Modules

cli

CLI surface (clap derive): find <path> [knobs] and init [path]. Behind the cli feature. Exit code is always 0 (the tool reports, it does not gate). This is the thin shell around the crate::scan facade: it resolves config, renders progress to stderr, and writes the report to stdout — the pipeline run itself is the testable core.

config

TOML configuration schema (Config/Scan/Locales/Match/Normalize/ Output) plus defaults. Parameterizes every pipeline stage.

report

Render candidate groups as text (colored) or json, with identical structure.

Structs

CandidateGroup

A reported duplicate-candidate group over one tier.

KeyId

Identity of a translatable key. domain is part of the identity, so messages.po:X and django.po:X are distinct keys. Hashable and ordered for stable, deterministic bucketing.

Report

The run’s data result. No timings, no formatting — rendering is the shell’s job.

Skip

A catalog that failed to parse. Recorded, never fatal.

Enums

Cell

A key’s value in one locale.

Error

A failed scan: discovery, no input, or a pipeline error.

PipelineError

A run that could not produce a report at all (distinct from a skipped file).

PipelineEvent

A progress event emitted during a run. The vocabulary is here; the rendering is the sink’s.

PoError

WalkError

Traits

ProgressSink

The one-method progress seam. Implement it to render, time, or record.

Functions

scan

Scan a directory tree for duplicate translation keys across its .po catalogs — the one-call library entry point. Discovers files under root (overriding config.scan.roots), parses them, and runs the pipeline.

scan_with

Like scan, but drives a ProgressSink so callers can render progress.