@dittowords/text-extract

Finds strings in a codebase that are potentially user-facing.

Usage

A whole directory:

import { runExtract } from "@dittowords/text-extract";

const { candidates, summary } = await runExtract({ inputPath: "/path/to/repo" });

Skips vendored, build and test trees, and honors .gitignore.

One file:

import { extractFile } from "@dittowords/text-extract";

const { candidates, skipped } = await extractFile({
  relPath: "src/Button.tsx", // used verbatim as `location.file`
  source: fileContents,
  framework: ["react", "next"], // optional
});

Scans whatever path it's given — no exclusions are applied. framework is the same token list runExtract derives from package.json files and native project markers; omitting it only costs a classification hint.

skipped is set when the file wasn't scanned at all (unsupported_language, unconfirmed_i18n_file, minified) — not the same as scanning it and finding nothing.

Both entry points reach the same verdict on the same file.

Notes

• Candidate order may not be stable between runs, because globby's file iteration order varies. The set is stable; compare accordingly.
• summary.filesFailed counts files an extractor threw on. They're dropped from the results, so a non-zero count means strings are missing.
• summary.linesStoppedEarly counts lines that were skipped because they are longer than MAX_LINE_LENGTH or slower than LINE_TIME_LIMIT_MS to tokenize. A skipped line has no candidates. summary.stoppedEarlyLines lists up to 20 skipped lines with their files. extractFile returns all skipped lines of its file in stoppedEarlyLines.

Development

yarn test
yarn typecheck
yarn build