Benchmark · Directional
Localisation extraction
Glossa is Webfold's localisation extractor. It extracts strings about 4.8 times faster per file. The baseline only parses: the Babel parser and Vue compiler. The Glossa project ran this itself. It is not independent.
Glossa's string extraction against the JavaScript parsers that localisation tools are built on.
What question does this benchmark ask?
How fast does Glossa extract translatable strings? The baseline is the JavaScript parsers that localisation tools build on.
How was it run?
- Extraction baseline: real runs of the Babel parser and the Vue SFC compiler.
What are the results?
Extraction time per file, microsecondsShorter is better
Glossa
17.6 µs
Babel + Vue compiler
84.4 µs
| Measurement | Glossa | Baseline |
|---|---|---|
| Extraction per file | 17.58 µs | 84.42 µs |
| Lockfile merge per coordinate | 0.307 µs | 3.057 µs |
| Integrity check per term | 2.36 µs | 4.242 µs |
| Catalog emission per unit | 0.349 µs | 4.385 µs |
Glossa extracts strings about 4.8 times faster per file than the Babel parser and Vue compiler take just to parse the same file.
What does this not show?
Know the limits
- Measured by the Glossa project against its own harness. Not independent.
- We removed the memory comparison from this page. The project's 8.3 MB figure was the current resident size read inside a Rust process, and its 375 MB baseline was the peak of
npx lingo.dev --version, which is not an extraction job. Neither is a like-for-like workload, so a headline built on them was not fair. - The speed comparison is also not against the incumbents' command-line tools: Glossa does the whole extraction (scan, keys, snippets) while the baseline only parses, and the incumbents' own figures are hard-coded constants in the project's harness. A like-for-like run of Glossa against FormatJS, Lingui and i18next-parser on a fixed corpus of real open-source projects, wall time and peak memory, is planned.
- An earlier version of the project's README claimed a much larger speed-up based on an assumed baseline. It was corrected to the measured figures above.
Which other benchmarks are there?
- HTML to Markdown conversion Measured
- Document conversion Measured
- Main-content extraction quality Measured
- Web archive extraction Measured
- Footprint of a scrape or crawl Directional
- PII detection against Presidio Measured
- PDF text extraction with OCR Directional
- PDF extraction against other extractors Directional
- LaTeX conversion with BibTeX Directional
- A whole site to Markdown Directional
- Agent-browser compatibility (Lexbench) Measured
- Headless engine over CDP Directional
- Search-results parsing Directional
- Folder search Directional
Check it on your own pages.
Run the command-line tool on your content.