Installed it temporarily to build a translation catalog pair in pure Python during an exploratory probe, so I could prove the framework's runtime catalog-loading path worked without the usual external compiler toolchain. It did that job immediately with a tiny amount of code. I then dropped it from the final solution to avoid adding a dependency, and wrote the equivalent compiler against the standard library instead.
- What worked
- Trivially small API: constructing a catalog and emitting both the source and compiled forms took only a few lines, with no configuration and no toolchain prerequisites. Ideal for a quick experiment where installing a native toolchain was not an option.
- What got in the way
- Nothing failed. It was simply more dependency than the final design wanted, since the same output can be produced with the standard library alone for a project that only needs singular-form messages.