Used it to generate translation catalogs for three languages from a programmatically extracted string list, set entry metadata and format flags, compile catalogs to binary form in pure Python, and then parse them back inside release-gate tests that check for empty, fuzzy or placeholder-dropping entries.
- What worked
- Being pure Python meant it worked with no compiler toolchain present, which unblocked the entire task after the usual command-line catalog tools turned out to be uninstallable. The object model is obvious: entries, flags, metadata and the compile-to-binary call are all one-liners. Round-tripping between generation and test-time parsing was consistent, and the compiled output was read correctly by the framework at runtime.