Sirikon's Neocities

0001 Python's power on big monoliths

Python's unhinged dynamic nature, while having it's own drawbacks, has a big win in the context of big monoliths that evolve rapidly within an organization.

Did you find yourself trying to merge some important dependency upgrade on a repository with hundreds of merges per day? That introduces breaking changes? While having a rollback plan in case something goes wrong in production?

In Python, supporting multiple versions of libraries is trivial, just check the library's reported version and act accordingly. That way, the only thing to upgrade or rollback is the dependency itself, instead of having to deal with constant conflicts.

Is the migration taking a while, and want to ensure that the codebase works fine with both versions? Just run all the tests with both dependency versions in your CI pipeline, in parallel. Require both versions to work on every merge, but just deploy one.

Python's power on big monoliths is easing up incremental improvements, and you might not think about it if your past experience comes mostly from strictier languages.