home.social

#mashumaro — Public Fediverse posts

Live and recent posts from across the Fediverse tagged #mashumaro, aggregated by home.social.

fetched live
  1. Just published #mashumaro 3.15 with the #JSONSchema plugin system! 🎉 I guess the next thing will be plugins for serialization and deserialization. Who knows, maybe this will be my ticket to strict validation and speeding up with #cython. On a side note, I believe I’m among the last developers to discontinue support for #python 3.8.

    github.com/Fatal1ty/mashumaro/

  2. Well, adding support for #python 3.13 in #mashumaro is less painful this time. No more edge cases like last year.

  3. After two months of coding in free time, I merged the long-awaited pull request that opens up the door to new ways of using #mashumaro. Previously, one could only work with dataclasses that inherit a specific mixin, e.g. DataClassJSONMixin. It was a stop factor for many and I was getting requests to support top-level array serialization. This requirement is now history, as anyone can construct a decoder and encoder for any type, even with no dataclasses at all.

    #python

    github.com/Fatal1ty/mashumaro#

  4. Just released a new version of #mashumaro with Python 3.12 support and new features. One of its optional dependencies “pendulum” hasn’t been released with 3.12 support, but it’s optional anyway.

    github.com/Fatal1ty/mashumaro

    #python #python312

  5. Well, not everything went smoothly with #python 3.12 support for #mashumaro. Some tests failed and I started investigating. Turned out, I came across an edge case with comprehension iteration variables isolation which apparently was introduced by comprehension inlining. Fortunately, there is still time before the final release of Python 3.12. Looks like I just unlocked the release blocker achievement 🎉

    github.com/python/cpython/issu

  6. I was curious why there were so many #mashumaro downloads on #python 3.7. This investigation by @hugovk might me the answer mastodon.social/@hugovk/110934. The post caught my eye literally after I committed TODO “switch to cached_property after dropping python 3.7 support”.

  7. @agturcz I had the same question during the development of #mashumaro when I needed to decide how to deserialize user data into typed sequences. It was critical for me to squeeze out the maximum performance and different benchmarks showed that list comprehensions just work way faster. So, it’s clear to me who to give preference to.

  8. Just finished writing a pyperf based benchmark for #mashumaro. The most difficult part was to create some kind of orchestration in #python which turned out to be a very bad idea. I ended up giving on beauty and just wrote a bash script.

    Interesting part. Rust gave a huge boost to #pydantic but mashumaro is still somewhat faster and not to be written off.

    github.com/Fatal1ty/mashumaro#