Improve decoder performance on SQL call #281
Conversation
| # Create Decoder once and reuse it for all encoding calls, | ||
| # to avoid paying setup cost for every call. More info: | ||
| # https://jcristharif.com/msgspec/perf-tips.html#reuse-encoders-decoders | ||
| _decoder = msgspec.json.Decoder() |
There was a problem hiding this comment.
Hi - msgspec author here.
The main benefit of creating and reusing a decoder is avoiding processing extra config kwargs on every call to decode, especially the type kwarg. Since you're not passing in any keyword arguments here, using the msgspec.json.decode convenience method should be comparable in performance. No harm in the change you've made here, but also likely no measurable benefit. I'll update our docs page to better clarify when you may want to preallocate a decoder.
There was a problem hiding this comment.
Thanks for clarifying!
The encoder example and the phrase "The same goes for decoding" disoriented me. I understood this in such a way that resources are spent on creating the object itself, so in my projects I created both an encoder and a decoder - then reuse both. And where I know the model in advance, I create an additional decoder.
P.S.: thanks for great product!
Pull Request Checklist
Description
orjsontip withmsgspecmsgspecdecode performanceClose Issue(s)