starlite VS api-performance-tests

For the past one and a half years 😁. Checkout our docs https://starliteproject.dev/

#1286 is the issue tracking the removal of our last internal Pydantic datastructure, and PR #1382 has the addition of attrs based signature modelling. Once those are merged you will be able to use Starlite without Pydantic.

You can go to Open an Issue and open an issue for a feature request/enhancement and describe basically the behavior you are wanting to see.

Falcon FastAPI Sanic Starlite (disclosure: I do work here)

for server side , starliteproject.dev/, only python async framwork worth looking into.

Well, I'm not objective but I'd say Starlite (https://github.com/starlite-api/starlite) is a might fine codebase. You can also learn a lot about tooling and typing going through it.

I have mentioned your concerns to the Starlite discussions board - https://github.com/starlite-api/starlite/discussions/1183

You can find the complete changelog on our v1.51.0 GitHub release page.

Since version v1.39.0 Starlite (released on the 12.11.22) no longer has any dependency on Starlette. Between this version and v1.45.0 that was released today (11.12.22), we have invested significant effort into benchmarking and optimizing code. One of our maintainers, @provinzkraut (Janek Nouvertné), has done amazing work rewriting our benchmarking framework. You can read more about this in here and run the benchmarks on your own by cloning the benchmark repository. The results are pretty impressive if I may say so myself:

Another important point I want to address is the framework's performance. The latest iteration of the TechPowered Benchmarks came out and they really surprised us. Starlite appeared rather slow there, and FastAPI was almost 200% faster thatn Starlette, which isn't technically possible given the structure of this library. Obviously something weird is happening there (no clue what), but this compelled us to invest some more efforts into our own benchmarking, the repository for which you can see here.

here: https://github.com/Goldziher/api-performance-tests

I used autocannon for this - which is a very cool benchmarking tool (I am not involved with it in any way). You can see the tests and more details in this repo. In a nutshell - I used identical setup for all 3 frameworks. Since both FastAPI and Starlite use Starlette as a basis, I also compared with Starlette.

I looked but still don’t understand… https://github.com/Goldziher/api-performance-tests/blob/main/results/json-starlite-sync-json-abc-1.json has 32k requests (per second?) but your graphs talk about 107

Legend: - a-: async, s-: sync - np: no params, pp: path param, qp: query param, mp: mixed params **Note*\*: only Starlite uses orjson by default - using their default settings both Starlette and FastAPI are significantly slower in serializing/deserializing JSON. You can find the tests in this repo, which also documents the test setup etc. As always, I invite you to checkout Starlite and maybe ⭐our repo. Starlite is a collaborative project - we have a discord server, which you can find here. You can find my previous reddit posts related to Starlite in these links: 1. Post regarding the builtin support for SQL Alchemy and the plugin system 2. Post introducing Starlite You can also read the article in medium (no pay gate) introducing Starlite.

You can find the tests in this repo, which also documents the test setup etc.