/japronto

Screaming-fast Python 3.5+ HTTP toolkit integrated with pipelining HTTP server based on uvloop and picohttpparser.

Primary LanguageCMIT LicenseMIT

Japronto!

irc: #japronto Gitter japronto/Lobby Build Status PyPI PyPI version

Japronto (from Portuguese "já pronto" /ˈʒa pɾõtu/ meaning "already done") is a screaming-fast, scalable, asynchronous Python 3.5+ HTTP toolkit integrated with pipelining HTTP server based on uvloop and picohttpparser. It's targeted at speed enthusiasts, people who like plumbing and early adopters.

You can read more in the release announcement on medium

Performance

Here's a chart to help you imagine what kind of things you can do with Japronto:

Requests per second

As user @heppu points out Go’s stdlib HTTP server can be 12% faster than the graph shows when written more carefully. Also there is the awesome fasthttp server for Go that apparently is only 18% slower than Japronto in this particular benchmark. Awesome! For details see squeaky-pl#12 and squeaky-pl#14.

These results of a simple "Hello world" application were obtained on AWS c4.2xlarge instance. To be fair all the contestants (including Go) were running single worker process. Servers were load tested using wrk with 1 thread, 100 connections and 24 simultaneous (pipelined) requests per connection (cumulative parallelism of 2400 requests).

The source code for the benchmark can be found in benchmarks directory.

The server is written in hand tweaked C trying to take advantage of modern CPUs. It relies on picohttpparser for header & chunked-encoding parsing while uvloop provides asynchronous I/O. It also tries to save up on system calls by combining writes together when possible.

Early preview

This is an early preview with alpha quality implementation. APIs are provisional meaning that they will change between versions and more testing is needed. Don't use it for anything serious for now and definitely don't use it in production. Please try it though and report back feedback. If you are shopping for your next project's framework I would recommend Sanic.

At the moment the work is focused on CPython but I have PyPy on my radar, though I am not gonna look into it until PyPy reaches 3.5 compatibility somewhere later this year and most known JIT regressions are removed.

Hello world

Here is how a simple web application looks like in Japronto:

from japronto import Application


def hello(request):
    return request.Response(text='Hello world!')


app = Application()
app.router.add_route('/', hello)
app.run(debug=True)

Tutorial

  1. Getting started
  2. Asynchronous handlers
  3. Router
  4. Request object
  5. Response object
  6. Handling exceptions
  7. Extending request

Features

  • HTTP 1.x implementation with support for chunked uploads
  • Full support for HTTP pipelining
  • Keep-alive connections with configurable reaper
  • Support for synchronous and asynchronous views
  • Master-multiworker model based on forking
  • Support for code reloading on changes
  • Simple routing

License

This software is distributed under MIT License. This is a very permissive license that let's you use this software for any commercial and non-commercial work. Full text of the license is included in LICENSE.txt file.

The source distribution of this software includes a copy of picohttpparser which is distributed under MIT license as well.