Small, sharp, auditable HTTP parser.

[lovelydinosaur]

Parsers

Here's what'll be incoming here...

HTTPClientParser

writer = io.BytesIO()
reader = io.BytesIO(
    b"HTTP/1.1 200 OK\r\n"
    b"Content-Length: 23\r\n"
    b"Content-Type: application/json\r\n"
    b"\r\n"
    b'{"msg": "hello, world"}'
)
p = httpx.HTTPClientParser(writer, reader)

# Send the request...
p.send_method_line(b"GET", b"/", b"HTTP/1.1")
p.send_headers([(b"Host", b"example.com")])
p.send_body(b'')

# Receive the response...
protocol, code, reason_phase = p.recv_status_line()
headers = p.recv_headers()
body = b''
while buffer := p.recv_body():
    body += buffer

HTTPServerParser

reader = io.BytesIO(
    b"GET / HTTP/1.1 200\r\n"
    b"Host: example.com\r\n"
    b"Accept: */*\r\n"
    b"\r\n"
)
writer = io.BytesIO()
p = httpx.HTTPServerParser(reader, writer)

# Receive the request...
method, target, protocol = p.recv_method_line()
headers = p.recv_headers()
body = b''
while buffer := p.recv_body():
    body += buffer

# Send the response...
p.send_status_line(b'HTTP/1.1', b'200', b'OK')
p.send_headers([(b'Content-Length', b'23'), (b'Content-Type', b'application/json')])
p.send_body(b'{"msg": "hello, world"}')
p.send_body(b'')

MultiPartParser

p = MultiPartParser(stream, boundary)
p.recv_start()
while not p.recv_complete():
     headers = p.recv_headers()
     body = b''
     while buffer := p.recv_body():
         body += buffer

[lovelydinosaur]

IDK how many folks will appreciate this, tho it's looking really neat. This swaps out h11 for a smart little HTTP parser with a similar design.

>>> import httpx
>>> import io
>>>
>>> writer = io.BytesIO()
>>> reader = io.BytesIO(
...     b"HTTP/1.1 200 OK\r\n"
...     b"Content-Length: 23\r\n"
...     b"Content-Type: application/json\r\n"
...     b"\r\n"
...     b'{"msg": "hello, world"}'
... )

>>> # Create ourselves a lil parser...
>>> p = httpx.HTTPParser(writer, reader)
>>> p
<HTTPParser [client IDLE, server IDLE]>

>>> # Send a request...
>>> p.send_method_line(b'GET', b'/', b'HTTP/1.1')
>>> p.send_headers([(b"Host", b"example.com")])
>>> p.send_body(b'')

>>> # Check the current state and output
>>> p
<HTTPParser [client DONE, server IDLE]>
>>> writer.getvalue()
b'GET / HTTP/1.1\r\nHost: example.com\r\n\r\n'

>>> # Read ourselves a lil response
>>> p.recv_status_line()
(b'HTTP/1.1', b'200', b'OK')
>>> p.recv_headers()
[(b'Content-Length', b'23'), (b'Content-Type', b'application/json')]
>>> p.recv_body()
b'{"msg": "hello, world"}'
>>> p.recv_body()
b''

>>> # Completed request/response cycle
>>> p
<HTTPParser [client DONE, server DONE]>
>>> p.start_next_cycle()
>>> p
<HTTPParser [client IDLE, server IDLE]>

We can doc out an example integrating this with https://www.encode.io/httpnext/networking
This gets us to zero dependencies, with very limited stdlib imports.

Now... I need to get things out into the public domain.

[lovelydinosaur]

Resolved in #122

[carltongibson]

This is cool. BTW.