URL parsing is broken

[wolever]

With Python 2.5 and 2.6, url parsing is broken because urlparse doesn't know how to interpret the ws scheme. Here's a fixed version that works correctly:

def _parse_url(url):
    """
    parse url and the result is tuple of 
    (hostname, port, resource path and the flag of secure mode)
    """
    if ":" not in url:
        raise ValueError("url is invalid")

    scheme, url = url.split(":", 1)
    url = url.rstrip("/")

    parsed = urlparse(url, scheme="http")
    if parsed.hostname:
        hostname = parsed.hostname
    else:
        raise ValueError("hostname is invalid")
    port = 0
    if parsed.port:
        port = parsed.port

    is_secure = False
    if scheme == "ws":
        if not port:
            port = 80
    elif scheme == "wss":
        is_secure = True
        if not port:
            port  = 443
    else:
        raise ValueError("scheme %s is invalid" % scheme)

    if parsed.path:
        resource = parsed.path
    else:
        resource = "/"

    return (hostname, port, resource, is_secure)

[liris]

It sounds strange. My python 2.6.1 and 2.5.4 on Mac woks fine.
urlparse method doesn't care of scheme type.

Please tell me the traceback or exception detail.

[wolever]

Wow, that is strange. Neither the 2.6.1 or 2.5.4 on my Mac work.

To recreate:

$ python --version
Python 2.6.1
$ cd /tmp/liris-websocket-client-673470f/
$ cp websocket.py examples/
$ cd examples/
$ python echo_client.py 
/private/tmp/liris-websocket-client-673470f/examples/websocket.py:27: DeprecationWarning: the md5 module is deprecated; use hashlib instead
  import md5
Traceback (most recent call last):
  File "echo_client.py", line 5, in 
    ws = websocket.create_connection("ws://localhost:5000/chat")
  File "/private/tmp/liris-websocket-client-673470f/examples/websocket.py", line 110, in create_connection
    websock.connect(url, **options)
  File "/private/tmp/liris-websocket-client-673470f/examples/websocket.py", line 205, in connect
    hostname, port, resource, is_secure = _parse_url(url)
  File "/private/tmp/liris-websocket-client-673470f/examples/websocket.py", line 76, in _parse_url
    raise ValueError("hostname is invalid")
ValueError: hostname is invalid

The behaviour with 2.5 is identical. The distribution is from https://github.com/liris/websocket-client/zipball/v0.4.1 .

[liris]

Oh, I see the problem. The parsed hostname is broken.

[brandonmoser]

This issue was reintroduced in a recent version. Changing the Scheme back to HTTP when parsing the URL allows the library to work on 2.6.1 (I haven't tried 2.5.x).