Various improvements to Protocol::WebSocket::Client

[greg-kennedy]

This request wraps up several changes to Protocol::WebSocket::Client to make it more friendly to end users.

• Fix missing on_eof callback
• Additional callbacks for ping and pong events, plus functions to send these messages
• A default pong reply to ping for keepalive
• Comments and misc. cleanup
• Documentation!

All changes are supposed to be backwards compatible with previous users, but they add some convenience methods and safeguards against improper use. I have tested this with a custom driver program and it seems to work pretty well. However, I know there are dependencies (e.g. https://github.com/plicease/AnyEvent-WebSocket-Client) that rely on this module, and I haven't checked them thoroughly for issues.

[vti]

Thanks Greg, those are very useful changes. A couple of suggestions if I may?

• squash those commits that fix each other (like the ready-related stuff);
• use 4 spaces instead of 2;
• do not use a void map.

Thanks :)

[greg-kennedy]

Well, don't mind the mess there... I was able to reduce it to five commits and fix those issues. I think it's ready to go : )

[vti]

Well...

t/client.t ........................... 1/? Protocol::WebSocket::Client: write() on unconnected WebSocket. at /home/vti/dev/protocol-websocket/lib/Protocol/WebSocket/Client.pm line 139.

    #   Failed test at t/client.t line 70.
    #          got: ''
    #     expected: '�'
    # Looks like you failed 1 test of 1.

#   Failed test 'write close frame on disconnect'
#   at t/client.t line 71.
Protocol::WebSocket::Client: write() on unconnected WebSocket. at /home/vti/dev/protocol-websocket/lib/Protocol/WebSocket/Client.pm line 139.

    #   Failed test at t/client.t line 81.
    #          got: ''
    #     expected: anything else
    # Looks like you failed 1 test of 1.

#   Failed test 'call on_write on write'
#   at t/client.t line 82.
# Looks like you failed 2 tests of 7.

[vti]

I guess they are failing because it now checks if it's connected or not.

[vti]

On disconnect the close frame is sent as masked, are you sure about that?

[vti]

Found this in the spec Close frames sent from client to server must be masked as per Section 5.3.

[greg-kennedy]

Yeah, I guess the tests are wrong, because it seems they are trying to do things (write, disconnect, etc) without an actual session in place.

I guess you could modify the test to futz with $client->{state} and set it to 1 before you do something else with it.

That said I will modify client.t and see where we can get.

[vti]

I already fixed the tests, just added a connect. I guess this is the correct way anyway. Thanks!