Closing the websocket

[afcondon]

A very small issue: the example in the repo closes the socket with two parameters, both given as Nothing (and interpreted in the WebSocket.js closeImpl method as mCode and mReason respectively.

But this will always fail to dereference and in any case it seems to me that the spec [1] doesn't actually take any parameters, tho' perhaps an antecedent spec did, i don't know.

[1] http://www.w3.org/TR/2009/WD-websockets-20091222/#websocket

[zudov]

Hi.

the spec [1] doesn't actually take any parameters

A more modern version of specification seem to say

void close([Clamp] optional unsigned short code, optional USVString reason);

with a more detailed description of close method here.

tho' perhaps an antecedent spec did

Yes, 2009 might be a bit too old.

But this will always fail to dereference

I'll take a look at this when I wake up, I remember testing the example, but probably I didn't notice that, and adding some automated tests which make sure that it works (e.g. in phantomjs) wouldn't hurt anyway. I'll also try to lookup how far does the browser support goes for this parameters, probably it's actually quite new, hence the failures.

It seems that closeImpl passes undefined to close() when it gets Nothing, not sure if it might cause problems (I wouldn't be too surprised if it can).

[zudov]

Depending on internal purescript details (value0) isn't good anyway, so it certainly has to be patched.

[afcondon]

Hahaha, okay that's embarassing that i didn't notice that my cursory Google search was to an ancient version of the working document. :-)

thanks for following up!

[zudov]

In order to properly fix .value0 thingy we need to first resolve https://github.com/zudov/purescript-undefinable. I made that library, but there were some ideas on IRC about adding it to purescript-nullable or integrating with purescript-foreign, so I am not publishing it before we decide what to do.

[zudov]

As of original problem, I can't really reproduce the "always fail to dereference" thingy.

When I run the example, it executes the socket.close Nothing Nothing and after some time receives the CloseEvent with status code 1005, which "Indicates that no status code was provided even though one was expected.".

[afcondon]

Maybe it's browser specific? Here's what i see running example in Chrome:

[zudov]

Ok, I see the problem now. Seems that if you do socket.close(undefined, undefined) in Chrome it would interpret undefined as 0. Doing socket.close() is allright though.

I think I should change the close type to something like:

close :: Maybe CodeAndReason -> Eff _ Unit

I am not yet sure what would be a good type (and name) for CodeAndReason I'll think about it a bit.

[zudov]

Please take a look at PR.