[Haskell-cafe] Network Connect Timeouts

[Mark Fine]

Hi Chris,

Yeah, definitely looks like I can reconstitute *runTCPClient* with a
wrapped *getSocketFamilyTCP* - messy, but should work. Thanks!

Mark

On Sat, Oct 14, 2017 at 5:17 PM, Chris Wong <lambda.fairy at gmail.com> wrote:

> Hi Mark,
>
> From reading the source code for *runTCPClient* [1], it looks like a thin
> wrapper around *getSocketFamilyTCP* and *close*.
>
> It should be sufficient to copy that definition into your project and wrap
> the *getSocketFamilyTCP* call in a *timeout*. You might need to import
> *Data.Streaming.Network.Internal* to get access to the *AppData*
> constructor.
>
> Hope this helps!
>
> [1] https://hackage.haskell.org/package/streaming-commons-0.1.
> 18/docs/src/Data-Streaming-Network.html#runTCPClient
>
> On Sun, Oct 15, 2017 at 9:08 AM, Mark Fine <mark.fine at gmail.com> wrote:
>
>> Hi Chris,
>>
>> In my application, I'm using *Data.Conduit.Network* from *conduit-extra* directly
>> (*Data.Streaming.Network* from *streaming-commons* indirectly). The
>> timeout I'm interested in managing is the connection establishment timeout
>> - how long does the connect call wait to establish a connection. I haven't
>> been able to figure out how to control this outside of applying a
>> non-conditional timeout to the connect call (which is a timeout on the
>> lifetime of a connection, not a timeout on the establishment of a
>> connection). Thanks!
>>
>> Mark
>>
>> On Sat, Oct 14, 2017 at 3:09 AM, Chris Wong <lambda.fairy at gmail.com>
>> wrote:
>>
>>> Hi Mark,
>>>
>>> What networking library are you using?
>>>
>>> There should be a lower level interface which allows for managing the
>>> lifetime of a connection by hand.
>>>
>>> Chris
>>>
>>> On Oct 14, 2017 15:17, "Mark Fine" <mark.fine at gmail.com> wrote:
>>>
>>>> I'd like a configurable network connect timeout. Has anyone solved this
>>>> reasonably?
>>>>
>>>> *System.Timeout.timeout* times out connections, but works on the
>>>> granularity of the life of the connection. Is there a reasonable way to
>>>> configure a timeout around just establishing the connection? Maybe
>>>> something like a conditional timeout that enables an action to disable the
>>>> timeout once it expires?
>>>>
>>>> As a workaround, I'm spinning trying to successfully connect first
>>>> before trying to connect for real:
>>>>
>>>> -- | Try the TCP connection and see if you can connect...
>>>> --
>>>> tryTCPClient :: Int -> ClientSettings -> IO ()
>>>> tryTCPClient microseconds settings = do
>>>>   ok <- newIORef False
>>>>   void $ timeout microseconds $ runTCPClient settings $ const $
>>>>     writeIORef ok True
>>>>   ok' <- readIORef ok
>>>>   unless ok' $
>>>>     tryTCPClient microseconds settings
>>>>
>>>> -- | Wrap runTCPClient with a connect timeout.
>>>> --
>>>> -- Tries the TCP connection first, and the runs the regular
>>>> runTCPClient.
>>>> -- Of course, this only enforces a TCP connect timeout on the first
>>>> connect.
>>>> -- The second TCP connect has no timeout :(
>>>> --
>>>> runTCPClient' :: Int -> ClientSettings -> (AppData -> IO a) -> IO a
>>>> runTCPClient' microseconds settings action = do
>>>>   tryTCPClient microseconds settings
>>>>   runTCPClient settings action
>>>>
>>>> I've also tried running the *tryTCPClient* in its own thread
>>>> concurrently with *runTCPClient* and throwing an exception if it can't
>>>> connect in *microseconds*. None of these offer an actual true
>>>> connection establishment timeout.
>>>>
>>>> Has anyone found a way to solve this? Thanks!
>>>>
>>>> Mark
>>>>
>>>>
>>>>
>>>> _______________________________________________
>>>> Haskell-Cafe mailing list
>>>> To (un)subscribe, modify options or view archives go to:
>>>> http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe
>>>> Only members subscribed via the mailman list are allowed to post.
>>>>
>>>
>>
>
>
> --
> Chris Wong (https://lambda.xyz)
>
> "I had not the vaguest idea what this meant and when I could not remember
> the words, my tutor threw the book at my head, which did not stimulate my
> intellect in any way." -- Bertrand Russell
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.haskell.org/pipermail/haskell-cafe/attachments/20171016/a33a0be9/attachment.html>