Thanks for your responses.
Of course I already test every pure parts of my program easily.

For now, user input is treated in 2 different ways:
- direct IO with hgetLine and like functions (blocking)
- the same but I directly push the strings on a TChan for another thread to treat them (not blocking).

The second is easy to test, since all I have to do is to insert test strings on the TChan to simulate user's input.
Maybe I can do everything this way.

But for example I had a bug of two threads inter-blocked on the handle.
These tests were not able to show it, since the test takes place not on the handle itself but a little after.
Maybe "mock handle" can help.

Corentin


On Tue, Oct 26, 2010 at 11:32 PM, Antoine Latter <aslatter@gmail.com> wrote:
On Tue, Oct 26, 2010 at 11:11 AM, Dupont Corentin
<corentin.dupont@gmail.com> wrote:
> Hello again café,
>
> I have a command line program that takes input from various handles
> (actually network sockets) like this:
>
>> s <- hGetLine h
>> etc.
>
> I'd like to unit test this. How can I do?
> I'd like to inject data on the handle so that all the input chain is tested.

I haven't tested it yet, but this "mock handle" will eventually work in GHC 7:

http://hackage.haskell.org/trac/ghc/attachment/ticket/4144/ByteStringHandle.hs

At least, the bug report about how it didn't work has now been closed :-)

I don't remember which operations worked and which didn't - getChar
would work, but getContents would fail.

The others have great advice, that you should try to isolate the logic
of your code from IO for precisely this reason - so it is easily
testable.

Antoine