These are chat archives for helium/airship

8th
Oct 2015
Reid Draper
@reiddraper
Oct 08 2015 00:29
oh no -- what's up?
Charles O'Farrell
@charleso
Oct 08 2015 00:34
See the PR I just raised
Huge f**k up on my behalf
We just hit it in production, I hope I haven't caused any problems for you
Well production-test I should say
Tim McGilchrist
@tmcgilchrist
Oct 08 2015 00:45
fun times
think this'll be a great forcing function to add more tests all round
Reid Draper
@reiddraper
Oct 08 2015 01:40
@charleso not sure I 100% see the issue -- was it calling each callback in its own transformer stack?
so whatever you did in one callback wasn't visible in another?
Charles O'Farrell
@charleso
Oct 08 2015 01:53
Yup :(
@reiddraper So we were using a "pure" state and it was reset for each handler field
I feel pretty stupid, I should have known
But, hey, it compiled so of course it works
Reid Draper
@reiddraper
Oct 08 2015 01:55
right, no worries
it looks like the approach in the PR gives us 80% of what we wanted anyway
the biggest difference is we can't have per-resource transformer stack, right?
Tim McGilchrist
@tmcgilchrist
Oct 08 2015 01:57
I'd personally really like per-resource stack
we're going to look into a few other ways we can get that
Reid Draper
@reiddraper
Oct 08 2015 01:58
so -- maybe we merge in #58 for now, to get things working again, and then re-evaluate our options when we don't feel a time crunch?
ditto re-introducing the parameterized Request and Response, if possible (though I think less important than per-resource transformer stack)
Tim McGilchrist
@tmcgilchrist
Oct 08 2015 02:00
we're using a git submodule of the ambiata repo so we've got something working again
that sounds like a good start, the way it is isn't working
Tim McGilchrist
@tmcgilchrist
Oct 08 2015 02:06
just as an aside, we had issues with the http-types bump where some packages had upper bounds on http-types that didn't work nicely with 0.9
Reid Draper
@reiddraper
Oct 08 2015 02:08
happy to revert that too if we want
Tim McGilchrist
@tmcgilchrist
Oct 08 2015 02:09
that might just be our issue, we have a mixture of scotty and airship right now
the problem will go away when scotty does
scotty-web/scotty#182
Reid Draper
@reiddraper
Oct 08 2015 02:10
ah, right
let me see what Wai wants
just >= 0.7
merged #58
Tim McGilchrist
@tmcgilchrist
Oct 08 2015 02:27
cheers, again sorry about that
Charles O'Farrell
@charleso
Oct 08 2015 02:43
@reiddraper Thanks. Let me know if you want the parametized Request/Response back, I think that should work. I don't have strong opinions about whether it's worth it, but feel bad about the run-around.
Reid Draper
@reiddraper
Oct 08 2015 02:47
I'd say if we can, it's nice
would be nice to be able to run QuickCheck without using the https://hackage.haskell.org/package/QuickCheck-2.8.1/docs/Test-QuickCheck-Monadic.html module
Charles O'Farrell
@charleso
Oct 08 2015 10:10
Hmm, might be a little trickier than I thought. I've created a branch with the revert and merge conflicts fixed if anyone is bored and wants to have a play:
The problem is both fromWaiRequest and toWaiResponse pin the 'm' to IO
Charles O'Farrell
@charleso
Oct 08 2015 10:58
I was right the first time, can do request but not response:
helium/airship#59
helium/airship#60
And I'm spent