simendsjo 2017-02-05 04:08:18
Xnuk: Very informative, thanks. I'll study your solutions carefully.
boccato 2017-02-05 04:25:21
Should I still put package boundaries on project.cabal build-depends if I am using stack with a resolver?
Polarina 2017-02-05 04:30:23
boccato, yes.
rx 2017-02-05 04:31:05
How useful is haskell in the industry?
rx 2017-02-05 04:31:30
Hello?
davenpcm 2017-02-05 04:31:36
What are the major examples of http client libraries in Haskell? Googling found me quite a few more than I liked.
davenpcm 2017-02-05 04:32:02
rx - It's fairly useful, but has some trouble with adoption in some orgs.
opqdonut 2017-02-05 04:32:07
rx: used quite a bit in financial tech
rx 2017-02-05 04:32:14
Just wanted to know if anyone in here worked on anything fun personally.
rx 2017-02-05 04:32:21
But yeah, I can imagine.
opqdonut 2017-02-05 04:32:40
our project uses postgrest, and I realised only after a month or so that it's written in haskell
opqdonut 2017-02-05 04:32:59
(http://postgrest.com/en/v0.4/)
boccato 2017-02-05 04:33:17
Polarina: Why is that? Won't it use the versions on the chosen resolver?
rx 2017-02-05 04:33:20
I'll take a look, thanks!
hpc 2017-02-05 04:33:43
a huge number of people use pandoc
opqdonut 2017-02-05 04:33:52
yeah pandoc is a great example
rx 2017-02-05 04:34:17
Lol where has this software been all my life
Polarina 2017-02-05 04:34:31
boccato, it would, yes. However, your package will most likely be broken for non-stack users.
boccato 2017-02-05 04:35:53
Hmm, haven't thought about that.
maerwald 2017-02-05 04:38:29
boccato: use upper bounds only if you _know_ it doesn't work with later ones, not if you are not sure
hpc 2017-02-05 04:39:42
usually you should add an upper bound on major version numbers anyway, because semantic versioning says those are where breaking changes are allowed to happen
boccato 2017-02-05 04:40:17
Any tips on lower bounds?
maerwald 2017-02-05 04:40:34
hpc: that's still guessing
hpc 2017-02-05 04:40:35
as far as you feel like testing if it works
hpc 2017-02-05 04:41:08
maerwald: not really
maerwald 2017-02-05 04:41:11
hackage is currently "rolling release"... and should be treated as such
boccato 2017-02-05 04:41:21
I am browsing around cabal files on github and it seems a few packages get their bounds defined but most dont.
maerwald 2017-02-05 04:41:34
hpc: yes it is, just because breaking changes are allowed to happen doesn't mean they actually do happen for package XY
maerwald 2017-02-05 04:41:37
it's just guessing
maerwald 2017-02-05 04:42:02
if you look at any rolling release distro, none of them use upper bounds, because it makes the depgraph explode
hpc 2017-02-05 04:42:04
https://pvp.haskell.org/?rdfrom=https%3A%2F%2Fwiki.haskell.org%2Findex.php%3Ftitle%3DPackage_versioning_policy%26redirect%3Dno
maerwald 2017-02-05 04:42:18
instead you fix stuff as you go
maerwald 2017-02-05 04:42:41
unless a package is really "broken" and cannot just be fixed... then you set an upper bound
maerwald 2017-02-05 04:43:31
hpc: I'm aware of that, it's still guessing as I pointed out, because it depends on used features etc
hpc 2017-02-05 04:43:32
"When publishing a Cabal package, you SHALL ensure that your dependencies in the build-depends field are accurate. This means specifying not only lower bounds, but also upper bounds on every dependency."
maerwald 2017-02-05 04:43:46
hpc: yeah and that suggestion is wrong
hpc 2017-02-05 04:44:09
then get it changed
hpc 2017-02-05 04:44:11
but that's the policy
maerwald 2017-02-05 04:44:25
I don't blindly follow policies just because someone put them on some wiki
hpc 2017-02-05 04:44:31
it's not on the wiki
hpc 2017-02-05 04:44:35
it's on pvp.haskell.org
maerwald 2017-02-05 04:44:58
so?
