Perl 6 - the future is here, just unevenly distributed

IRC log for #darcs, 2015-08-15

| Channels | #darcs index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
01:23 mizu_no_oto joined #darcs
02:04 mizu_no_oto joined #darcs
02:22 Riastradh joined #darcs
02:34 mempko joined #darcs
02:47 mizu_no_oto joined #darcs
03:09 mizu_no_oto joined #darcs
03:33 mizu_no_oto joined #darcs
05:11 mempko joined #darcs
11:28 lf94 joined #darcs
12:31 mizu_no_oto joined #darcs
12:36 mizu_no_oto joined #darcs
12:48 mizu_no_oto joined #darcs
14:00 mempko joined #darcs
14:03 mizu_no_oto joined #darcs
14:26 notdan sm: ping
15:17 mempko joined #darcs
15:23 pointfree "darcs stash" --> darcs subspace ?
15:23 pointfree space: a continuous area or expanse which is free, available, or unoccupied [by clutter]
15:23 pointfree subspace: a space that is wholly contained in another space, or whose points or elements are all in another space.
15:23 pointfree ...For those who missed the earlier ( http://irclog.perlgeek.de/darcs/2015-08-13#i_11050753 ) discussion this is about naming a future darcs feature that would hide the effects of all patches not matching a regex, tag, hash, index-range from the working copy thus giving us the part we like about git-style branches.
16:20 aristid_ joined #darcs
16:40 aristid joined #darcs
16:58 mizu_no_oto joined #darcs
17:25 sm hey notdan
17:57 dolio Strictly speaking, I think such a command would probably bad in the same ways as mercurial and git branches.
17:58 dolio I keep everything in separate directories even when using those, because switching branches in place destroys build information and such.
17:59 dolio The only useful thing they allow me to do (from a darcs perspective) is fetch a whole bunch of related repositories at once.
18:15 sm dolio: au contraire, switching branches in place allows you to *reuse* build information, which is considered a feature for large projects
18:32 notdan sm: will you be able to do a meeting/merge soon?
18:45 sm notdan: sure thing how about now
18:46 sm looking
18:50 * Heffalump appears
18:54 sm notdan, I got stuck at http://hub.darcs.net/co-dan/darcsden-to-merge/patch/7ed56300b9896c1d00d2e96c09b48ab02d806ba5 because it seemed complex/tricky
18:55 sm is this the simplest use of the type system to get what we need do you think ?
18:55 sm easy for future contributors to work with etc ?
18:57 Heffalump sm: the way DenConfig is defined (existential quantification of a bunch of type variables to get a single non-parameterised type) is pretty standard
19:28 sm notdan: ping, would like to hear your thoughts as well
19:28 sm I'm some kind of intermediate haskeller, and I found this hard to understand on first and second look. So I'm just asking if there a simpler alternative, or is this the simplest way ?
19:29 sm I have more questions which I'll get into next
19:44 notdan Hm well I don't know any other way
19:44 notdan The existential is needed to abstract away the "parsing" code in `getDenConfig`
19:45 notdan (it's the existentially quantified DenConfig that got your attention, right?)
19:45 notdan those features are used in e.g. darcs itself
19:55 mizu_no_oto joined #darcs
20:18 Heffalump darcs does go relatively far down the road of advanced type system features, particularly when handling patches
20:20 sm notdan, Heffalump: sorry, back. It's the whole thing
20:20 sm starting with Class.hs
20:22 sm when we haskellers use sophisticated type system trickery, someone often comes along the next day and says oh, you could have done that more simply with plain old functions and parameters
20:22 sm darcsden definitely aims to require less expertise from contributors than darcs
20:24 sm it's the web app that's fun to hack new features in while you're stumped on a darcs bug!
20:27 sm so I'm still unclear if there is a less clever implementation we should consider, or is this the only way haskell can deliver the feature we want (ie selecting between different modes & content based on a command-line flag)
20:28 sm notdan, Heffalump what's your opinion ?
21:30 gh_ joined #darcs
21:35 mizu_no_oto joined #darcs
22:33 mizu_no_oto joined #darcs
22:39 dolio sm: No, you can't. It used to cause all sorts of problems at work when I tried to do that. And I'd end up having to clean after switching branches all the time.
22:40 dolio So now I have a directory for each branch, and the differences in building between them no longer cause any problems.
22:48 dolio I could imagine it working in simple cases. But the larger and more complex your project get, the less sound it is going to be to reuse build artifacts across branches.
23:19 sm dolio: ok. I remember some projects talking about this as a killer feature for them, but maybe they are special
23:19 sm and slapdash!
23:20 dolio If your branches are a couple patches off some main branch, I can see it.
23:20 sm there are things I do like about in-place branches, I must say
23:20 sm less cd'ing sounds silly but it's a plus

| Channels | #darcs index | Today | | Search | Google Search | Plain-Text | summary