Perl 6 - the future is here, just unevenly distributed

IRC log for #darcs, 2014-10-15

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

All times shown according to UTC.

Time Nick Message
01:28 stepkut` joined #darcs
01:30 rieper_ joined #darcs
01:31 gh_ joined #darcs
01:31 wavis_ joined #darcs
01:31 c74d joined #darcs
01:32 siel_ joined #darcs
01:40 aristid_ joined #darcs
01:45 c74d joined #darcs
03:41 dolio joined #darcs
04:42 dolio joined #darcs
05:23 lelit joined #darcs
07:20 gal_bolle joined #darcs
09:39 stepkut` joined #darcs
11:17 mizu_no_oto joined #darcs
12:13 gh_ joined #darcs
12:15 gh_ hi
12:31 gal_bolle joined #darcs
13:09 stepkut` joined #darcs
14:03 mizu_no_oto joined #darcs
14:16 gh_ I'm (attempting to) implementing send --minimize-context
14:16 gh_ it seems to be not much code to write but it requires concentration
14:16 gh_ and probably coffee
14:49 gh_ so we have both removeFL and fastRemoveFL…
15:35 byorgey joined #darcs
15:54 gh_ FL is an instance of Commute! how cool is that?
15:55 c74d joined #darcs
16:07 dolio joined #darcs
16:09 alexei_ joined #darcs
16:35 Heffalump :-)
16:36 Heffalump you'd expect it to be, wouldn't you?
16:36 Heffalump what's the heuristic for minimizing the context?
16:36 Heffalump a nice-to-have along with send --minimize-context would be for the same to work with obliterate -O
17:33 gh_ joined #darcs
17:36 gh_ Heffalump, good idea obliterate -O
17:36 gh_ Heffalump, no heuristic so far, just trying to get something working but I think I get it now
18:15 gh_ type witnesses just saved me from doing something wrong and slightly difficult to detect
18:25 Heffalump excellent :-)
18:27 gh_ and now they're giving me headache.. I don't know how to tell that some return witness type may be equal to some other witness type appearing as argument of the same function
18:28 Heffalump do you mean how to tell the type system that they must be equal?
18:28 gh_ that they can
18:28 Heffalump if they only might be equal, then you don't need to say anything
18:29 gh_ let me paste the function and ghc's reaction
18:29 Heffalump you assume they are different
18:29 gh_ http://lpaste.net/6811693774972911616
18:30 gh_ oops, line 272 is:   rmContext NilRL necessary to_be_sent' = necessary :> to_be_sent'
18:30 Heffalump what if you say wD instead of wZ?
18:31 Heffalump in the type signature
18:31 gh_ I don't think I want that because I'm commuting out patches from that whole thing, so the final context may become different
18:32 gh_ the patch I'm trying to commute is paTop
18:32 Heffalump does it type-check if you say they are the same?
18:32 gh_ lemme try
18:32 Heffalump I think the things that you're commuting out end up in the left-side of the (:>) being returned
18:33 gh_ if I do that replace, it complains elsewhere, at the final line of the function
18:34 gh_ no, they are simply dropped
18:35 Heffalump ok, so if you have a free type variable wZ you're saying that the caller can choose what the end context is
18:35 Heffalump if you want to choose, you should use Sealed
18:36 gh_ aha, I hadn't thought about that
18:36 gh_ indeend I don't want them to choose
18:37 gh_ *indeed
18:38 Heffalump right, I see, the return is supposed to be anchored at wA but has some new end-point, because you have commuted out some other patches. You need the anchoring so you can produce a valid context in the send output as well as the valid patches following that context. Sealed sounds just right for that.
18:40 gh_ omg it compiles
18:41 gh_ I was expecting it to complain about the returning wA in the first case of the function. but since the first argument is NilFL, this implies that wA = wB so we're good.
18:41 gh_ :)
19:10 ilbot3 joined #darcs
19:10 Topic for #darcs is now http://darcs.net/ | logs: http://irclog.perlgeek.de/darcs/ | darcs 2.8.5 is out http://darcs.net/Releases/2.8
19:42 gh_ still struggling with types
19:45 gh_ done.
19:45 mizu_no_oto joined #darcs
20:51 byorgey joined #darcs
20:59 gh_ obliterate -O --minimize is coming along
21:02 Heffalump yay
21:03 Heffalump if this works out, any reason we shouldn't make --minimize the default?
21:06 gh_ I've written a very simple shell test and it works, now if we have ~100 patches in the context, I don't know
21:06 Heffalump darcs should be a fairly good test case
21:06 Heffalump but we can see how it goes before making it default
21:10 gh_ the patches are in screened
21:10 gh_ thanks for the help Heffalump
21:12 Heffalump ah, you don't stop at tags? That might be prohibitive in practice.
21:13 Heffalump I think in practice a lot of patches have to be visited to discover what you can and can't remove
21:13 Heffalump anyway, cool, should be easy enough to tweak
21:14 gh_ hmm yes I do stop at tags, since I'm only minimizing the 1st argument of PatchSet, which are the patches more recent than the last clean tag. unless I'm wrong?
21:51 Heffalump that's right

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