Perl 6 - the future is here, just unevenly distributed

IRC log for #darcs, 2015-08-19

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

All times shown according to UTC.

Time Nick Message
00:10 mempko joined #darcs
01:47 ilbot3 joined #darcs
01:47 Topic for #darcs is now http://darcs.net/ | logs: http://irclog.perlgeek.de/darcs/ | darcs 2.10.0 is out http://darcs.net/Releases/2.10
01:52 mizu_no_oto joined #darcs
02:32 mizu_no_oto joined #darcs
03:11 favonia joined #darcs
06:58 sm joined #darcs
08:37 favonia joined #darcs
11:05 mizu_no_oto joined #darcs
11:20 mizu_no_oto joined #darcs
12:31 mempko joined #darcs
15:05 Riastradh joined #darcs
17:18 mempko joined #darcs
18:54 * Heffalump appears
19:04 notdan Hi Heffalump!
19:05 notdan I've pushed some patches to http://darcs.updog.xyz/updog/darcsden-local
19:05 notdan and did some rebasing so you probably should wipe the latest ones
19:06 Heffalump great, will try now
19:06 notdan So, this is the situation: https://privatepaste.com/9821267917
19:07 notdan I have tested the 'den' command with the HTML page and it works fine for me, but I am not sure how it is on Windows
19:07 notdan The '.' repository should be expanded now
19:08 notdan The non-repositories in darcsden_repos should be ignore -- with a caveat
19:08 notdan "The situation: the presense of a non-repository in the list makes the whole darcsden throw errors. Should we check for the "_darcs" directory or should we use tryIdentifyRepo? In the former case we still might stumble upon the problem of the whole darcsden not working because of one non-repo with a _darcs subdir."
19:08 Heffalump can't you patch darcsden not to throw errors in that case?
19:09 notdan I've also made a flag for enabling/disabling record, but I haven't put the button on the page yet
19:09 notdan hm
19:10 Heffalump I think it should list the repo and then complain when you try to browse into it
19:10 Heffalump either with the normal backend or the local backend
19:10 notdan I can try; it is kinda hard to locate the source of the error in darcsden, because the exceptions are just propagated without any useful information
19:10 notdan hm
19:10 notdan yes
19:10 Heffalump you can try searching for the text of the error, in both darcsden and darcs
19:12 notdan Yeah, that's always useful
19:15 notdan OK, I think I can pinpoint it to two places
19:16 notdan I had another problem with getting rid of parsing problems -- it is really hard, as it turns out :(
19:16 Heffalump ooh, it just launched the right URL on Windows!
19:17 notdan I tried a bit of regex trickery, but taggy fails on something as simple as "<script ></script >"
19:18 notdan Then I tried various parsers.. the only one I found that parses the HTML correctly is tagsoup
19:18 notdan But it doesn't have a lens interface, so using it means rewriting all the logic
19:19 notdan there is actually a lens interface to tagsoup, IIUC, but it is part of the hexpat library which needs a C dependency IIRC
19:19 Heffalump how about manual string processing?
19:19 notdan yeah, that's what I ment by regex trickery
19:19 Heffalump what is 'taggy' in that context?
19:20 notdan Oh, the libraries 'taggy' and 'taggy-lens'
19:20 notdan I read about it on github and reddit some time ago and using them I wrote the tester
19:20 Heffalump how about literally doing string processing by hand?
19:21 Heffalump character by character
19:21 Heffalump you just need to transform the HTML you actually get into HTML your existing code can parse
19:23 notdan It is possible, but not expressible by a fold/map :(
19:24 notdan It should be possible, it should involve just context-free grammar
19:25 Heffalump auto-launch seems to work fine on Linux too, apart from the race condition
19:25 Heffalump can you give an example of the "bad" and "good" HTML?
19:25 notdan wait, the race condition is still happening? argh :{
19:25 Heffalump oh, I thought you hadn't fixed that yet?
19:26 notdan "bad": "<script    \n></script   \n>", "good": "<script></script>"
19:26 notdan Heffalump: no, but I thought having a separate HTML page would fix it
19:26 Heffalump why would it? The timing of launch still matters
19:27 notdan I just though launching a browser/opening a web page would give it "enough" time to start the server
19:28 Heffalump you're like the people that fix race conditions by inserting a sleep :-)
19:28 notdan hehe
19:28 * notdan blushes
19:28 notdan yeah, that's just silly
19:29 notdan OK, I had the JS ajax polling code, I will try to use that
19:29 Heffalump why not just poll in the Haskell code?
19:29 * notdan just realized that he can simiulate the data race on his computer for testing purposes
19:30 notdan hm
19:31 notdan I used to think it would be slower, but I don't have any rational reasoning behind that
19:35 Heffalump so '.' seems to work and den seems to launch URLs, making my general experience much happier :-)
19:35 Heffalump anything else to look at?
19:37 notdan Hm, I think I managed to get darcsden not to scream all over
19:37 notdan and only report an error when you actually try to open the repository
19:37 Heffalump cool
19:37 notdan but besides that, no, not really
19:38 notdan I still don't get how to make the HTML parsable :( tbh
19:38 Heffalump sorry, missed your response about that, only thinking about it now
19:38 notdan oh yeah
19:38 Heffalump my idea is just to implement a state machine in Haskell code directly
19:39 Heffalump foo _ ('<':xs) = '<':foo TagOpened xs
19:39 notdan sorry it's kinda of hard to keep track of the interlocular's activity when talking not face-to-face :)
19:40 notdan yeah, a function like that won't work, because we need an unbounded look-ahead
19:40 Heffalump it can always be two pass
19:40 Heffalump it doesn't have to be nice, it just has to work
19:49 Heffalump does that sound plausible?
19:50 notdan It should be, I was just uh out of luck :(
19:50 notdan But your suggesting about having several passes made me think
19:51 notdan and I reckon that the 'foo' might be expressable as a fixed point
19:51 Heffalump that still sounds over-complicated :-)
19:51 Heffalump is there a single pass algorithm on the reversed string?
19:52 notdan Possibly
19:52 Heffalump but a fixed point is fine too if it works
19:52 notdan the worst thing is that I don't really know what is "bad" html and what is "good" html
19:55 Heffalump good html is html that makes your tests work..
19:55 notdan Right, I meant I don't know in advance what's good and what's not; I have to test everything basically
19:56 Heffalump but you have tests to verify whether you've succeeded
19:56 Heffalump I mean, I'd rather you didn't just hardcode the expected (bad HTML, good HTML) pairs, but it doesn't matter too much if it's a bit shonky, at least initially.
19:58 notdan right
19:59 Heffalump btw, re regexes, why can't you just replace (\s*)> with > ?
20:00 notdan OK, I think I got it. Now just to make sure that it doesn't pass the tests it shouldn't pass..
20:01 notdan Heffalump: hm I think there were greediness issues, but I am not sure
20:01 Heffalump :-)
20:02 * Heffalump happily explores around his repositories
20:06 Heffalump anything else to discuss? I'm not quite sure when I'll be available tomorrow.
20:06 notdan Hm, nothing that I can think of off top of my head
20:06 Heffalump remember to think about merging as well as the outstanding implementation items
20:07 Heffalump i.e. as soon as you can carve out something mergeable, do it and let sm know
20:07 notdan So, the agenda seems to be: textarea field for /record; button for /record; polling in order to avoid data races
20:07 notdan yea
20:08 Heffalump and test fixing
20:08 notdan so speaking about merging, what did we decide on <http://hub.darcs.net/co-dan/darcsden-to-merge/patch/7ed56300b9896c1d00d2e96c09b48ab02d806ba5>?
20:08 Heffalump is that a question for sm?
20:09 notdan I meant, sm asked your opinion at some point, and you greenlit it, right?
20:10 Heffalump oh, the existential?
20:10 Heffalump yes, I thought he was going to merge that
20:10 notdan sm: hi! Regarding merging -local changes. I think we got stuck on the naming issues. So what exactly do you want to change in terms of that?
20:15 Heffalump re next meeting, how about you report on progress on here as and when, and I reply when I can?
20:17 notdan Alright
21:46 maerwald joined #darcs
22:01 mempko joined #darcs
22:11 mizu_no__ joined #darcs
22:12 mizu_no__ joined #darcs
22:21 favonia joined #darcs

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