Camelia, the Perl 6 bug

IRC log for #metacpan, 2013-09-09

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

All times shown according to UTC.

Time Nick Message
00:33 shmuel joined #metacpan
01:10 klapperl_ joined #metacpan
01:37 omega oalders: yeah, I know. That one is a mix of not following the standard, but following it cloesly enough that something gets parsed, and our relaxed group parsing because of ribasushi "refusal" to follow the spec, while we try to parse it anyways
03:06 haarg there's a few things wrong with how metacpan treats the changes file
03:06 haarg and i don't know parsing well enough to know the proper fix
03:07 haarg but generally, changelog entries starting with * shouldn't be treated as groups unconditionally
03:07 haarg lots of dists use * instead of -
03:12 omega yeah, I know, thats one of the problems
03:12 omega I think we have to rip that part out again
03:12 omega a line based parser can only work so well for complicated situations
03:13 haarg indentation seems like the only thing you can really depend on
03:14 omega yeah, which means it isn't really groups, but nesting
03:14 haarg combined with recognising lines vs continuations
03:14 haarg which you can do a rough job by checking for single non word characters probably
03:15 haarg unfortunately, the cpan-changes spec isn't much of a spec
03:16 omega nope
03:16 omega and unfortunately, it sometimes parses things that aren't according to spec (like dbix-class, which I tried to fix, which lead to this mess)
03:16 omega and it parses that in a bad way
03:17 haarg well
03:17 haarg dbic doesn't follow the spec if you read the english
03:17 haarg but it follows the technical part of the spec just fine
03:17 omega "doesn't follow the spec" was wrong to say
03:17 omega I meant, gets parsed in a broken way, rendering the end result useless
03:18 haarg yeah
03:19 haarg that's mainly because the spec itself is pretty meaningless
03:20 haarg in an attempt to follow all possible existing patterns, it made itself so general that it has almost no value as a spec.
03:22 mauke_ joined #metacpan
03:24 preflex_ joined #metacpan
03:28 omega yeah, hence the idea that "our" parser could be more relaxed, because we don't really care about specs, we care about getting information out there
03:28 omega but it is hard to do right
03:31 haarg there are two options i can think of that seem reasonable
03:32 haarg first is nested entries, where sub-entries need to be [*-whatever_else]\s*
03:33 haarg and otherwise indented entries are continuations
03:34 haarg the second would be to define a more strict format, and if it fails just show the raw text data, with white-space: pre
03:50 omega the second option gets complicated by getting the latest release only
03:54 haarg i mean more by following the spec strictly, so version headings would still be respected
03:54 haarg but everything else that was indented after it would be treated as the changes for that version.
04:12 ether_ joined #metacpan
05:31 ribasushi <omega> the second option gets complicated by getting the latest release only <--- how so?
06:08 plu joined #metacpan
06:45 omega Well, not if we still expect "version lines" to be spec conformant I guess
06:45 omega but I understood it as wanting to match more releases that cpan::changes doesn't
06:45 omega there is a lot of weird stuff going on in Changes files on cpan :)
06:46 shmuel left #metacpan
06:46 omega (Other than that someone needs to write that code)
06:47 haarg being strict about the version lines makes sense
06:58 * ribasushi still thinks that attempting to parse the file is a step in the wrong direction
06:59 ribasushi even trying to read the version itself is already a step into the woods of heuristics
07:00 dpetrov_ joined #metacpan
07:06 haarg it's a least worst situation
07:06 haarg i think parsing is more reasonable than diffing
07:06 ribasushi and I think exactly the opposite ;)
07:07 ribasushi haarg: pistols or rapiers?
07:08 * haarg points to https://metacpan.org/release/Moose as evidence that diffs won't always work well
07:08 haarg broken changelog entry, likely to be fixed
07:08 haarg which reminds me
07:08 ribasushi haarg: but it *will* work
07:09 ribasushi haarg: -U1 with taking only one hunk (the longest one)
07:09 ribasushi yes, still heuristic, but expecting the relevant changes to be in one hunk is beyond reasonable
07:10 omega except when someone reformats the entire changelog :)
07:10 bouncy joined #metacpan
07:10 ribasushi omega: and then you display the entire changelog - how much worse is that than the half-assed stuff we have now?
07:10 ribasushi another avenue is to tighten up the spec and not show *anything* that does not strictly conform
07:11 * ribasushi will be happy to just have no changes display for DBIC at all, if it chokes on them ;)
07:11 * ribasushi likes all-or-nothing behaviors
07:14 haarg i do think the spec is next to useless as it is.  making it more strict can only be an improvement.
07:20 * haarg needs to file an issue to have the syntax highlighter disabled for large enough files
07:33 daxim joined #metacpan
07:55 bouncy joined #metacpan
10:21 ribasushi haarg: https://github.com/CPAN-AP​I/metacpan-web/issues/948
10:21 dipsy [ Suggestion to revert 7695cea and go for a more strict "all or nothing" approach · Issue #948 · CPAN-API/metacpan-web · GitHub ]
10:21 ribasushi figured I'll bring everything under one roof
10:21 ribasushi mo: ^^
10:21 ribasushi oalders: ^^
10:21 ribasushi omega: ^^
12:02 lestrrat no biggie, but why doesn't metacpan list my .xs modules and stuff ? they should be in xs/* directory. https://metacpan.org/source/DMAKI/​Crypt-DH-GMP-0.00011/lib/Crypt/DH
12:02 dipsy [ Browse DMAKI/Crypt-DH-GMP-0.00011/lib/Crypt/DH - metacpan.org - Perl programming language ]
12:02 lestrrat oh wait
12:02 lestrrat wrong directory.
12:02 lestrrat Browse link was taking me into lib/Crypt, I expected it to be at the top level.
12:03 lestrrat is that normal?
12:04 mauke browse generally takes you to the directory the current file is in
12:04 mauke so if you're looking at .../GMP.pm, it takes you to lib/Crypt/DH/
12:06 daxim joined #metacpan
12:08 lestrrat I guess I was used to search.cpan.org behavior (since there's only one Browse link, it always goes to the same location :)
12:08 daxim_ joined #metacpan
14:15 rjbs joined #metacpan
14:19 rjbs left #metacpan
16:19 rwstaune` joined #metacpan
17:42 bowtie joined #metacpan
18:31 rwstauner left #metacpan
18:31 rwstauner joined #metacpan
19:09 Topic for #metacpan is now Have you installed your MetaCPAN VM? https://github.com/CPAN-API/metacpan-developer | Chat logs available at http://irclog.perlgeek.de/metacpan/
20:20 daemon joined #metacpan
20:20 shibayu36 joined #metacpan
20:21 mo joined #metacpan
20:21 mauke joined #metacpan
20:21 burnersk joined #metacpan
20:21 rGeoffrey_zzz_ joined #metacpan
20:21 preflex joined #metacpan
20:22 Kovensky joined #metacpan
20:23 ribasushi joined #metacpan
20:23 lestrrat joined #metacpan
21:04 meraxes joined #metacpan
23:06 mauke joined #metacpan
23:07 preflex joined #metacpan
23:12 BinGOs_ joined #metacpan
23:13 idn_ joined #metacpan

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