tags. No need for headingOffset since PPFrame::expand() adds the [17:59:34] appropriate heading markers. The new preprocessor lacks the capricious stripping [17:59:36] of trailing whitespace present in the old LST, so I updated the parser tests to [17:59:38] suit. My other parser test changes are trivial, and all tests pass. [17:59:40] I didn't use XPath in the end, because I really want to migrate away from DOM at [18:04:44] meh [18:04:46] Tim committing stuff [18:04:49] and not showing up here [18:05:40] ah, thats some old message [18:06:06] domas Got time for an off-topic question? [18:06:08] About performance of MySQL vs. MSSQL [18:06:19] I'm not that much of an expert of mssql [18:06:20] :) [18:06:48] Do you have any idea how the performance compares? [18:06:59] sure mysql is faster! [18:07:03] =) [18:07:21] but mssql is useful. ;) [18:07:25] I was in a project handover meeting the other day, and was a bit surpised to see the hardware requirements for MSSQL was so big. [18:07:55] well, I've seen big mysql boxes too [18:08:16] 2 DUAL-core Xeons and 4 GB memory to do exactly the same as a project I once made, running on a flat database on a P-II with 512MB [18:08:32] And the database was pretty stressed to perform as well. [18:08:56] So I'm wondering if the non-enterprise version of MSSQL is intentionally dumbed down somehow [18:09:41] or it was run by an idiot [18:09:52] I don't think so [18:10:04] i've configured small oracles instances on my desktop that perform comparably to mysql, i'd be surprised if mssql was different [18:10:08] -s [18:10:17] The project had used a consultant to get to that point [18:11:01] We're talking a database with 500k records tat are continiously changed. [18:13:25] if i indent inside a nest of ifs, am I going to get that whitespace considered as wikitext? [18:13:38] http://pastebin.com/m353e8102 [18:13:40] fr example. [18:13:45] Wegge, maybe the consultant was incompetent. Have you actually seen the table structures and queries? [18:14:00] Yes [18:14:19] And they're all perfectly correct and sane? [18:14:22] Seems surprising. [18:14:27] We've been doing this for ages, so I'd be surpirised if it was that bad. [18:14:35] You're not cynical enough. [18:14:45] Maybe not [18:14:57] Most people have no idea how to run databases efficiently, and that probably includes a high percentage of self-appointed consultants. [18:15:04] Experience doesn't necessarily equate to knowledge. [18:15:10] I'm tempted to make a trial on a MySQL instance, just to see if there is any real difference. [18:15:24] With the same code, syntax converted to MySQL? [18:15:33] Would be interesting. [18:15:39] when you say 'non-enterprise version', do you mean the developer edition, or something else? [18:15:47] Of course MySQL can be faster because it has no features, especially 4.0! [18:15:52] We don't need silly stuff like subqueries! [18:16:08] Yes, otherwise it would not be a usefull comparision [18:16:38] I mean the product you get when you just order a Microsoft SQL server 2005, without specifying extra words like Enterprise &c [18:16:45] Not the developer version [18:19:38] alalalala [18:19:41] I'm hunting for parser bug [18:19:55] hello [18:20:05] 03siebrand * r29506 10/trunk/extensions/ (4 files in 2 dirs): [18:20:05] * use wfLoadExtensionMessages for Filepath [18:20:05] * add version in extension credits [18:20:05] * update Translate extension [18:20:19] domas Which one? [18:20:25] what is the new log in german "Verschiebelogbuch"? [18:20:48] no. Vereinigungslogbuch [18:20:49] sry [18:21:13] heuler06 Sounds like a log for the merging of page histories [18:21:40] but it's not possible to do that with only one action? [18:21:48] It is now [18:21:50] could someone take a look at a template i just made and see if i can shorten it somehow? [18:22:08] ah [18:22:11] [[Special:Mergehistory]] or somesuch [18:22:34] http://pastebin.com/m79826672 [18:22:45] i wanted more whitespace to ease readability, but the whitespace became wikitext. [18:22:51] thx, wegge [18:22:54] cu [18:22:59] you call it like {{Death by|axe|monkey|lower=1}} [18:23:55] Morbus, you can put as much whitespace as you want around template parameters. That gets stripped. [18:24:02] Or should. [18:24:18] Simetrical: http://pastebin.com/m353e8102 [18:24:22] that was my first attempt, and it made it a pre [18:24:36] *Simetrical does have to wonder about someone who needs a template to accommodate all the different categories of images of Death by X [18:24:38] i was hoping whitespace was stripped between | | [18:24:49] Hmm. [18:24:55] It's supposed to be. [18:24:57] I think. [18:25:14] Simetrical: horror wiki. there's a lot of different deaths ;) [18:25:45] Simetrical: supposed to be stripped? hrm [18:25:52] I think so. [18:25:57] Maybe it's not done properly. [18:26:33] there's no shortening tips to my second pastebin though? that's just the way it's gotta be? [18:26:47] ok, the problem [18:26:57] any magic word used decreases parser cache time to 1h [18:26:58] i want to accept up to five descriptors for one death scene, and then offer named booleans of "lower[case]" or "display" [18:27:21] any solutions? :) [18:28:29] brion: would people be very unhappy if we had magic word-screwed parser cache increased to one day, instead of one hour? [18:28:49] ARE YOU KIDDING?! WHAT J00 SAY?! [18:28:51] *Morbus spittakes. [18:29:33] domas Do we have {{CURRENTHOUR}} ? [18:29:47] we have stupid stuff [18:29:57] currenttimestamp [18:30:12] currenthour [18:30:15] numberofedits [18:30:17] etc [18:30:38] ok, I add a feature to control that [18:30:40] caching [18:30:45] and set it to 1d on live site [18:31:00] \o/ [18:31:18] So things like {{CURRENTHOUR}} are actually never more than an hour off? [18:31:24] yup [18:31:30] that sucks, eh? [18:31:38] well, stuff is still cached in squids [18:31:53] and there magicwords don't matter [18:31:58] Hello [18:32:31] well, i think the difference is that squid is supposed to cache generated content, and mediawiki is supposed to generate as active a content as possible. [18:32:41] $wgParserCacheDynamicPageExpireTime; [18:32:42] \o/ [18:32:52] well, if someone is really unhappy [18:32:54] Don't you dare. [18:32:59] we can add exclusions [18:33:02] Half the length of that variable name, please. [18:33:17] $wgPCachDynPExp [18:33:20] nah [18:33:20] lol [18:33:22] ++ [18:33:23] I like long names [18:33:41] I like descriptive names, but not *excessively* long names. [18:33:49] domas $wgParserCacheExpireTime = 86400; [18:33:51] this is not excessively long [18:34:01] I have 23" monitor [18:34:03] Wegge, no, it's for dynamic pages only. [18:34:06] and my editor isn't even maximized [18:34:17] Wegge: wgParserCacheExpireTime exists, yes [18:34:27] the code now looks like: [18:34:27] if( $parserOutput->containsOldMagic() ){ [18:34:27] $expire = $wgParserCacheDynamicPageExpireTime; # 1 hour [18:34:27] } else { [18:34:27] $expire = $wgParserCacheExpireTime; [18:34:28] } [18:34:28] Simetrical: ok, so the difference seems to be the starting whitespace. Compare http://pastebin.com/m353e8102 to http://pastebin.com/m177a72be [18:34:31] $wgParserCacheDynamicExpire would be marginally better. [18:34:38] that's my first non-working, and my third working attempt. [18:34:48] the third working is marginally better, but forces the delimiters to the start of the line. [18:35:16] i can't get rid of the nested }}'s on line 6 though [18:35:18] domas, that variable name alone is half of an 80-char line. [18:35:48] Is it possible to create a new Article instance with some prepared content in it and launch edit page on that, from an extension ? [18:36:05] domas Ideally we do away with all that, and offload the work to some clientside .js [18:36:09] I'm pretty sure EditPage.php is hardcoded to use $wgArticle everywhere. [18:36:29] Wegge, relying on JS unnecessarily is not acceptable. [18:36:34] As it is now, the squids will happily keep an outdated page [18:36:49] Relying solely isn't [18:37:20] But oflloading that kin of stuff will make >95% of the users happy [18:37:32] And the rest is already screwed by squid caching [18:37:56] you can't put most uses of magic words into js, though [18:38:10] like the main use of the current date stuff is including per-day pages [18:38:10] If we wanted to, we could [18:38:10] Right, you can't do something like {{#if: something generated by JS}}. [18:38:12] That too. :) [18:38:31] Question: I just did an edit on a page, hit save, all that... in the DB, page.page_latest = 1409, however, text.old_id only is going up to 1408... where is this text being stored in the meantime? [18:39:17] Andrew: text.old_id does not correspond to page_latest [18:39:25] page_latest = rev_id, rev_text_id = old_id [18:40:16] hmm, k, looked like page_latest was a foreign key for text... [18:40:38] AndrewBuis, it's explained in maintenance/tables.sql. [18:42:18] building a program to go through and basically do a search on our web pages for words in our wiki, create a link to them as well as have title=(text of the wiki page) [18:42:59] 03catrope * r29507 10/branches/ApiEdit_Vodafone/includes/ (42 files in 3 dirs): [18:42:59] ApiEdit_Vodafone: Partial svnmerge of latest revisions from trunk. [18:42:59] * Merge split up to prevent hanging mail clients [18:43:07] domas: http://lists.wikimedia.org/pipermail/foundation-l/2008-January/037280.html [18:43:39] 'But now it was just taking a piss at people doing their job and trying to get projects supported, extended and elevated.' [18:43:50] 03catrope * r29508 10/branches/ApiEdit_Vodafone/ (12 files in 6 dirs): ApiEdit_Vodafone: svnmerge part 2 [18:43:54] hmm, not sure about that [18:44:02] certainly a possibility ;) [18:45:18] 03catrope * r29509 10/branches/ApiEdit_Vodafone/languages/messages/ (59 files): ApiEdit_Vodafone: svnmerge part 3 [18:46:18] 03catrope * r29510 10/branches/ApiEdit_Vodafone/languages/ (54 files in 3 dirs): ApiEdit_Vodafone: svnmerge part 4 [18:47:33] I thought API edit was already merged into trunk? [18:47:48] Ehh this was svnmerging from trunk TO apiedit_VODAFONE [18:48:03] Meh, even hangs my mail client for a while when I split it in four [18:48:11] Doesn't hang as long, though ... [18:48:38] get a better mail client ;) [18:49:07] Well it's pretty much Thunderbird's only disadvantage [18:49:27] Is it possible to create a new Article instance with some prepared content in it and launch edit page on that, from an extension ? [18:49:29] Native support for multiple mail accounts is important for us (4-person family, each with a different account) [18:49:44] Telemac you can't create an Article with some content in it [18:50:03] Err to whoever said that: EditPage.php uses $wgArticle *nowhere* [18:50:05] i changed from TB to the bat recently, as well as being faster i find it a lot less annoying :) [18:50:23] RoanKattouw, oh, that's nice. [18:50:31] *Simetrical just uses Gmail's web interface [18:50:33] Telemac you can just create an EditPage object and set its $mArticle member [18:50:53] Oh wait you already have to pass an Article object to EditPage's constructor [18:51:07] Telemac: You can also check out the docs at http://svn.wikimedia.org/doc/classEditPage.php [18:51:50] RoanKattouw: thanx [18:53:23] ok, where was I [18:53:36] alternative is to maintain list of deterministic/nondeterministic magic words [18:53:36] mornin [18:53:41] hi brion! [18:53:53] I'd use Tim as authority, but you fit too. :) [18:53:59] \o/ [18:54:07] brion: {{MAGICWORDS}} make parsercache work just for one hour [18:54:20] and {{MAGICWORDS}} are so much abused in all templates, that generally whole enwiki has this problem [18:55:09] so, is bumping up that hour very bad thing [18:55:15] The downside is that caching content using magic words too long makes it outdated [18:55:26] thats upside! :) [18:55:40] domas: was this newly added? [18:55:46] It's kind of lame when some page tells you Election Day is 3 days away because the cache is only purged every week... [18:55:49] brion: probably not [18:55:52] i don't remember us actually having such a thing [18:55:55] RoanKattouw: allo [18:56:00] AzaTht: Hi [18:56:10] brion: I extended memcached arena to 160gb today [18:56:15] woot [18:56:17] from previous 50g [18:56:20] RoanKattouw, you can manually purge the page if necessary. [18:56:24] I know [18:56:30] With a bot or whatever. [18:56:33] and now I'm sad, as most of objects have insanely short TTLs because of deterministic magic words [18:56:47] RoanKattouw: have a couple of things regarding the api (as usual) [18:56:48] But that results in a lot of manual purges, and essentially accomplishes nothing save for a lot of annoyances ;) [18:56:49] Maybe we should allow some tag like that takes cron-style input, and purges at those intervals. [18:56:53] I thought you would [18:56:56] Hi I have a Mediawiki Site which has a problem. Users cant change existing pages anymore, but can create and change new ones? What could be wrong? [18:56:58] Except not more than, say, once an hour. [18:57:10] RoanKattouw: hey, you never here [18:57:19] petzlux: After they've created them, can they edit them a second time? [18:57:22] Even I as Admin has the same problem [18:57:38] RoanKattouw: Yes, a newly created page can be edited fine [18:57:39] RoanKattouw: prop=imageinfo, iihistory should be changed into iilimit imo [18:58:06] check out www.manipedia.eu , thats the wiki with the problem [18:58:26] RoanKattouw: I would be happy to be able to use backlinks as a prop in an generator [18:58:32] AzaTht: Change iihistory to iilimit? Why? [18:58:38] RoanKattouw: makes sense [18:58:47] Backlinks as prop? Backlinks is a list... What do you mean? [18:58:54] Its frustrating as there is no error message you just get bumped back to the main index page [18:58:58] AzaTht: You mean iilimit defaulting to 0? [18:59:16] yea, or 1 [18:59:46] would make sense to have same sort of limits as other revisions related lists [19:00:22] RoanKattouw: why would backlinks be more of a list than langlinks? [19:00:57] AzaTht: backlinks lists pages, like allpages and friends do. Langlinks lists links, like prop=links, prop=templates and friends do [19:01:00] Makes perfect sense [19:01:06] damn [19:01:19] Keep in mind that API modules are often categorized based on what they return (pages, revs, other stuff...) [19:01:27] goddamn FUCK [19:01:27] The iilimit thing is on my todo list [19:01:37] spotlight seems to reindex my drive *every fucking time* the machine dies [19:01:41] RoanKattouw: isn't a template a page? [19:01:44] *brion stabs apple [19:02:01] AzaTht: Yes, but a langlink isn't [19:02:19] domas: we should get Tim to make quantum magic words [19:02:37] Also, prop=links|langlinks|templates|categories is about stuff *on* the page, list=backlinks|embeddedin|imageusage is about stuff on other pages *referencing* the page [19:02:41] RoanKattouw: but templates is a prop [19:02:42] I guess that's the logic [19:02:56] RoanKattouw: can understand that [19:03:16] brion: enjoying your non-MS box? [19:03:18] petzlux: ALL clicks on preview or edit throw me back to the main page [19:03:19] 03brion * r29476 10/trunk/phase3/ (RELEASE-NOTES includes/RawPage.php): [19:03:19] * (bug 12505) Fixed section=0 with action=raw [19:03:19] PHP's fuzzy type comparisons strike again! A check for $this->mSection != '', [19:03:19] probably intended to protect against an unset value, matched for integer values [19:03:19] of 0 as well. (The fun part is that 0=='' and 0=='0' but '0'!='' :) Since the [19:03:20] parameter is validated through getIntOrNull(), only the null check is necessary [19:03:24] here. [19:03:25] come back to the dark side [19:03:27] but it's rather cumbersome when you have a genrerator that only return one page to be unable to connect list on that page directly [19:03:33] 03dale * r29475 10/trunk/extensions/MetavidWiki/ (9 files in 4 dirs): updates to web interface to manage stream files [19:03:34] It is the only way! buhahaha!!!! [19:03:36] Err brion didn't you commit that one earlier? [19:03:40] HUH?!? [19:03:44] CIA-40 has gone crazy [19:03:51] Reports 29476 before 29475 [19:03:52] RoanKattouw: Okay, but this is recent behaviour, thats what I dont understand ? [19:04:04] wow, yet another fuzzy comp. bug [19:04:06] RoanKattouw: didnt change anything knowlingly [19:04:16] petzlux did you change anything recently? Stuff in LocalSettings.php, updated to a new version, that sort of thing? [19:04:16] AaronSchulz, PHP is great that way, huh? [19:04:21] RoanKattouw: the thing I has is I from a category grabs 1 image [19:04:24] That's like what, four in the past year? [19:04:29] Not like we're a very big project, either. [19:04:44] *RoanKattouw is still angry at PHP for WONTFIXing $wgUser->getEffectiveGroups()[0] [19:04:46] Simetrical: heh, don't rank on PHP :( [19:04:53] it's lazy programming [19:04:59] No it's not. [19:05:02] It's a stupid language. [19:05:09] AaronSchulz: haven't had an MS box of my own in a long long time :) [19:05:22] RoanKattouw: would be rather useful to use lists as props in some sort of generator [19:05:38] *AaronSchulz hugs PHP :'( [19:05:46] Use lists as props? But there isn't really a *difference* between lists and props, is there (other than that props generally don't have limits) [19:05:50] AaronSchulz, have you tried Python or Perl? [19:05:58] Or Ruby, or any other scripting language? [19:06:18] AzaTht: Oh you mean like do generator=something&list=backlinks ? [19:06:21] and about the user thingi, what about an usergenerator, that for example can be used on revisions and allusers, and can have props like groups, contribs etc... [19:06:26] RoanKattouw: yea [19:06:29] That *used* to work, but doesn't anymore [19:06:35] I know :( [19:06:38] AzaTht: The userthingy is on the long-term todo list ;) [19:07:01] 'Cause backlinks only lists backlinks for one page, and a genny can return more than one page (or none at all) [19:07:03] brion: people should use === and such for strong comps. [19:07:08] *simetrical [19:07:33] RoanKattouw: the issue here is that I for the generator generates ONE page гѓ„ [19:07:43] I know [19:07:47] But that's not always the case [19:07:49] brion: Come on, don't you want to tast some sweet Microsoft love again? [19:07:57] What should the software do when the genny returns multiple pages? [19:08:01] AaronSchulz, if I say $var == '', could that ever possibly reasonably be interpreted as including 0? [19:08:11] RoanKattouw: prop=categories, dunno if it's a bug, but the clprop doesn't work when using the prop as an generator [19:08:16] gclprop then [19:08:20] AaronSchulz, if I wanted to match 0 or '' or any variant thereof, I would just use if( !$var ). [19:08:26] AzaTht: That's right [19:08:35] PHP's handling of comparisons makes no sense, which is why it causes bugs like this on a regular basis. [19:08:43] damn [19:08:44] fuck [19:08:44] shit [19:08:48] RoanKattouw: ok [19:08:52] even {{cite}} breaks caching [19:08:57] argh [19:09:00] frustration :) [19:09:00] Simetrical: 0 == "" is fucked up, yeah [19:09:10] 'Cause if you do like generator=categories&gclprop=something&prop=links , all the ApiQueryCategories module returns is an array of pageids that prop=links works its magic on [19:09:21] AaronSchulz, that's the whole problem I'm objecting to. 0 == "any arbitrary string that doesn't start with a number" is even more insane. [19:09:24] There's no way to pass that clprop stuff along [19:09:31] ok [19:09:35] domas: try to keep all of your curse marbles in a row, kthx :) [19:09:37] I think MagicWord.php needs new method, telling the TTLs for each of them [19:09:40] fuck [19:09:41] shit [19:09:41] damn [19:09:42] :) [19:10:19] domas: do that at CZ [19:10:27] domas: none of those things touch caching unless someone added a new feature for that [19:10:29] RoanKattouw: for imageusage, would it possible to avoid bailing out if the given title isn't an image [19:10:30] so first find the feature [19:10:33] then look at it [19:10:34] they delete and replace all incivility with a fun template [19:10:42] Err AzaTht: why [19:11:01] imageusage only takes one title right? [19:11:02] as categorymembers doesn't bail out if the given title doesn't exists [19:11:09] Simetrical: huh? [19:11:12] brion: http://svn.wikimedia.org/viewvc/mediawiki/trunk/phase3/includes/ParserCache.php?view=diff&r1=14200&r2=14201 [19:11:22] oh wait, wrong one [19:11:28] an backlinks etc... [19:11:31] wrong [19:11:33] again [19:11:36] domas [19:11:37] AaronSchulz, aryeh@aryeh-desktop:~$ php -r "var_dump( 0 == 'x' );"bool(true) [19:11:38] strange, annotate shows that [19:11:43] domas: :D [19:11:44] AzaTht: That's because non-existent categories/pages can still be linked [19:11:46] Er, it mashed the output on the end. [19:11:55] Non-existent images can never be included, though, only linked [19:12:05] AaronSchulz, in comparing a string to a number, the string is cast to a number. Strings that don't start with a number are cast to zero. [19:12:16] RoanKattouw: true, but when it bails, it stops other queries as well [19:12:22] Wow, 14201, that's an oldie [19:12:27] Ergo, if you type $var == 0, it will return true for almost any string whatsoever. [19:12:27] I have list=imageusage|backlinks [19:12:31] Right [19:12:42] Hmm, I guess that could be done [19:12:42] Simetrical: really? [19:12:44] if imageusage bails, backlinks won't return anything as well [19:12:46] I can't believe that [19:12:52] AaronSchulz, look at the docs. [19:12:58] AaronSchulz, or try it yourself. [19:12:59] brion: does this heathen speak Teh truth? [19:13:06] AaronSchulz, just run php -r "var_dump( 0 == 'x' );" [19:13:09] brion: it was in virgin 2517 revision of ParserCache [19:13:10] AaronSchulz, bool(true). [19:13:11] AzaTht: I'm on the imageusage thing right now [19:13:34] ok [19:13:37] thanks [19:13:41] though back then it was [19:13:41] if( $parserOutput->containsOldMagic() ){ if ( $value ) { [19:13:41] 20 $expire = "1 HOUR"; # Delete if article has changed since the cache was made [19:13:41] 21 } else { if ( $value->getTouched() != $article->getTouched() ) { [19:13:42] 22 $expire = "7 DAY"; $wgMemc->delete( $key ); [19:13:44] 23 } [19:13:46] ;-) [19:14:01] oh, good old wfQuery() times [19:14:42] AaronSchulz, just run php -r "var_dump( 0 == 'x' );" [19:14:45] bool(true). [19:14:48] AaronSchulz, just run php -r "var_dump( 0 == '1x' );" [19:14:51] bool(false) [19:15:03] opps, extra cp stuff there [19:15:04] AaronSchulz, yup. '1x' == 1. [19:15:14] AaronSchulz: hehe [19:15:22] WHAT [19:15:23] THE [19:15:25] FUCK??? [19:15:26] '2007 was a great year for computing' == 2007. [19:15:33] Simetrical: yes [19:15:40] What about 2007 == '2007 rocked' ? [19:15:47] I am still having a hard time believing this [19:15:49] AaronSchulz, do you believe me now that PHP is a moronic language? [19:15:50] brion: so I can assume it is historical cruft and throw it out? :-D [19:15:51] RoanKattouw: don't think order matters [19:15:57] That was crafted by lazy idiots, for lazy idiots? [19:16:08] RoanKattouw, of course, that's true too. [19:16:09] I guess that when int and string are compared, the string is converted to a string [19:16:10] AaronSchulz: it's lazy evaluation [19:16:21] ", the string is converted to a string"? [19:16:23] RoanKattouw, the string is converted to the numeric type, yes. [19:16:26] Simetrical: Try '2007 rocked' - 2000 [19:16:27] plenty of good things are called "lazy", this is beyond that [19:16:30] To an int, sorry [19:16:34] hehe [19:16:37] '0' == 0 is good lazy [19:16:38] The rationale is that it makes sense to have '007' == 7. [19:16:41] " 2007" != 2007 [19:16:41] hehehe [19:16:42] http://svn.wikimedia.org/viewvc/mediawiki/trunk/phase3/includes/ParserCache.php?revision=2517&view=markup [19:16:52] 'meow' == 35 is not [19:17:00] 'meow' == 0, not 35. [19:17:05] thats how mediawiki used to look like :) [19:17:28] how do you set the default page to be something other than Main_Page [19:17:32] Simetrical: why isn't just removing forward zeros and then doing a '3' = 3 type of comp [19:17:35] AzaTht: Are you having trouble passing Image:Nonexistent to imageusage, or Talk:Nonexistent ? Cause the latter isn't an image [19:17:37]