[09:42:51] if i want to report bug about the new editbar in vector, what product & component to use the best? [09:43:33] MediaWiki extensions / UsabilityInitiative [09:56:43] is there any tracking bug for editbar? [09:59:08] No [09:59:57] then i probably will start one to have em all together [10:00:37] OK [14:44:19] adam_miller: Dude your new highlighting code just went live and it's awesome :D try typing "ph.d" and "ph d" into the search box on enwiki [14:45:09] awesome [14:45:20] thank god it works [14:45:57] i was a bit nervous being that we couldn't test it against lucene [14:46:06] "ph d" even highlights "PH (d...)" correctly [14:48:22] yeah i think i got it pretty close [14:49:42] it doesn't do the namespaced stuff, like turning "wp:" to "wikipedia:" and it doesnt do well with non latin characters like matching "e" to "??" [14:50:27] That's true [14:50:42] But you'd probably need Lucene support for that [14:51:39] right, when i got into that stuff i started thinking we were going at it from the wrong end [15:23:39] is it live on cswiki too? so i could check the issues we had with it? [15:25:56] RoanKattouw: i do not agree with marking of that h3-removing revision as fixme, i'll put some comment there in few mins, please check it and consider oukeying it again, thx [15:26:17] From what I hear it affects print display [15:26:54] Yes, it's live on all wikis [15:34:58] RoanKattouw: that is just logical delusion, i'll have arguments and evidence there against [15:39:01] so the comment is there [15:39:11] *Danny_B is going to check the search suggest [15:40:17] I see, looks good [15:40:35] TBH I hadn't even clicked those things, I just wanted to get this stuff the hell out of the door [15:40:50] I'll deploy it later today [15:40:55] it doesn't bold all lines if they don't match the search string, is that intended behavior? [15:41:25] adam_miller: ^ [15:41:56] can you rephrase that? [15:42:27] adam_miller: open http://cs.wikipedia.org/wiki/Speci%C3%A1ln%C3%AD:Pr%C3%A1zdn%C3%A1_str%C3%A1nka [15:42:57] type in search (letter by letter and wait for the suggestions): [15:43:27] looks right [15:43:40] e.g. type in Sp [15:43:52] type p??????e [15:43:56] letter by letter [15:44:00] and see how it changes [15:44:07] The S looks wrong though [15:44:14] even thoug: [15:44:22] type ?? only [15:44:28] it won't get hilited at all [15:44:30] It highlights the first S of every word, so that's two Ses in Spojen??_st??ty_americk?? [15:44:51] Danny_B: Try capital ?? [15:45:13] RoanKattouw: bah, it should ignore the case as it always used to [15:45:29] and capital doesn't work too [15:45:49] How do you do case-insensitivity for non-Latin chars in JS? [15:45:52] (i even tried other letters with accent to prove it's not cached) [15:46:24] You're right that the capital letter doesn't work either, that's weird [15:46:35] i don't know how you do the suggest, but to me it seems as simple algorithm: [15:46:44] send ajax way the searched string [15:47:01] receive json with array of titles [15:47:22] adam_miller: Why does typing ?? not highlight occurrences of ?? at the beginning of the title, even though there's no small letter / capital difference or anything involved? [15:47:26] iterate through the array and bold the appropriate number of letters in every item [15:47:29] then display [15:47:31] simple [15:47:37] Define "appropriate number of letters"? [15:47:38] so why it isn't working? [15:47:58] appropriate number of letters == length of input in searchbox [15:48:00] Hah [15:48:04] Yeah that's what we had before [15:48:08] Now type in WP:A [15:48:15] That's four letters, right? [15:48:26] RoanKattouw: i've got no clue...it gets weird with non latin characters [15:48:27] So that highlights [b]Wiki[/b]pedia:AN/I [15:48:35] regex assumes their breaks in words at times [15:48:42] so i would guess that's what's happening there [15:48:47] i know, you have to return the namespace aliases as well [15:48:56] i use the /b flag to look for the beginning of words [15:48:59] So, it's not that simple [15:49:08] or even return hilited stuff already [15:49:14] hilite serverside [15:49:16] Yes, that's where we want to go [15:49:20] Highlight on the Lucene side [15:49:29] Because Lucene is doing all the crazy accent stuff [15:49:46] adam_miller: ?? _is_ latin character :-P [15:50:41] non western characters? [15:51:01] wtf do i call letters i have to hit the option key to type? [15:51:24] RoanKattouw: i was right.....try this in console [15:51:38] "??anidu".match(/\b??/) [15:51:40] vs [15:51:41] "non (lower)ascii" probably [15:51:45] "Zanidu".match(/\bZ/) [15:52:51] if you had (either in response or any wg var) list of namespaces and aliases you could simply use it for matching [15:53:08] simply, eh [15:53:20] want to write it? [15:53:56] i think the right thing to do is to have lucene do it [15:54:55] typing "1 (" suggests all kind of years 1999 etc... - intended behavior? [15:55:24] why did we actually turned off the previous suggest algorithm? [15:55:42] which was matching titles alphabetically [16:02:56] adam_miller: why do you use \b in the beginning? [16:04:26] so it doesn't match every instance of the character 'h' when you type 'h' [16:04:42] RoanKattouw: that spojene staty americke hiliting of both "s" is apparently because of "\bs" [16:05:03] adam_miller: "^" wouldn't work? [16:05:13] no [16:05:28] what was the problem with it pls? [16:05:40] do you know what ^ does? [16:05:55] that's begin of string/input [16:06:05] right, and i didnt want that [16:06:35] i thought bolding should bold only the appropriate part _from the start_ not in the middle [16:06:38] if you type "fun" and you get "fun fun fun fest" as a result, i want each fun to be highlighted [16:06:58] it leaves out a lot of articles [16:07:50] but is it still searching alphabetically? [16:09:25] if yes, bolding in the middle is not good, sicne it gives the impression the auto-suggest can search in other manner than just alphabetically [16:11:58] i don't know what you mean by alphabetically...it searches against an index based on the users query [16:13:15] as in "starting with that string" [16:14:21] so you think it should only match the first instance of the users query? [16:14:34] yes. [16:14:40] that is what used to be before [16:14:42] i can't use ^ because of things like "(I Can't Get No) Satisfaction" [16:15:08] and i feel like there are other more interesting cases which i can't recall right now [16:16:26] well, hiliting in the middle doesn't have a sense since you do not return anything else but titles beginning with such string, not containing it [16:16:38] unless you plan to do so of course [16:19:36] Danny_B: can you file a bug with specific queries you feel are highlighted incorrectly? [16:19:46] and stuff like "(I Can't Get No) Satisfaction" still can be processed on the client-side first to escape any regexp-like characters [16:21:14] so let me summarize to prevent misunderstandings and unnecessasry work: [16:22:37] we want suggest to returnfrom server anything matching the alphanum chars typed into searchbox [16:23:34] then we want to hilite everything from the beginning matching the string (case ignore) and ignore any non alphanum chars [16:23:50] do we care about the number of chars between? [16:24:17] should typing "ab" return and hilite also "a+b=c" ? [16:26:08] or "a b" should hilite "a+b=c" [16:26:11] or none? [16:26:27] i'd say the second [16:26:52] the first one should be only for dots [16:27:17] ie. typing CIA should result in C.I.A. too [16:27:53] in cs typography it should be c. i. a. [16:28:46] well, then maybe local settings for each wiki? [16:30:27] yeah, just shove it with as many settings as possible and let loacl wikis decide in flames...i mean discussions [16:39:23] interesting: [16:39:50] "i can" returns also "(i can't get no) satisfaction" [16:39:52] but [16:40:00] "i can t" does not [16:40:13] one would assume ' is also being ignored [17:03:07] RoanKattouw: could you hide NTOC from the prefs when users are not allowed to use it ? Like atm ? [17:03:19] Yes, been meaning to do that forever [17:03:32] it came up again at nl.wp [18:25:40] why do we use _image_ to create _border_? [18:25:48] Ask Trevor [18:25:58] When he's around and not having lunch (presumably) at OSCON [18:26:01] :) [18:26:25] I knew the reason once, but I forget [18:26:44] there is no reason to do that [18:43:51] I think it's due to the degradation [18:43:58] and IE... [18:44:11] Yes, there is /some/ reason [18:44:17] I suggest you e-mail Trevor to ask why [18:48:10] Platonides: ie doesn't have any problems with borders. and i can't imagine any degradation. i really wonder why. will catch trevor and ask him about that [18:49:39] monobook used borders and since vector is based on monobook, there should be no issues [18:50:13] That is a hilariously untrue statement [18:50:39] Because it assumes that Vector is based on Monobook more than it really is and that Monobook had no issues [18:50:40] vector is not based on monobook? [18:50:45] To a degree, yes