Wikipedia:Bots/Requests for approval/VoxelBot 2


 * The following discussion is an archived debate. Please do not modify it. To request review of this BRFA, please start a new section at WT:BRFA. The result of the discussion was Symbol keep vote.svg Approved

VoxelBot 2
Operator:

Time filed: 02:56, Thursday January 31, 2013 (UTC)

Automatic, Supervised, or Manual: Supervised

Programming language(s): AutoWikiBrowser, Python

Source code available: AWB, Standard pywikipedia

Function overview: Replace substitutes for Romanian letters used before Unicode 3 was released (Ş, ş, Ţ, and ţ) with their proper letters in the Romanian alphabet: (Ș, ș, Ț, and ț). It will replace everywhere except for image links and interwiki/external links.

Links to relevant discussions (where appropriate): Bot requests/Archive 52

Edit period(s): One time run

Estimated number of pages affected: Hundreds of thousands. Working off the Geography of Romania Category

Exclusion compliant (Yes/No): Yes

Already has a bot flag (Yes/No): Yes

Function details: From a database scan of pages with the characters in their titles list of pages in categories related to Romanian Geography, move them to their correct name if they contain incorrect characters. Then, replace the characters defined above with their correct letters in the Romanian alphabet and fix the double redirects created.. The input pages will only be strictly Romanian, since the current letters are correct in non-Romanian languages. The input pages won't just be taken from the base category (Geography of Romania) but will be sub-categories recursed by AWB that are manually checked.

Discussion

 * Will it do any other kinds of gen-fixes when making these edits? Will most users be able to see the difference in the letters or will it be purely an underlying language code change?  MBisanz  talk 03:23, 31 January 2013 (UTC)
 * In some manual test edits ( for example) you can see that it does make a visible change. It is set up to also do genfixes (of course only if it matches the regex as well), but this can be disabled as you know. It's already editing so genfixes (which many of these pages need) are quite good in this case. Vaca  tion  9  03:42, 31 January 2013 (UTC)

I believe this must be supervised, not automatic. --MZMcBride (talk) 03:52, 31 January 2013 (UTC)
 * What exactly do you mean by supervised? If you mean checking the edits as they are made, this is completely possible. Vaca  tion  9  04:08, 31 January 2013 (UTC)
 * Yes, I mean checking the edits. "Supervised" instead of "automatic" (which is the equivalent to "unsupervised"). I think human review will be needed for each of these edits, as find and replace is notoriously tricky on any large body of work. The number of edge cases is simply overwhelming. --MZMcBride (talk) 18:54, 31 January 2013 (UTC)
 * That's completely possible, and I've noted the bot as Supervised instead of Automatic. Vaca  tion  9  22:28, 31 January 2013 (UTC)


 * Another problem has arisen: we have to move pages with incorrect characters in them as well. Thus, I came up with a three-step plan. First, a database scan for articles with incorrect characters in them, which outputs to a file. Then, (after review of the articles) using Pywikipediabot or a similar framework I mass move the pages. Then, when doing the AWB scan, we can both correct the redlinks and correct everything else in the page. This will need some interesting code but I think it would be worth it. Thoughts? Vaca  tion  9  13:34, 31 January 2013 (UTC)
 * We can actually just use the API to find the list of the pages in the Category, then use our code to see if the title needs to be moved, and if it does then move it. Vaca  tion  9  13:45, 31 January 2013 (UTC)
 * Any comments? Vaca  tion  9  03:10, 7 February 2013 (UTC)
 * Sorry, I've had a cold and been otherwise busy in real life. Are you ready for a trial?  MBisanz  talk 03:09, 9 February 2013 (UTC)
 * It might be a couple days before the trial starts, but if we're approved for trial that would be great. I feel for you; managing both BAG and CHU? :O Vacation nine Public 13:40, 9 February 2013 (UTC)
 * sure thing. Haha it's what I signed up for.  MBisanz  talk 17:22, 9 February 2013 (UTC)
 * Alright, I've got the page mover set up (just need to implement the WhatLinksHere checker which will change the link from a double redirect). I can't test the AWB part until you or another admin adds VoxelBot to the AWB Bots approval list. Still trying to figure out the regular expressions here; for some reason even if I select the "Skip external/internal links, etc" option in the RegEx menu, it still edits links? Which is obviously not the behavior needed here. Any ideas? I saw somewhere about an option for this but I've searched AWB far and wide and found nothing. Vaca  tion  9  13:41, 10 February 2013 (UTC)
 * I have added VoxelBot to the awblist here. What regular expression are you using and what do you desire it to do? :)  ·Add§hore·  T alk T o M e ! 13:44, 10 February 2013 (UTC)
 * It's just a simple find and replace; not even a regular expression persay. It's replacing the incorrect characters with the correct ones. But of course this shouldn't apply in links as it would break them. Any ideas? I don know whythe ignoe comments, links, etc. option is not working. Vacation nine Public 13:49, 10 February 2013 (UTC)

Completed a total of 50 edits in both page moves and AWB replacements. The reason it took so long is because I was dealing with some problems with AWB not skipping links along with unicode errors. I have got everything running smoothly now however. Here is my planned workflow: 1. Using a python script I created, get a list of pages from categories (manually reviewed categories) relating to Romanian geography. 2. Run these pages in another python script I created which takes the page names in and performs a page move with the correct characters. Sample moves using this script are here:. 3. In the same script, VoxelBot automatically, using Backlinks (like WhatLinksHere but through the API), corrects links to the page. Examples are here:. 4. Using AWB (thanks Addshore ), from a list of manually reviewed categories like before, replace all other instances of the incorrect characters with correct ones (ignoring external/internal links and templates, but not notes like before (using a custom regex)). Examples:. General fixes are run as well. All of the steps run by script or AWB will be supervised, but not manual. Vaca tion  9  22:15, 13 February 2013 (UTC)
 * All looks good  ·Add§hore·  Talk To Me! 22:08, 13 February 2013 (UTC)
 * The above discussion is preserved as an archive of the debate. Please do not modify it. To request review of this BRFA, please start a new section at WT:BRFA.