Wikipedia:Bots/Requests for approval/28bot

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 Approved.

28bot

• (BRFA
• Approved BRFAs
• talk
• contribs
• count
• SUL
• logs
• page moves
• block log
• rights log
• flag)

Operator: 28bytes (talk · contribs)

Time filed: 01:48, Thursday March 31, 2011 (UTC)

Automatic or Manually assisted: Automatic

Programming language(s): Python

Source code available: Not yet; still in development

Function overview: Remove text inserted in articles by accidentally clicking on edit bar.

Links to relevant discussions (where appropriate): Edit filter discussion

Edit period(s): Continuous

Estimated number of pages affected: 500-1000 a day, judging by this

Exclusion compliant (Y/N): N/A - article space edits only

Already has a bot flag (Y/N): N

Function details: This bot would remove accidental edits of the type previously flagged by edit filter 18. These are the literal strings '''Bold text''', ''Italic text'', [[Link title]], [http://www.example.com link title], == Headline text == and similar strings listed in the now-disabled edit filter. Requests have been made to re-enable the edit filter, but after testing the filter I believe a bot is preferable, as the edit filter uses too many resources, and having an edit filter prevent the edit (rather than simply stripping out the accidentally inserted markup after the edit is made) would annoy and confuse new users. The bot would remove the accidentally inserted text without removing the legitimate changes (if any) made by the editor, and would only operate in article space.

Discussion

What evidence is there that it would help to leave "good" edits? I have occasionally reverted someone who has made one of these kinds of edits, and I am pretty sure that in all cases I reverted their other edits as well. If a bot removes just the blatantly obvious crud, that would appear to other page checkers that something has "approved" the current content, so reducing the tendency for other edits to be checked. For example, if an editor adds obvious nonsense in one edit to a BLP, while changing the birth date without reason in another edit, any commonsense editor would revert both edits as "unsourced" or whatever. Johnuniq (talk) 03:30, 31 March 2011 (UTC)

You make a good point. I manually cleaned up a few hundred of these myself, and in almost all cases the "non-accidental" part of the edit was better off not having been made. I'd be just as happy to have the bot undo the entire edit (or edits, as XLinkBot does) and possibly leave a friendly note on the editor's talk page explaining why, if consensus here is that that approach would be better. 28bytes (talk) 04:51, 31 March 2011 (UTC)

Needs wider discussion. Few people watch these BRFAs, so you are best to advertise this somewhere else (e.g. VP). In essence, reverting the whole edit could revert a good edit; reverting only the problematic part could hide the possible bad parts from editors. —  HELLKNOWZ  ▎TALK 12:09, 3 April 2011 (UTC)

Fair enough, I will start a discussion there. 28bytes (talk) 15:20, 3 April 2011 (UTC)

I have started a discussion at the Village Pump here. 28bytes (talk) 17:18, 5 April 2011 (UTC)

Seems like consensus is pretty clear to implement the simple case portion of the bot. Maybe you can run that for a bit, then return and request to add the complex if there is consensus (seeing your bot perform effectively may change the minds of some)? Regards, MacMed^talk_stalk 20:07, 9 April 2011 (UTC)

Agreed, and I'll elaborate below. 28bytes (talk) 20:25, 9 April 2011 (UTC)

From VP discussion. Approved for trial (50 edits). Please provide a link to the relevant contributions and/or diffs when the trial is complete. For cases with one, several, or combinations of the edit bar default strings without any other changes (except whitespace). Additionally, tag user pages with {{uw-test1}} or higher where appropriate by the usual WP:UWUL user message guidelines. —  HELLKNOWZ  ▎TALK 20:21, 9 April 2011 (UTC)

Thanks. I will implement it this way, and if there are no objections, I'd also like to have it create a subpage in the User:28bot/ space that will track what it would have done with complex cases so that a future discussion can use that as a data point for deciding whether to expand the functionality. 28bytes (talk) 20:25, 9 April 2011 (UTC)

I don't see any harm in that. Makes sense. Regards, MacMed^talk_stalk 20:29, 9 April 2011 (UTC)

Go ahead, some empirical evidence would be nice. —  HELLKNOWZ  ▎TALK 20:30, 9 April 2011 (UTC)

Is there something wrong with filter 18? It's not matching anything at the moment, and the bot hasn't made any edits. It seems you may have disabled it here. Also, why not just do the whole thing with a slightly more refined edit filter? - Kingpin¹³ (talk) 15:57, 23 April 2011 (UTC)

The filter had been disabled since 2009[1]; I re-enabled it for about 3 minutes in March (see your diff) before deciding to pursue the bot route. I thought a bot would be better suited to the task since a filter would be unable to perform selective/partial reverts as proposed in the VP discussion. The bot doesn't have any edits yet as I am still coding it. 28bytes (talk) 00:15, 25 April 2011 (UTC)

Still coding this? What's the status of this request - Kingpin¹³ (talk) 14:17, 22 May 2011 (UTC)

Yes, but I hope to have it up and running the trial edits this week. 28bytes (talk) 21:45, 22 May 2011 (UTC)

The bot's up and running now. Question: for the 50-edit trial, is that 50 mainspace edits, or 50 mainspace + user talk edits? Or total edits? It's over 50 total edits now, but most of that is just it logging to User:28bot/edit-tests-found the pages it found that have "mixed" test edits. 28bytes (talk) 06:29, 6 June 2011 (UTC)

It's not a strict number. In this case revert+notification would be 1 "unit of edit". So just do 50 or so reverts. —  HELLKNOWZ  ▎TALK 06:35, 6 June 2011 (UTC)

Thanks. I'll stop back by when it hits that. Not a lot of people doing "pure" edit tests this weekend, it seems... 28bytes (talk) 06:36, 6 June 2011 (UTC)

Also here is an example of the bot hiding a possibly bad edit: [2]. This revision remained. I wonder if reverting all recent edits by the user would be better? —  HELLKNOWZ  ▎TALK 06:39, 6 June 2011 (UTC)

Yeah, I noticed that. Reverting all the edits (a la XLinkBot) would be my preference, but multiple consecutive edits by the same editor didn't really get brought up in the VP discussion, so I'm not sure what to do about that. 28bytes (talk) 06:55, 6 June 2011 (UTC)

Trial complete. OK, it's just passed 50 removals of edit tests in the mainspace. I've checked each one of them, and they all appear to be correct. 28bytes (talk) 15:00, 17 June 2011 (UTC)

 Approved. I still have some concerns on how it will handle sneaky/complex vandalism, but since it replicates a filter, I'm ok. MBisanz ^talk 14:19, 24 June 2011 (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.