Wikipedia:Bots/Requests for approval/BU RoBOT 11

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

Withdrawn by operator.

BU RoBOT 11

Operator: BU Rob13 (talk · contribs · SUL · edit count · logs · page moves · block log · rights log · ANI search)

Time filed: 04:25, Monday, April 11, 2016 (UTC)

Automatic, Supervised, or Manual: Automatic

Programming language(s): AWB

Source code available: AWB

Function overview: Substitute transclusions of a template to complete merges or carry out other consensus decisions at WP:TFD

Links to relevant discussions (where appropriate): Varies by template, but substitution will always follow a close as "merge", "substitute and delete", or similar at WP:TFD

Edit period(s): Multiple runs, based on need at WP:TFD/H and future TfD discussions

Estimated number of pages affected: Depends entirely on the template(s) involved. Could be anywhere from a couple hundred to tens of thousands in a single template.

Exclusion compliant (Yes/No): Yes

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

Function details: Implementing consensus at TfD discussions often require semi-automated or automated work to merge/substitute templates. I've filed BRFAs in the past for simple substitution of wrappers to implement the consensus of various TfD discussions. See Task 8 for the most recent (including an example of the regex that would be used for this task). I believe there were others in the past as well. I'm hoping to get permission to substitute templates in that way more broadly, mostly to avoid clogging up BRFA with non-controversial bot tasks that are technically trivial. {{About3}} is currently pending substitution, and I'll use that for any trial if necessary (although you could also treat Task 8 as a large trial, which had no issues whatsoever).

I want to be very clear on what this task is not. This task will not be construed to be approval for anything involving adding/removing/changing/editing of parameters within a template using regex. It will not be used to invoke modules, as was the case in Task 2, since the use of a module is an extra layer of technical complexity that warrants scrutiny. It will only substitute transclusions of templates to implement a TfD consensus.

Discussion

To demonstrate need for this, {{Unicode}} needs substituting now, too (over 50,000 transclusions). ~ Rob^Talk 13:21, 12 April 2016 (UTC)[reply]
- I note that, if simple substitution is what you're after, you could just put the template into Category:Wikipedia templates to be automatically substituted and add it to User:AnomieBOT/TemplateSubster force. Anomie⚔ 18:45, 12 April 2016 (UTC)[reply]
  - @Anomie: Hmm, didn't even know that was a feature. Still, I think it's best to get this task approved because there may be situations where I need to substitute only in certain situations (i.e. if a template doesn't use parameter X, then substitute. Otherwise, do not and humans will come through later and take care of it.) Also, how would your bot handle a situation where 50,000+ transclusions need substituting? Would it "crowd-out" the type of work your bot usually does? ~ Rob^Talk 19:03, 12 April 2016 (UTC)[reply]
    - It shouldn't crowd out other work. Even the tasks on the same task runner shouldn't be affected too badly, since I usually code the jobs to yield after running for about 5 minutes. Anomie ⚔ 19:22, 12 April 2016 (UTC)[reply]

This is asking for blanket approval, however the most recent bot trial on this account still contains errors that could be caught with dry-runs (e.g., this trial had this bug twice within the first 10 edits) 10 edits is the absolute bare-minimum to check for unforeseen edge cases before allowing a bot to make edits. If we're being asked here to grant blanket approval, I'd only be comfortable supporting such a thing when at least a couple of conditions are met:

Trial safety — That is, it's clear the owner's proactive with testing and debugging their own code before "putting it on production." This means there's dry-run testing and/or manually-approved-edits happening, and that there's no obvious evidence to the contrary. A reasonably sized sample (basically think 25-50 edits—the same as most official bot trials) before allowing the bot to edit unsupervised is a decent chunk.
BRFAs by the owner develop a clear trend of being low-maintenance — you're essentially saying "I'm confident that I can bypass BAG, who'd otherwise check that the bot's going to run reasonably safely and within the bounds of site policies and guidelines, because I've taken clear steps and have sufficient wisdom to avoid needing them." There needs to be a clear trend of evidence to support that.

I'm concerned these are not yet met. For example, the most recent bot trial on this account contained errors that clearly would have been caught with a reasonable supervised run (i.e., this trial had this bug twice within the first 10 edits). While I'm fully aware that bugs can happen even after extensive testing, I feel this is evidence toward the lack thereof and therefore evidence for a continued need to involve BAG in testing before edits are made live—at least until a few more BRFAs show otherwise. In the grand scheme of things, it's not a huge bug, but it's a reflection of bug-detecting/bug-prevention methodology, which is at the core of a blanket approval.

--slakr^\ talk / 05:22, 16 April 2016 (UTC)[reply]

@Slakr: The regex that I'm requesting to use here does not change, however, and I've published it in Task 8. The only thing that changes is the template name. I'm asking for blanket approval to use that exact same regex with zero modifications to its structure on more templates. There were zero errors in that bot run (and its trial). The only situation in which I might use different regex is to restrict the sample, but that doesn't change the edits on the page, only what pages are used. If a mistake is made there, it merely results in less edits. Ultimately, it depends how BAG wants to handle things. I'm happy to spam trivial tasks that all look the same here, but it's just going to crowd up this process even more.

As a side note, I think we have different opinions on what trials are. I consider the trial the supervised run. I directly supervise the first 10 edits or so and then, if all is well, let the next 40 run and check them immediately afterward. This is mostly because I don't care to have the bot edits on my account and I can't, under bot policy, do a supervised run on my bot account before the trial. Obviously, I would check the first edits made by my bot when I swapped out template names, although it really doesn't change anything substantive. Better safe than sorry. ~ Rob^Talk 12:49, 16 April 2016 (UTC)[reply]

By the way, if blanket approval for these simple substitutions is not possible, please approve a trial here for just substituting {{Unicode}} in this BRFA to save us all at least a little time. ~ Rob^Talk 15:29, 18 April 2016 (UTC)[reply]

{{BAG assistance needed}} ~ Rob^Talk 06:41, 22 April 2016 (UTC)[reply]

Withdrawn by operator. Just going to submit BRFAs individually as these tasks come up in the future. ~ Rob^Talk 01:50, 30 April 2016 (UTC)[reply]

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.