Yoink! I'm making a check for uncaptioned images.

I think my main feedback so far is that includes/ResourceLoaderData.php is hard to find, but it is necessary to specifically set variables such as minimumCharacters for functions such as mw.editcheck.BaseEditCheck.prototype.getModifiedContentRanges() to work properly.

I've been rebasing onto the new multicheck code, and discovering that I didn't understand the codebase as well as I thought. Progress is happening!

Further update on the experience of creating a check:

• Functions that depend on includes/ResourceLoaderData.php values need defaults. Possibly we could add explicit comments in existing checks about where these are set, seeing as they're examples as well as features
• Likewise, it's probably worth noting that a new check needs adding to extension.json and editcheck/modules/init.js
• It's fairly straightforward to understand that one should create an onDocumentChange or onBeforeSave function which returns a list of EditCheck objects
• Although I'm already slightly familiar with the document model, understanding how to manipulate it well took most of the time:
  • How to get nodes of a particular type
  • How ranges relate to selections, and how selections relate to fragments
  • Range manipulation, such as determining whether a node was fully within an edited range, knowing whether to use outer range
  • I didn't end up navigating the tree much for my check, but I did in the process of building it. Some of this comes down to simply understanding ve's data model
• Having correctly identified the targets of my check, it's time to act()
  • It was a useful sidequest to understand how selections and context works, and it's certainly a possibility that giving users boilerplate code or a helper function to activate a sensible context action might make life a lot easier
  • I had to go and re-learn how windows work, eg that they often exist in a closed state and need to be opened with instance-specific data. I'm a little fastidious here, other developers will likely copy code and get working code but I wanted to be sure I was getting things right. It was useful to find out how to pass the fragment so that the media dialog would open in the correct mode
  • It's probably useful to also document how you might present options other than accept and reject in an action dialog
  • I didn't want to implement a dismiss function because I knew that there was going to be a more generic one coming down the pipe.

Edit: sorry for the edit to create the second half of the comment, I shift-entered when I meant to enter

Here's a patchdemo: https://patchdemo.wmcloud.org/wikis/dd1e6b2bec/wiki/Main_Page

Change #1083873 had a related patch set uploaded (by Zoe; author: Zoe):

[mediawiki/extensions/VisualEditor@master] WIP: Add an edit check for uncaptioned images

https://gerrit.wikimedia.org/r/1083873

Next Step
Per today's team meeting, Editing Engineering will review and respond T374066#10337785 seeing as how the Edit Check API has evolved since this feedback was shared.