Disclaimer: I work for or provide services to the Wikimedia Foundation. However, the Foundation does not vet all my activity, so edits, statements, or other contributions made by this account may not reflect the views of the Foundation.
Mon, Feb 17
Thu, Feb 13
There's questions regarding whether we should include any sort of "this comment was edited" notice on an edited comment. Meeting discussion leans towards "no", in part because there's no good way to link to the revision history showing the change without making a second edit to add that link in.
Wed, Feb 12
@Mayakp.wiki I've made the very small required change to the schema here: https://meta.wikimedia.org/w/index.php?title=Schema:EditAttemptStep&type=revision&diff=19802113&oldid=19486181
Tue, Feb 11
I did some research into hiding tags: it hides the tag entirely if the associated message is present-but-disabled. (If there's no message at all, it just shows whatever the tag name is.) Messages are disabled by setting them to literally the empty-string or to -.
My theory is that @Ryasmeen has NWE enabled, and so "edit source" isn't technically a navigation event (because it's just replacing the contents of the current page). We need to special case detecting and canceling this somehow.
Mon, Feb 10
But Maya and I have spoken about it, and we know what needs to happen.
Fri, Feb 7
@Esanders: I'm actually not sure what's going on there with 1. For me, locally, exact matches are working just fine. See:
Thu, Feb 6
This is pretty tight, so we should some a top margin (0.5em?) as seen in Phab:
Mm, I do agree. I've updated the patch with a top margin 0.5ex, which looked better to me than 0.5em:
Wed, Feb 5
Why is it that a reply made using DiscussionTools in visual mode would receive three tags instead of two?
My initial logic was just that it'd be simpler for search, rather than anyone interested in the data having to know up front what all the possible modes were. That said, I can certainly remove the pure discussiontools tag, and you can remember when searching what the options are.
Fri, Jan 24
How – if at all – should the text input mode (e.g. source vs. rich text) be tracked?
Thu, Jan 23
EditAttemptStep needs to be tweaked a little so it can apply to DiscussionTools being used to edit a page. I think what's needed is:
Wed, Jan 22
I'm working on a patch for this which will apply placeholder tags for likely edit types. The bulk of the actual technical work is adjusting the VisualEditorEdit API to allow the tags to be customized, anyway.
Tue, Jan 21
This is more work to audit than I have time to handle. The difficulty is mostly that any random wiki could have a gadget / extension which adds a message on arbitrary criteria... and even for just the core / extension code, there's enough different ways to trigger something like this that it's a real pain to search for it happening.
Jan 9 2020
I do find @ssastry's heredoc argument on the grounds of "it's already being implemented for similar purposes elsewhere" to be quite compelling. I also agree on the useful clarity of it being a fairly generic thing that can be used to bypass wikitext newline behavior.
That's the experience with a margin added so that the table is pushed in by a margin the size of the side context items. I can put it up on the prototype if you'd like to try it personally.
Jan 7 2020
This is a question which is lightly holding up this ticket in code review: https://phabricator.wikimedia.org/T229839
Dec 20 2019
It's happening to other edits as well, so restbase is just down in general on beta, not in a talk-specific manner. E.g. https://en.wikipedia.beta.wmflabs.org/api/rest_v1/page/html/Cats/407373?redirect=false&stash=true
Dec 16 2019
Server-side PHP sends EditAttemptStep events, right? And those are limited to action='init' & action='saveSuccess'?
@mpopov Yeah, that's my logic. It does also do saveAttempt and saveFailure server-side. Perhaps checking for saveAttempt would be most inclusive, since that'll always fire.
Oh, and on the debian-10.0-buster (deprecated 2019-12-15) image, apparmor is installed as apparmor/stable,now 2.13.2-10 amd64 [installed,automatic]. Automatic implies it was pulled in as a dependency, I think, so perhaps some part of debian (lxc?) stopped explicitly depending on apparmor in the last few weeks?
I also verified that a working vagrant VPS set up on the debian-10.0-buster (deprecated 2019-12-15) image does have apparmor installed. I can't say whether that's the image or some puppet change to the role::labs::mediawiki_vagrant role that has occurred in the last few weeks.
Dec 14 2019
Define “edits done” — do you want all visits to the edit page, or can we restrict it to successfully completed edits? If the latter, I think we could use analysis of current EditAttemptStep data to compare sessions that are just made of ‘init’ and ‘saveSuccess’ to ones that also include ‘ready’/‘firstChange’.
Dec 13 2019
Dec 12 2019
Is it just that list item pr everything else on the page beyond the list item? Reading the title and description, I got the impression it is just the table in that list item and hence is limited.
I'm not convinced this should be invalid -- it's a pretty bad page-corruption issue if bad wikitext goes through Parsoid.
Dec 10 2019
Ah, hold on, I think you already fixed that in https://gerrit.wikimedia.org/r/#/c/mediawiki/skins/Vector/+/555626/ and I was just using an unfortunately timed checkout on the VPS. So ignore me. :D
https://gerrit.wikimedia.org/r/#/c/mediawiki/skins/Vector/+/554284/ seems to have caused some issues if you're using non-default languages. On a fairly stock mediawiki install, if I use the header selector to change my language to en-GB, I see this:
Maybe a good first step would be to make a new server and see there's any problems getting vagrant going on it, then we can move to migrating needed test data to it if that works out?
Dec 5 2019
I think this is ultimately a duplicate of T158617?
This is definitely still an issue, and I just had to work around it while setting up a cloud VPS.
Dec 2 2019
I've raised this question before and not seen an answer: what edit_count data is being used to assign people to the junior/senior buckets? Is it assigned at time-of-posting or time-of-analysis? Are we counting all edits or just article-namespace edits towards this? This could make a notable difference if we, say, made it much easier for people to post a lot of comments quickly...
Nov 22 2019
Jess and I were talking about this and we agreed to try out adding a slight margin to tables on mobile only when they have focus, to see how that affects the experience.
Nov 14 2019
That last patch is a not-to-be-merged-but-useful-to-exist-for-testing one. The actual username completion will exist over in DiscussionTools, but that's not actually in a state where it's easy to use yet.
Nov 7 2019
I don't actually know anything about the Graph extension, but a quick glance at the code suggests that it's trying to load a graph-spec from a page that doesn't exist, and doesn't have a sensible check for that failure case.
Nov 6 2019
I'm interested in the ORES scores over reversion-rates as well, mostly because it lets us avoid all sorts of confusion about what counts/should-count as a revert, and side-effect like the mentioned "what if we overwhelm the people who'd be doing the reversions with a flood of content".
Will the prototype servers really require too much work to migrate? They're running mediawiki-vagrant, as I recall, so as long as it's in a currently viable state for installs we should be able to spin a new one up from scratch.
On the RESTBase side, the most numerous recorded cases were the ones where VE does not provide an If-Match header during transforms - there were 1122 such occurrences in the last 7 days, so I think we need to prioritise finding out how and why that happens.
Nov 5 2019
Thanks for looking into it! I'll just mention that, the schema can be changed to make the fields optional, which will then give you proper metrics showing how often this condition occurs.
That patch implements the strategy I described. Adding $wgVisualEditorRestbaseParsoidVariant = 'php'; to the config file triggers the appropriate header-sending.
I'm thinking that we can't just push it everywhere, because that'd break any non-foundation user of VE who hasn't configured restbase to know about that variant. (E.g. my updated-yesterday vagrant install falls over and breaks as soon as I start sending that header.)
Nov 4 2019
It's not just loaded_timing, there's also ready_timing.
Oct 31 2019
For now I went with making a puppet/hieradata/local.yaml file and including npm::node_version: 10 within it.
This is probably a regression of T218238, which indicated that it was going to be making eventgate optional for cases which don't strictly require it.
Oct 29 2019
I want to second @Jdrewniak's point -- if we can refactor skins first so that skins really are a lighter presentation layer on top of a sensible base page structure, then whatever we choose for the rest of this doesn't matter so much. If we're not going to do that, there's a minefield ahead.
Oct 15 2019
Sorry, just to be explicit:
When does the 'dialog-review' and 'dialog-preview' get fired ?
When you click the "review your changes" or "show preview" buttons that are at the bottom of the save dialog.
During testing I also found a bunch of other actions being stored under the mwSave feature. We can ignore them if required as the number of times these actions appear is very very small..but it would help to know if its a bug:
Those are all probably interactions with other code. The logging we do just does logging based on the ids that are present in the code. If other extensions / gadgets / userscripts insert things, they'll get included in the logging. wpReviewEdit is from the FlaggedRevs extension, for instance.
These are all from the last patch to that ticket, yes.
Oct 8 2019
Safari 13 has added Visual Viewport support, which can plausibly help with this.
Oct 3 2019
That said, the "Yes, but it's getting fired twice" shouldn't be firing twice. But I can't reproduce that myself, so I might need more information.
Everything Rummana saw looks correct, it's just a confusion about what the event name is. For the VisualEditorFeatureUse events it's two-part, there's a feature and action component, with the name being written here as [feature].[action].
Sep 25 2019
It should all be mwSave as the feature and then the dialog-whatever as the action.
Sep 24 2019
There must be some Etag manipulation going on on the client side (either in the PHP side of JS of the VE code) that alters it in some instances or removes it in other, given that we have evidence both cases mentioned in the description happen regularly (albeit with different frequencies).
Sep 23 2019
For what it's worth, I can't find anything in VE which ever alters an ETag. I.e. the weird suffix case should just be us passing whatever Restbase gave us when we loaded the article back to it.
Sep 20 2019
Sep 18 2019
T232790 is basically the QA process, after all.
This one was a pain to track, but it's a caught-and-swallowed exception inside TreeModifier. (Swallowed because it occurs in a jQuery promise, where exceptions are discarded.)
Here's the stack trace:
Sep 17 2019
Buckets are all assigned server-side, yes -- I posted a link to where it's done above.
Sep 11 2019
@ppelberg They're assigned to the buckets server-side. They then log their bucket as the assigned one, regardless of which editor loads.
Sep 10 2019
Okay, current patches implement firstChange in: desktop VisualEditor, desktop WikiEditor, MobileFrontend wikitext and visual.
We should also write patches for those editors.
Okay! That cluster of patches implements the firstTransaction step on desktop and mobile. It logs it with the timing as "time since the document was ready".
I have updated the schema for now, with time-since-ready: https://meta.wikimedia.org/w/index.php?title=Schema:EditAttemptStep&diff=19364181&oldid=19273145
@MNeisler - Sorry, I didn't think to ask -- for the timing, would it be more helpful if I stored time-since-initialization or time-since-ready? That'd sort of be whether this event is conceptually grouped with the editor-loading cluster or the saving cluster.
Sep 9 2019
From the patch where we started discussing this:
When a contributors taps...
|“<” in “Save your changes” dialog||added in patch: mwSave.dialog-abort|
|“<” in “Review your changes” view||added in patch: mwSave.dialog-review|
|to "Preview"/"Review" their changes||added in patch: mwSave.dialog-review/preview|
|"Visual" or "Wikitext" in “Review your changes” view||added in patch: mwSave.review-switch-source / mwSave.review-switch-visual|
|the "This is a minor edit" checkbox in the Save dialog||added in patch: mwSave.checkbox-wpMinoredit|
|the "Watch this page" checkbox in the Save dialog||added in patch: mwSave.checkbox-wpWatchthis|
|the “>” in v1 of the toolbar||already present in EditAttemptStep as saveIntent|
@Neil_P._Quinn_WMF / @MNeisler : Given that the "starts an edit" part requires a change to the EditAttemptStep schema, could you verify that all I'd need to do is add, say, "firstTransaction" to the action enum, and add a first_transaction_timing integer? No other extra data required?
Sep 7 2019
For the code-side generation of this distribution, this is where it's defined: https://github.com/wikimedia/mediawiki-extensions-MobileFrontend/blob/master/includes/MobileFrontendEditorHooks.php#L48
Sep 3 2019
Aug 31 2019
Aug 30 2019
@Ryasmeen: for testing, the most likely side effect of this patch would be somehow breaking the ability to use the save shortcut again while the save dialog is open to trigger a save.
That patch is the most-limited version of this: it's just on mobile, and it's also just the mediawiki link inspector. If greater scope is desired, I can adjust this.
Actually, more extensive question: should this be just the link modal, or should this be all inspectors? If we're invoking a consistent visual language, after all...
Is this a mobile-only change?
Aug 29 2019
Aug 27 2019
After some research, it's a side-effect of a89c79dc7907fb32707c0fe535230cc04c38cae5 rather than being something tied to the save dialog actions -- intent is that if an accesskey hits the save-button while the save dialog is open, it triggers a save.
Aug 22 2019
Educational DVDs can get them too! Though I admit I didn’t bother to look up whether there’s any distinguishing characteristic in the code.