@jrieken Looks like something changed in the filtering logic of suggestions. Your recent change around the inComplete:true maybe?

The emmet completion provider returns a completion list with inComplete set to true.
In stable, every letter typed would call the completion provider, but : would just filter existing results.

Stable:

In the current Insiders, every letter typed still calls the completion provider, but : filters out everything

Current Insiders:

Yes, I was curious in the first moment why it stopped working.

Oh, bummer... That's a regression from a change I made yesterday to fix #28400. I'll might revert that change...

Sorry to bother you. :/

Hm, reverting 5ad1618 (which is the fix for #28400) doesn't help. @ramya-rao-a It seems that : is registered as a trigger character and when we see that we invoke the provider in question and merge its results with the existing result. No, the problem is that that provider doesn't return anything and that the existing results don't match anymore because we compare html: with html etc. Needs more drilling...

@jrieken Looks like something changed in the filtering logic of suggestions. Your recent change around the inComplete:true maybe?

So, we did make a change yesterday (I think too early for the version this was reported for) which does the following: When a suggestion result is incomplete and if the word as stopped we don't re-trigger suggestions but filter the suggestions we currently have. So, when typing html: the word ends and we stop asking for new suggestions (that might be a problem with expansions form html:foo"bar but I am unsure that exists). However, since the results already contain html:5 it will be shown and can be selected.

I wonder if this was caused by having another provider that registers with the : trigger character (I believe 7ba1dcf). So what happens, and I don't think much has changed there, is that hitting a trigger-character only asks that provider (which doesn't return a result). Its (empty) result is being merged with the existing completions (from the incomplete provider) which didn't get a chance to update and hence fail the filter condition.

In short, there is something off. I am not yet convinced that a recent change caused this but things are complex. I acknowledge that there is a race between incomplete suggestions and trigger-characters, however I am a little scared to make a change late in the endgame.

Ok, the competing provider is the html-extension that registers those completion item provider: https://github.com/Microsoft/vscode/blob/master/extensions/html/server/src/htmlServerMain.ts#L75. Still not sure what as changed but I'll think about a fix

@ramya-rao-a I do have a change ready that makes sure to also resolve incomplete suggest results when a trigger character is being hit. However, when I ask emmet for completions with the html: prefix I just get that back, so no html:5. Fixing this bug uncovers another bug...

I believe that this used to work by accident (more than one thing is off here) and that's because before the completion item would have a just the label html:5 and now it has a filterText which ishtml. The filter text, when provided, must match, but it doesn't because it's shorter than html:.

Final word: There is something off when resolving incomplete suggest results when a trigger character is being hit, changing that is too risky IMO, and reveals other bugs in emmet (e.g. hit Ctrl+Space after html: in any version). So I'd opt for removing/fixing the filter text and things should be back to normal.

Thanks for all the digging @jrieken
I agree its too late in the end game for risky fixes.

I believe that this used to work by accident

I have added : as trigger character for emmet completions and updated the completion provider to return expected items.
So after 42ae1fc and microsoft/vscode-emmet-helper@5875048, we get expected outcome.

Still an issue for abbreviations like ul*2
The suggestions disappear when * is typed.

Re-opening

Yeah, whenever we hit a word stops with incomplete results and when the result isn't matching that we cancel. That was done in 5ad1618. Feel free to revert and reopen the corresponding issue. I am on vacation now... 🏖

Happy holidays @jrieken 😃

I add the * and $ to trigger char list as well. So we are good on those fronts.

Note to verifier:

• Try abbreivations with * $ : in your abbreviations and check you dont lose the suggestion list
• Refer to https://docs.emmet.io/cheat-sheet/
• Few things from the cheat sheet is not working due to upstream issue. See Few of the abbreviations from the cheatsheet don't work emmetio/atom-plugin#27

@ramya-rao-a

Just tested emmet suggestions with : and * in the latest insider build and there work fine. $ also works in my php file but I am not sure if that's coming from emmet or the language server.

Thank you so much for your great work.

Thanks for all the feedback @nanaasumadu and @jens1o