Make tests faster & better (use completions API) #103
Conversation
|
Also I think there is something wrong with end of lines were in changed files. Currenty end_of_line in editorconfig is set to crlf (I don't remember why) maybe it should be set to |
|
I'll definitely take a look on that as this is a pain in the ass to wait for those tests so long. |
|
Intention of this PR is to eliminate delay calls and make things as fast as possible, I will probably include other test improvements here. First part seems okay as I can see on ci. |
zardoy
changed the title
| @@ -2,7 +2,7 @@ import * as vsc from 'vscode' | |||
| import * as ts from 'typescript' | |||
| import { invertExpression } from './utils/invert-expression' | |||
|
|
|||
| export const NOT_COMMAND = 'complete.notTemplate' | |||
| export const NOT_COMMAND = '_postfix.applyNotTemplate' | |||
There was a problem hiding this comment.
To be honest I have an idea to drop this command at all and replace this with multiple versions of the template, so instead of picking not template and then selecting from quick pick one would see not#1, not#2 (or sth similar)
This would not only help with those tests here but also make it more friendly to the user in my opinion + less clicks to apply the template.
There was a problem hiding this comment.
Interesting idea. While it can make sense in case when I need to apply not on identifier (identifier not will always go first and it takes less clicks), but when I need to always revert whole binary expression it might be painful for users that have:
- Other templates starting with
notegnotZwill be last - For js only: word-suggestions
It will be hard to apply last variant, so I don't see any advantages of it (for example web storm also provides selector as second step).
But we definitely need to add at selector (0 or -1) support for not template. So I can have postfixes that always revert either identifier or whole binary epxression.
BTW I renamed it here so its not shown in keybindings.json command autocomplte and other locations as commands starting with _ are internal.
There was a problem hiding this comment.
Ok, for complex binary expressions that could be tricky, but who does more nesting that just several levels on the other hand? :)
Anyway even if I decide to pursue this way this should probably be optional as some people might prefer current way.
|
thx to github, got notification of commit above. I don't think we need to bump engine until actually new API is used within extension (it is used outside, in our tests) |
True that but maybe it's time, I haven't bumped it for a while and I'm pretty sure 95%+ users are updating on regular basis so this should be just fine especially that 1.72 is from September 2022. |
I tested on singleline tests only: on my macbook m1 its reduced time 58s -> 8s
Would like to receieve feedback on it!
I decided to use API until microsoft/vscode#166094 is resolved, because, awaiting workspace.applyEdit works properly.