Fix #40941 - Suggest trigger at language boundaries #43349
Conversation
| @@ -46,7 +46,19 @@ export class LineContext { | |||
| } | |||
| const pos = editor.getPosition(); | |||
| model.tokenizeIfCheap(pos.lineNumber); | |||
| const word = model.getWordAtPosition(pos); | |||
|
|
|||
| // When we are at an embedded language boundary, check the word at the previous character | |||
There was a problem hiding this comment.
This some pretty ugly logic. Let me know if there is some better way to express this
There was a problem hiding this comment.
Talked to @alexandrudima about this and the outcome is that getWordAtPosition should be fixed instead of the consumer. Today the function is biased towards the right but it should look left when (1) there was no result on the right and (2) when it started at a language-border.
|
Ok, I didn't do that initially because I was concerned about regressions. I've pushed an update that moves the fix into Take a look at the test case for |
Fixes #40941 **Bug** Suggestions not auto triggered when you `x` in `const z = <div x>`. See #40941 (comment) for explaination of root cause **Proposed fix** In the case where suggestions are triggered at the boundary between two languages, I believe we should trigger. This change adds special logic to handle that case by making sure we check the word at the previous character position instead of at the current character positon when at language boundaries
…r the left language
|
👍 |
Fixes #40941
Bug
Suggestions not auto triggered when you type
xinconst z = <div x>. See #40941 (comment) for explaination of root causeProposed fix
In the case where suggestions are triggered at the boundary between two languages, I believe we should trigger. This change adds special logic to handle that case by making sure we check the word at the previous character position instead of at the current character positon when at language boundaries