Remove suggestion mode for implicit array creation

[jcouv]

Customer scenario

Type Class[] x = new Class[]. You expect the Class on the right-hand-side to be completed. But actually, there was a suggestion mode for implicit array creation that prevented such completion.
I don't think such a suggestion mode makes sense. Suggestion modes prevent completion when the user might type a never-seen word, but that is not the case here. So I removed it.

Bugs this fixes

Fixes #24432

Workarounds, if any

You can always fight against the completion, but it's annoying.

Risk

Performance impact

Is this a regression from a previous update?

No

Root cause analysis

How was the bug found?

Reported by customer

FYI @CyrusNajmabadi @sharwell

[CyrusNajmabadi]

the issue here is that the user may type Foo[] f = new [. So i'm fine with any changes here as long as the above can be written out without intellisense completion interfering and causing you to write: ```Foo[] f = new Foo[".

[CyrusNajmabadi]

Suggestion modes prevent completion when the user might type a never-seen word, but that is not the case here

I agree we probably don't need suggestion mode. But we do need to ensure that the items is not hard selected.

[jcouv]

From my understanding, if the item were soft-selected, then the common scenario would become less convenient (user must use arrow to select).
I'm afraid this would also affect the object creation scenario (not just array creation), which is very common and should be preserved.

One potential way out would be to keep hard-selection for object creation, but use soft-selection when inferred type is an array. I'll try that.

[CyrusNajmabadi]

From my understanding, if the item were soft-selected, then the common scenario would become less convenient (user must use arrow to select).

Yes. That's intentional though. Otherwise the other normal scenario (implicit arrays) becomes painful. implicit arrays are totally reasonable and normal to type. It was intentional to make it so that we don't aggressively complete arrays and annoy the user when they type this.

[sharwell]

Here are a few cases I'll be looking to verify:

• Typing Foo[] data = new $$ should result in Foo being soft-selected ($$ is used as a position marker, not literal characters)
• Typing Foo[] data = new F$$ should result in Foo being hard-selected
• Typing Foo[] data = n should result in new being hard-selected
• Typing Foo[] data = n[ should produce Foo[] data = new[
• Typing Foo[] data = n [ should produce Foo[] data = new [
• Typing Foo[] data = new [ should produce Foo[] data = new [
• Pressing Ctrl+Space in Foo[] data = new $$ should hard select Foo

[CyrusNajmabadi]

Pressing Ctrl+Space in Foo[] data = new $$ should hard select Foo

Interesting! Does ctrl-space change soft-selection to hard selection?

Another test case would be:

Foo[] data = new<space><tab>

Should make "new Foo"

[sharwell]

Interesting! Does ctrl-space change soft-selection to hard selection?

The Edit.CompleteWord command is allowed to always assume that the user is attempting to insert a reference to something that already exists.

Another test case would be:

Foo[] data = new<space><tab>

This case is covered by my first typing case, assuming that pressing Tab completes a soft-selected item.

[jcouv]

@sharwell I added some tests cases as you suggested, although I think they relate more to completion on "new" which is a different provider.
I tested Tab, but wasn't sure how to test Ctrl+Space.

[CyrusNajmabadi]

I think ctrl-space is "complete word" right? If so, you can use testState.SendCommitUniqueCompletionListItem

[CyrusNajmabadi]

Note: i don't see anything in the complete-word codepath that would cause soft-selection to become hard selection. This is functionality i think we'd have to add.

[jcouv]

@sharwell ctrl+space sounds like a separate issue.
OK with current change?

[sharwell]

@jcouv It looks good. Do we have any tests yet for this syntax? I expect they will all pass but it'd still be good to have them.

int[] = { 0 };

[sharwell]

💡 Consider adding tests for the array initializer syntax if there aren't any already.

[jcouv]

@Pilchie I'm not sure the current ask-mode process for 15.7. Please approve if ask-mode in effect. Thanks

[jcouv]

Checked with Jared. No ask-mode approval required for this branch yet. I'll go ahead and merge. Thanks