Strikeout deprecated CompletionItems

[kamranayub]

Closes #50972

Changes

• Add CompletionItem.deprecated
• Add deprecated flag to SuggestDataDto
• Add inline styling to suggest widget

[kamranayub]

@jrieken I wasn't sure if the way I was applying styles was recommended so definitely looking for feedback there (and whether or not to externalize the CSS class name)

[KamasamaK]

Note that adding a dedicated deprecated property would conflict with @jrieken's proposal at
#23927 (comment)

[kamranayub]

@KamasamaK Confused though, because I thought it was already defined in the LSP:

https://microsoft.github.io/language-server-protocol/specification#textDocument_completion

Won't this change match that interface defined there? I assumed the interface is a 1:1 mapping.

[jrieken]

Won't this change match that interface defined there? I assumed the interface is a 1:1 mapping.

There is no 1:1 mapping, tho often a conceptual mapping between the extension API and the LSP. @KamasamaK raises a valid point here, not sure how to proceed. Adding a single boolean is obviously the easiest change but we don't wanna do that all the time, e.g. for private, static, readonly, etc, etc. So maybe some upfront design work is needed. Also checking what @dbaeumer's plans for LSP are?

[kamranayub]

Without a lot of background/context for this, my uneducated view is that private, static, readonly would be access modifiers while deprecated seems like a separate concern, like an annotation/tag/status.

[jrieken]

, my uneducated view is that private, static, readonly would be access modifiers while deprecated seems like a separate concern

Well, we don't wanna be to academic here. Sure 'public', 'private', 'protected', 'friend' etc are access modifiers but there is more, like 'static' or 'final' or Java's multithread modifiers like 'volatile' and 'transient'. I think there is no absolute truth and the API never tried to correctly represent actual language constructs but to be a common denominator of a UI model the reflects those things. So, one can think of a bit mask of such flags that add some extra rendering to a symbol, like a static method, or a private, deprecated constructor

[dbaeumer]

I would support adding a modifier instead of a separate deprecated field. There is no rule for a 1:1 mapping. And there are requests in LSP to support more than deprecated which would lead to a modifier as well.

[kamranayub]

Once I have a hot second I can explore switching to a modifier. 👍

[jrieken]

I will go ahead here and merge things and do more the remaining changes on master. Thanks @kamranayub

[kamranayub]

Awesome, thanks! 🙏

…

On Wed, Aug 21, 2019, 4:22 AM Johannes Rieken ***@***.***> wrote: Merged #78092 <#78092> into master. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#78092?email_source=notifications&email_token=AAEJU25MX3ZRVDW62IVVOLDQFUCNNA5CNFSM4IHSPBU2YY3PNVWWK3TUL52HS4DFWZEXG43VMVCXMZLOORHG65DJMZUWGYLUNFXW5KTDN5WW2ZLOORPWSZGOTFJUE4Y#event-2572370547>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAEJU24HVRK2WDRYK7MC2ETQFUCNNANCNFSM4IHSPBUQ> .

[jrieken]

No - thank you!