feat: initial feature-flagged, draft new autocompleter MONGOSH-2036 #2424
Conversation
| @@ -72,6 +72,7 @@ describe('snippet integration tests', function () { | |||
| shell.assertNoErrors(); | |||
| }); | |||
|
|
|||
| // TODO: port to the new autocomplete | |||
There was a problem hiding this comment.
I'll either do this in this PR or I'll file a follow-up ticket
| @@ -29,6 +29,7 @@ export interface InterpreterEnvironment { | |||
| */ | |||
| export class OpenContextRuntime implements Runtime { | |||
| private interpreterEnvironment: InterpreterEnvironment; | |||
| // TODO: we have to also port this to the new autocomplete | |||
There was a problem hiding this comment.
There was a problem hiding this comment.
(I might even do it in this PR)
| @@ -352,6 +363,14 @@ describe('MongoshNodeRepl', function () { | |||
| }; | |||
| const tabtab = async () => { | |||
| await tab(); | |||
| if (process.env.USE_NEW_AUTOCOMPLETE) { | |||
| // TODO: This is because autocomplete() is async and will either list | |||
There was a problem hiding this comment.
Less of a TODO than a NOTE.
There was a problem hiding this comment.
Yeah, I feel like setting a fixed timeout of 210 is either almost guaranteed flakiness or something that at least looks like it would be flaky ... can we somehow wait for the right event (e.g. some autocompletion data fteched event) at this point?
| line: string, | ||
| results: { result: string }[] | ||
| ): [string[], string] | [string[], string, 'exclusive'] { | ||
| // TODO: actually use 'exclusive' when we should |
There was a problem hiding this comment.
| @@ -352,6 +363,14 @@ describe('MongoshNodeRepl', function () { | |||
| }; | |||
| const tabtab = async () => { | |||
| await tab(); | |||
| if (process.env.USE_NEW_AUTOCOMPLETE) { | |||
| // TODO: This is because autocomplete() is async and will either list | |||
There was a problem hiding this comment.
Yeah, I feel like setting a fixed timeout of 210 is either almost guaranteed flakiness or something that at least looks like it would be flaky ... can we somehow wait for the right event (e.g. some autocompletion data fteched event) at this point?
| const hash = crypto.createHash('sha256'); | ||
| hash.update(uri); | ||
| return hash.digest('hex'); | ||
| } |
There was a problem hiding this comment.
This feels like it miiiight be worth caching on the Mongo instance, not sure if that would be premature optimization but while hashing is cheap, it might not be do-a-few-times-per-autocomplete cheap?
| return { | ||
| currentDatabaseAndConnection: () => { | ||
| return { | ||
| connectionId: connectionIdFromURI(this.currentDb.getMongo().getURI()), |
There was a problem hiding this comment.
Is this intended to be the ID that the autocompleter passes to things like connectionId? Would it be simpler to just assign an increasing counter id to each Mongo instance?
Still a draft, don't merge :)
Most of everything is stubbed or TODO and this needs tests. But hopefully it is reasonably close to the interface we're going for.