Adds missing parameters to ICompleterProvider API

[hbcarlos]

References

Fixes #14806

Code changes

see #14806

User-facing changes

N/A

Backwards-incompatible changes

N/A

[jupyterlab-probot]

Thanks for making a pull request to jupyterlab!
To try out this branch on binder, follow this link:

[hbcarlos]

Hey @krassowski, should we change the IProviderReconciliator and the logic to pass the necessary arguments, or does jupyterlab-lsp reimplements the handler and everything?

jupyterlab/packages/completer/src/tokens.ts

Lines 157 to 181 in 2161e6c

	export interface IProviderReconciliator {
	/**
	* Fetch response from multiple providers, If a provider can not return
	* the response for a completer request before timeout,
	* the result of this provider will be ignore.
	*
	* @param {CompletionHandler.IRequest} request - The completion request.
	*/
	fetch(
	request: CompletionHandler.IRequest
	): Promise<CompletionHandler.ICompletionItemsReply | null>;
	/**
	* Check if completer should make request to fetch completion responses
	* on user typing. If the provider with highest rank does not have
	* `shouldShowContinuousHint` method, a default one will be used.
	*
	* @param completerIsVisible - The visible status of completer widget.
	* @param changed - CodeMirror changed argument.
	*/
	shouldShowContinuousHint(
	completerIsVisible: boolean,
	changed: SourceChange
	): boolean;
	}

jupyterlab/packages/completer/src/handler.ts

Lines 295 to 304 in 2161e6c

	this._autoCompletion &&
	(this._reconciliator as IProviderReconciliator)
	.shouldShowContinuousHint &&
	(this._reconciliator as IProviderReconciliator).shouldShowContinuousHint(
	this.completer.isVisible,
	changed
	)
	) {
	void this._makeRequest(editor.getCursorPosition());
	}

jupyterlab/packages/completer/src/handler.ts

Lines 347 to 348 in 2161e6c

	return this._reconciliator
	.fetch(request)

[krassowski]

If we are changing the interface we should also implement the logic here. It will be useful beyond LSP implementation (for other completion providers). In the current draft the LSP implementation no longer needs to swap the completion handler (which was a nightmare previously), so hopefully we won't need to do this again.

[krassowski]

Thanks! One minor thing: a tests checking that trigger gets passed down would be useful here (because trigger is optional typing will not save us if someone deletes it accidentally in a refactor).

[fcollonval]

Thanks @hbcarlos

[fcollonval]

CI failure are not related

[fcollonval]

@meeseeksdev please backport to 4.0.x