The autocompletion system performs its own matching (against the text between from and to).
doesn’t fit some cases where all the options in a CompletionResult should be shown but not all their labels match the text between from and to (e.g., a completion source which is typo tolerance).
Not currently. The filtering does some fuzzy matching, but it does drop completions that it considers to not match well enough.
Does your completer only return completions that should be shown? And without span propery on the result? The main problem with supporting this is that the system allows multiple sources at the same time, and if it doesn’t score the completions (which is part of the filtering) it doesn’t know how to sort them relative to those from other sources.
Yes, the completer (provided via override) returns only the options should be shown.
No span provided.
For my case, the autocompletion service thinks:
Alice ≈ Álice (accent insensitive)
Lexer ≈ Lezer (small distance)
The autocompletion service may return options which are completely different from the current token (according to string similarity) but pertinent to the context.