Do we still need the auto complete fork?

[thien-do]

Afaik the fork is created for 2 primary reasons:

• Fuzzy search scoring Consider scoring of fuzzy matches in autocompletion codemirror/dev#236
• Highlight matched string Allow styling matched substrings in the completion dropdown codemirror/dev#234

Since both issues are addressed by codemirror's autocomplete module, do you think we can go back to the original module yet? Is there some plan for the fork?

[Nexucis]

It requires to be tested and validated we have the same functionality than today.

Note that the autocomplete fork is including a limitation of the autocomplete list. It's a performance issue fixed by @juliusv when you have to display 100k of element. It actually displays only the first 100 or 1000 ( don't remember exactly the limit)
And I don't think we reported this issue to cmn yet.

[thien-do]

CM is going to release a new version soon, which I will try to validate the functionality and report here :D hope it helps

Note that the autocomplete fork is including a limitation of the autocomplete list.

I think this should be implemented at the user land. It's difficult for the autocomplete module to decide a good limitation for all use cases. WDYT @juliusv ?

Afaik the above fix is the only difference between the fork and the recently updated original

[juliusv]

@dvkndn Yeah, in the general case this should likely be a configurable limit, but when I last looked at the code it seemed cumbersome to pass a user-supplied limit setting all the way down to the code path that applies the limit. Probably is necessary though.

[thien-do]

yes it's necessary. If we can't configure the autocomplete module, is it possible to apply this at user land? Like, I don't know, maybe after fetching from the server?

[juliusv]

@dvkndn As I understand it, this wouldn't be solvable by the autocomplete sources that we provide because they still need to provide all items for further filtering etc.. We only want to limit the display, not how many items are processed before that.

[thien-do]

Oh.. it needs to be handled in the module then. Let me find if there's a related issue on codemirror yet. To be honest it's good to limit the display in general for performance reason anyway :D I think we can suggest this to CM

[juliusv]

@dvkndn Yeah for sure, I think it makes sense for any user who has a huge number of underlying items, since it brings down the display time from multiple seconds to something like 100ms.

[juliusv]

@dvkndn Btw., I filed codemirror/dev#293.