Auto-complete should use symbols defined in the current program

[cdsmith]

Right now, auto-complete works against a fixed list of symbols from the standard library. In addition to that, it should look in the current file for things that look like definitions, and add them to the auto-complete list.

[Maaarcocr]

I would be interested in working on this feature for the Summer of Haskell. I would probably work on different others feature such as issue number 34: Auto-complete should be type-sensitive and issue number 129: Rainbow brackets. Would it be possible to further discuss about this possibility?

[cdsmith]

@Maaarcocr Sure, tackling a cluster of editor usability features would be a great proposal! Feel free to get in touch with me at cdsmith@gmail.com to discuss any questions you have.

[nixorn]

If it is actual, I can try.

[cdsmith]

I noticed when testing last night that something like this seems to be happening already? But only for definitions with a type signature. I do not know when this change was made. I'll track it down when I have a sec

[cdsmith]

It looks like a rough version of this was added in c63fd5c by a Summer of Haskell student two years ago. I missed it because it was one commit in a long list.

Remaining issues:

1. It only works for symbols with top-level type signatures. This is probably okay, since implementing type inference in the browser is less than ideal.
2. The newly added symbols are appended to the bottom of the autocomplete list, which makes this feature less discoverable than it could be.
3. User-defined symbols are not included in the newly implemented doc-on-hover feature. It would be great to add them there as well.

Points 2 and 3 together indicate that there's room for keeping some kind of canonical dictionary of known terms that is updated as editing continues, and is used as a common source for both hover and autocomplete.

[cdsmith]

@nixorn I noticed that you indicated you're working on this for BountySource. As I mentioned above, @three implemented this in the summer of 2017. I made a small improvement to sorting (which may have been broken by the doc-on-hover CL anyway, but I don't think so). The core feature here already exists. There are definitely improvements, but this gets messy in terms of the bounty. I'd have to award the bounty to @three if he wants to claim it; otherwise, if the two of you agree to something else, I'm happy to go along.

[nixorn]

@cdsmith I initiated this as refactoring of all hint-and-hover-stuff as you described in previous message. I plan to create one general variable with docs and annotations, implement updating of this variable on user change, make render functions for hovers and hints (to separate data and view) etc. And solve #135 along. I'm not against awarding bounty to @three, I going to follow this plan anyway.

[cdsmith]

Sounds great! The plan sounds great to me. You might want to limit how often you update the extracted list of hints/docs, because it might make typing annoying to scan the whole buffer on every keystroke for sizable projects on low-end Chromebooks that are used in many schools. That's not too hard to fix, though. Perhaps you could do something like: every time a change is made, we set a three second timer. If additional changes are made in that three seconds, the timer gets reset. When three seconds passes without another change, we scan the source and update the list." (Or you could do something like the way syntax highlighting works, where you store the state and only incrementally rescan starting from the location of the change. But that sounds more complicated.)

Thanks for the understanding about the bounty; I'm just trying not to leave anyone feeling they are owed something and didn't get it. But if Eric doesn't want to claim it, then I'm happy to give you the bounty instead for your extension work.

[nixorn]

Is this also should be closed via #799 ?

[cdsmith]

Yeah... though I think I need to revert that commit. It looks like we lost all the formatting in the auto-complete box. Did you test that at all?

[nixorn]

Yes, I tested that code.

Where is regression?