Partial parsing results for completions about superpower HOT 5 CLOSED

Hi! Nothing like this is built in - not sure how it could be welded on, but very interested to hear how you go if you come up with something!

Best regards,
Nick

from superpower.

@nblumhardt, to be honest, in such case I see only two solutions:

• as commonly done, you can always create additional parsers for the "wrong" situations
• avoid this problem altogether and provide completions only for correctly parsed programs (identifier completions, for example)

from superpower.

The first solution will partially solve the problem, however, we all know that you cannot physically cover all the scenarios. Also, the codebase will become a big mess, trying to make parsers for all the "wrong" paths a user can take.
The second one is feasible in the beginning, as most of the times you want exactly that type of completions. However, in that case you will have a very limited set of possible completions you can provide.

Either of those ways won't provide a full experience for the user. The only practical solution I see is having a parser provide its state in a form of a list of "possible" trees, to solve the possible ambiguity problem of parsing incomplete sources.

from superpower.

Also, you can try to use the type data of the previous successful compilations as the source of truth. It will allow you to query data about the failed token, based on that info and provide a meaningful set of completions. Hmm, it won't even require any kind of magic. Interesting!

from superpower.

Thanks for sharing your thoughts. I might close this one as I don't think there's a direction to follow here, yet, but keep us in the loop with your investigations! :-)

from superpower.