behavior of example content_assist

hi! my goal is implementing auto-completion for existing language. i’m tring it with simplified sample (fake) grammar, using chevrotain/examples/parser/content_assist as reference.

but it seems that example can’t handle my scenario well 😦

my sample grammar is,

// NL token is NewLine.
// _  token is WhiteSpace.

startRule -> 
  NL? stmt (NL stmt)* NL?

stmt -> 
    functionStmt
  | enumStmt
  | declareStmt

functionStmt -> 
  (visibility _)? (STATIC _)? FUNCTION _ IDENTIFIER

enumStmt -> 
  (visibility _)? (STATIC _)? ENUM _ IDENTIFIER

declareStmt ->
  (visibility _)? DECLARE _ IDENTIFIER

visibility -> 
  PRIVATE | PUBLIC

my expected behavior is,

input	candidates(next possible tokens)
	Private, Public, Static, Function, Enum, Declare
Private	Static, Function, Enum, Declare
Private Static	Function, Enum

please see my code on gist. https://gist.github.com/ayapi/e29c9982aacd0ca3f03b3b83e6c642e4

i cant get expected candidates because, maybe, before consume token, before settle rule, at these position, in parser internally.

is it possible?? i want to realize it somehow or other. actually, i’ve never seen parsing library that can handles well it easily. nearley stores all next probable tokens to table row, so i just have to collect them, but nearley has other issues(slow, hard to get rulestack/tree on error, etc…)). if it can be solved by chevrotain, chevrotain will be best-solution.

[bd82]

I’m working on a different approach which will (hopefully) be built into Chevrotain as a first class feature.

You can keep track on my progress here: https://github.com/SAP/chevrotain/issues/291

[bd82]

Great 😄.

And if you can share the grammar you are working on (link/access/…) I will review it, as it is useful both ways:

• Feedback to Chevrotain --> How do people actually use the library.
• Feedback to Grammar authors --> access to tips/tricks/best practices that are not always documented.

I will close this issue as the original problem was resolved. Additional feedback is always appreciated 👍 so new issues are very welcome.