Fix common prefix completion issue

[msujew]

Closes #1370

Simply updates our version of chevrotain-allstar to 0.4.2 which includes a fix for this. Also adds a test to ensure that we don't regress in the future.

[cdietrich]

@msujew i just have the new version a try by overrding yarn resolution.
now we see a ton of new parse errors.
is this expected?

[msujew]

I'll have another look 👍

[cdietrich]

we have the same pattern at some other places
but also our expressions break. but i could not nail that one yet

[cdietrich]

expression reproducer model

  ensure someOtherNumber / 10  

and grammar

entry Model: items+=Stmt*;

  Stmt: 'ensure' expr=Additive;

  Additive infers Expression:                                                                                                                                                                                                     
      Multiplicative
      (({infer Add.left=current} '+' | {infer Sub.left=current} '-') right=Multiplicative)*;                                                                                                                                      
                                                            
  Multiplicative infers Expression:
      Primary
      (({infer Mul.left=current} '*' | {infer Div.left=current} '/') right=Primary)*;
                                                                                                                                                                                                                                  
  Primary infers Expression:
      RefExpression | NumberLiteral;                                                                                                                                                                                              
                                                            
  RefExpression:
      target=PathOrMetadataAccess;
                                                                                                                                                                                                                                  
  NumberLiteral:
      value=NUMBER;                                                                                                                                                                                                               
                                                            
  PathOrMetadataAccess returns string:
      PATH | MetadataAccess;

  // '/' inside PATH collides with '/' as division operator                                                                                                                                                                       
  PATH returns string:
      ID | (ID ('/' ID)* ':' ID);                                                                                                                                                                                                 
                                                            
  terminal MetadataAccess: /@[_a-zA-Z][\w_-]*/;                                                                                                                                                                                   
  terminal NUMBER: /[0-9]+(\.[0-9]+)?/;
  terminal ID: /[_a-zA-Z][\w_-]*/;                                                                                                                                                                                                
  hidden terminal WS: /\s+/;  

hidden terminal ML_COMMENT: /\/\*[\s\S]*?\*\//;
hidden terminal SL_COMMENT: /\/\/[^\n\r]*/;

[msujew]

@cdietrich I've added your test, and probably found a good compromise. I had to add an incomplete flag to the lookahead to make this work - but I believe it does now work as expected. See #2147.

[cdietrich]

@msujew my first testcase still does not work
need to double check i did not screw stuff up in building it
can we still produce alphas?

[msujew]

my first testcase still does not work

I've reverted all changes related to the parsing phase - the only difference should now be in the completion phase.

can we still produce alphas?

I have pushed a next release for you, see 4.2.2-next.d9c47299.

[cdietrich]

ok the 2nd test as a escape problem in the ts backtick string literal
i am still evaluating the results in the completion tests

[cdietrich]

update: they look promising