Get different results when run the parser multiple times

[dongli]

I am writing a parser for a configuration file format. Here is the grammar that I have written

  start: expr*

  %import common.CNAME
  %import common.ESCAPED_STRING
  %import common.SIGNED_NUMBER
  %import common.WS

  COMMENT: /\/\/.*/
  OPERATOR: ">=" | "<=" | ">" | "<" | "=="
  TRUE: "TRUE"
  FALSE: "FALSE"

  %ignore WS
  %ignore COMMENT
  %ignore ";"

  expr: key "=" value

  key: CNAME
  value: dict
       | array
       | boolean
       | string
       | number
       | threshold
       | keyword

  dict: "{" expr* "}"
  array: "[" (value "," )* [value] "]"
  string: ESCAPED_STRING
  number: SIGNED_NUMBER
  boolean: TRUE | FALSE
  threshold: OPERATOR number
  keyword: CNAME

And the sample file is:

climo_mean = {

   file_name = [];
   field     = [];

   regrid = {
      vld_thresh = 0.5;
      method     = NEAREST;
      width      = 1;
   }

   time_interp_method = DW_MEAN;
   match_day          = FALSE;
   time_step          = 21600;
}

I also wrote a transformer to turn the tree into a dict object, when I noticed that the FALSE is parsed as boolean or keyword at different runs. I have already put boolean rule before keyword. Any idea why I got random results? Thanks!

RUN 1:

keyword NEAREST
keyword DW_MEAN
boolean FALSE
{'climo_mean': {'field': [],
                'file_name': [],
                'match_day': False,
                'regrid': {'method': 'NEAREST', 'vld_thresh': 0.5, 'width': 1},
                'time_interp_method': 'DW_MEAN',
                'time_step': 21600}}

RUN 2:

keyword NEAREST
keyword DW_MEAN
keyword FALSE
{'climo_mean': {'field': [],
                'file_name': [],
                'match_day': 'FALSE',
                'regrid': {'method': 'NEAREST', 'vld_thresh': 0.5, 'width': 1},
                'time_interp_method': 'DW_MEAN',
                'time_step': 21600}}

[erezsh]

Hi, I believe this is a known bug: #201

I'm going to roll out Lark v0.7 in a week or two, which should solve this problem.

Meanwhile, you can use the earley_sppf branch for testing, and to verify that this is indeed a known issue.

[erezsh]

Btw, as another solution meanwhile, you can try to play with rule weights. For example:

boolean.10: TRUE | FALSE

[dongli]

@erezsh I have tried rule weight, but it does not work.