Larger Max numerical suffix/idx for the DSL Methods

[fengliner]

09-2019 Update

This is now possible as Chevrotain no longer depends on Function.prototype.toString

• Remove usage of Function.prototype.toString() #998

The following DSL methods have been added in version 6.4.0 which accept an idx argument.

• consume
• CstParser.subrule
• EmbeddedActionsParser.subrule
• or
• option
• many
• atLeastOne

There are no respective _sep variations as the _SEP dsl methods are likely to be re-implemented
as macros in the future rather than a base "building block".

• See: User Defined Macros #1004

The maximum valid idx is 255, this is validated during grammar recording, an invalid value
will cause an error to be thrown during parser initialization.

---

Original Issue

What is max numerical suffix for the DSL Rules？Is 9?
What can I do if numerical suffix more than 9？

[bd82]

Do you have a real example where you needed more than 9 suffix?

The suffixes are part of the "identity" of the specific DSL rule,
What this means is that as long as the identity is unique you don't need a unique suffix,
For example

$.RULE("myRule", () => {
  // we dont need any suffixes here because the "Identity" also uses the Token Name...
  $.CONSUME(A)
  $.CONSUME(B)
  $.CONSUME(C)
  $.CONSUME(D)
  $.CONSUME(E)
)

What can I do if numerical suffix more than 9？

• You can create new suffix methods in your grammar based on the existing ones.

  • https://github.com/SAP/chevrotain/blob/master/src/parse/parser_public.ts#L624-L626

• You can refactor your rule and split it up into multiple rules.

[fengliner]

More than 9 suffix is normal to mysql grammar. Example, the tableOption rule, option suffix is more than 9.

tableOption:
ENGINE '='? engineName
| AUTO_INCREMENT '='? decimalLiteral
| AVG_ROW_LENGTH '='? decimalLiteral
| DEFAULT? (CHARACTER SET | CHARSET) '='? charsetName
| CHECKSUM '='? boolValue = ('0' | '1')
| DEFAULT? COLLATE '='? collationName
| COMMENT '='? STRING_LITERAL
| COMPRESSION '='? STRING_LITERAL
| CONNECTION '='? STRING_LITERAL
| DATA DIRECTORY '='? STRING_LITERAL
| DELAY_KEY_WRITE '='? boolValue = ('0' | '1')
| ENCRYPTION '='? STRING_LITERAL
| INDEX DIRECTORY '='? STRING_LITERAL
| INSERT_METHOD '='? insertMethod = (NO | FIRST | LAST)
| KEY_BLOCK_SIZE '='? fileSizeLiteral
| MAX_ROWS '='? decimalLiteral
| MIN_ROWS '='? decimalLiteral
| PACK_KEYS '='? extBoolValue = ('0' | '1' | DEFAULT)
| PASSWORD '='? STRING_LITERAL
| ROW_FORMAT '='? rowFormat = (
DEFAULT
| DYNAMIC
| FIXED
| COMPRESSED
| REDUNDANT
| COMPACT
)
| STATS_AUTO_RECALC '='? extBoolValue = (DEFAULT | '0' | '1')
| STATS_PERSISTENT '='? extBoolValue = (DEFAULT | '0' | '1')
| TABLESPACE uid tablespaceStorage?
| UNION '='? '(' tables ')';

[bd82]

Thanks for the example.

Personally I would not translate such a rule "directly" without changes.

• Option1: each one of the alternative could get its own nested rule.
  • This would enable better seperation for the semantics (e.g AST Building logic).
• Option2: Factor out the common left hand prefix and examine if the value on the right hand side is of the correct "type" in a post processing phase.

  •      tableOption:
            tableOptKey '='? tableOptValue
            |  TABLESPACE uid tablespaceStorage?
         ;
    

  • That would simplify the grammar, but add some complexity at a later phase.

Either of the options would avoid the need for really large suffixes.
Additionally there is an internal limitation that limits a suffix to 4bits (value 15).
Anything more could lead to unexpected errors due to the creation of invalid cache keys.

[bd82]

Closing this, please re-open if needed.

[bd82]

Re-opening as this is now possible.