What's Changed

• Add --state option for state machine inspection in #34
• Add state field to ParseError in #35
• Use ArrayVec for multiple reduce rules container in #36
• Add more test cases in #37
• Optimization for unused data in #38 #39 #40 #41
• Use static memory for parser table generation in #42

Full Changelog: v3.32.0...v3.34.0

What's Changed

• fixed can_feed() logic in #28
• fixed expected() logic in #29
• run code generation for sample projects in #30
• check if custom reduce action can be identity in #31
• Single rule optimization can be applied with custom reduce action in #32

Full Changelog: v3.31.0...v3.32.0

What's Changed

• fixed number of tags mismatch in split_off in #24
• fixed eof was not popped on pop_start() in #25
• Removed %runtime directive in #26
• Optimized out unused states and production rules, reduces runtime by 30% in #27

Full Changelog: v3.30.0...v3.31.0

What's Changed

• add IntermediateState for lightweight state generation in #22
• use truncate() instead of pop() for unused and shadowed variables in #23
• optimize counting logic in DataStack::split_off

Full Changelog: v3.29.0...v3.30.0

What's Changed

• fix command line arguments for note diagnostics in #19
• split location stack for reduce action, optimizing the stack management in #20
• replacing the enum-based TokenData approach with separate Vec stacks for each rule type by #21

Full Changelog: v3.28.0...v3.29.0

What's Changed

• use compact integer types instead of usize to reduce memory usage and binary size in #18
• fixed some memory was not freed in non-deterministic parser

Full Changelog: v3.27.0-1...v3.28.0

What's Changed

• optimized non-deterministic parser and increase performance by 10 times in #17

Full Changelog: v3.27.0...v3.27.0-1

What's Changed

• remove explicit eof token feed in #15
• Optimize dense state fields in #16
• support OR separator in paren syntax ( a b | c | )

Full Changelog: v3.25.0...v3.27.0

What's Changed

• change error token logic; refactored the parser to treat the error token as a terminal symbol in #13
• add function call and $sep syntax in #14
  $sep( P, P_separator, *|+ ): A repetition of P separated by P_separator. The repetition can be *, or + to indicate zero or more, or one or more repetitions respectively

Full Changelog: v3.24.0...v3.25.0

What's Changed

• add error-recovery to grammar parser
• Runtime operator precedence conflict resolution in #12
  Example:

%left '+';
%left '*';

E: E op=BinOp E %prec op { ... }
 | ...
 ;

// The precedence of `BinOp` is determined by the production rule that `BinOp` was derived from.
BinOp: '+'
     | '*'
     ;

In this example, the E: E op=BinOp E rule's precedence is determined by the BinOp non-terminal.
When the parser needs to resolve a conflict involving this rule, it will look at what BinOp was reduced from.
If BinOp was reduced from +, the precedence of BinOp will be the precedence of +, which is defined by %left '+'.
otherwise, if BinOp was reduced from *, the precedence of BinOp will be the precedence of *, which is defined by %left '*'.

If any non-terminal symbol was referenced in the %prec directive,
every production rule in that non-terminal must have operator precedence.

Full Changelog: v3.22.0...v3.24.0