# -* text -*-
## PE grammar for the grammar specification language accepted by the
## LEMON parser generator.
## Questions:
## Are directives always at the beginning of a line ?
## Or can they start within a line ?
## The ifdef/ifndef/endif directives are not documented.
## Nor are the cmdline options
PEG pg::peg::lemon (LemonGrammar)
LemonGrammar <- SPACE Statement + EOF;
Statement <- ( Directive
/ Rule
) SPACE;
Rule <- Identifier Label? ASSIGN Definition DOT Precedence? Codeblock? ;
Definition <- (Identifier Label?)* ;
Label <- LPAREN Identifier RPAREN;
Precedence <- LBRACKET Identifier RBRACKET;
Directive <- DINTRO ( Code / DefaultDestructor
/ DefaultType / Destructor
/ ExtraArgument / Include
/ Left / Name
/ Nonassoc / ParseAccept
/ ParseFailure / Right
/ StackOverflow / Stacksize
/ StartSymbol / SyntaxError
/ TokenDestructor / TokenPrefix
/ TokenType / Type
/ Fallback
);
Code <- DCODE Codeblock;
DefaultDestructor <- DDEFDEST Identifier Codeblock;
DefaultType <- DDEFTYPE Codeblock;
Destructor <- DDEST Identifier Codeblock;
ExtraArgument <- DEXTRA Codeblock;
Include <- DINCL Codeblock;
Left <- DLEFT Identifier+ DOT;
Name <- DNAME Identifier ;
Nonassoc <- DNON Identifier+ DOT;
ParseAccept <- DPACC Codeblock;
ParseFailure <- DPFAIL Codeblock;
Right <- DRIGHT Identifier+ DOT;
StackOverflow <- DSTKOVER Codeblock;
Stacksize <- DSTKSZ NaturalNumber;
StartSymbol <- DSTART Identifier;
SyntaxError <- DSYNERR Codeblock;
TokenDestructor <- DTOKDEST Identifier Codeblock;
TokenPrefix <- DTOKPFX Identifier;
TokenType <- DTOKTYPE Codeblock;
Type <- DTYPE Identifier Codeblock;
Fallback <- DFALLBK Identifier+ DOT;
# Lexical layer
match: Codeblock <- LBRACE ( Codeblock
/ C_COMMENT
/ Cplusplus_COMMENT
/ (!RBRACE .)
)* RBRACE ;
Identifier <- Ident SPACE;
match: Ident <- (<alpha>/'_') (<alnum>/'_')*;
NaturalNumber <- NatNum SPACE;
match: NatNum <- [0-9]+;
void: ASSIGN <- '::=' SPACE;
void: DOT <- '.' SPACE;
void: LPAREN <- '(' SPACE;
void: RPAREN <- ')' SPACE;
void: LBRACKET <- '[' SPACE;
void: RBRACKET <- ']' SPACE;
void: LBRACE <- '{';
void: RBRACE <- '}';
void: DINTRO <- '%';
void: DCODE <- 'code' SPACE;
void: DDEFDEST <- 'default_destructor' SPACE;
void: DDEFTYPE <- 'default_type' SPACE;
void: DDEST <- 'destructor' SPACE;
void: DEXTRA <- 'extra_argument' SPACE;
void: DINCL <- 'include' SPACE;
void: DLEFT <- 'left' SPACE;
void: DNAME <- 'name' SPACE;
void: DNON <- 'nonassoc' SPACE;
void: DPACC <- 'parse_accept' SPACE;
void: DPFAIL <- 'parse_failure' SPACE;
void: DRIGHT <- 'right' SPACE;
void: DSTKOVER <- 'stack_overflow' SPACE;
void: DSTKSZ <- 'stack_size' SPACE;
void: DSTART <- 'start_symbol' SPACE;
void: DSYNERR <- 'syntax_error' SPACE;
void: DTOKDEST <- 'token_destructor' SPACE;
void: DTOKPFX <- 'token_prefix' SPACE;
void: DTOKTYPE <- 'token_type' SPACE;
void: DTYPE <- 'type' SPACE;
void: DFALLBK <- 'fallback' SPACE;
# These have % in their definition because they
# are not part of the regular directives.
void: DIFDEF <- '%ifdef' SPACE;
void: DIFNDEF <- '%ifndef' SPACE;
void: DENDIF <- '%endif' SPACE;
# Directives we do not really know about. They are in the SQLite
# parse.y file, but are not documented for LEMON. We treat them as
# whitespace for now, see below.
Ifdef <- DIFDEF Identifier;
Ifndef <- DIFNDEF Identifier;
Endif <- DENDIF;
# Whitespace
void: SPACE <- ( [ \t\n\r]
/ C_COMMENT
/ Cplusplus_COMMENT
/ Ifndef / Ifdef / Endif
)*;
C_COMMENT <- CCOM_OPEN (!CCOM_CLOSE .)* CCOM_CLOSE;
CCOM_OPEN <- '/*';
CCOM_CLOSE <- '*/';
Cplusplus_COMMENT <- '//' (!EOL .)* EOL;
EOL <- '\r\n' / '\r' / '\n';
EOF <- !.;
END;
