ANTLR4: Correctly matching common prefixes - antlr4

I have the following ANTLR4 grammar:
grammar test;
start_symbol: '(FILE' line* ')' EOF ;
line: '(' ID ')' ;
ID: [a-zA-Z_] [a-zA-Z0-9_]* ;
White_space : [ \t\n\r]+ -> skip ;
... and it perfectly works on this sample input file:
(FILE
(LINE)
)
But I also want it to work on:
(FILE
(FILELINE)
)
This does not work. Obviously, the lexer generated an implicit '(FILE' token, which will also match the '(FILELINE' in the second line, which leads to an error.
How can I fix this?
Bonus: I also want to parse this:
(FILE
(FILE)
)
Thanks :)

Something like this would do it:
start : '(' FILE line* ')' EOF;
line : '(' id ')';
id : ID | FILE;
FILE : 'FILE';
ID : [a-zA-Z_] [a-zA-Z0-9_]*;
SPACE : [ \t\n\r]+ -> skip;

Related

Why isn't the program token recognized? ANTLR4

I have this grammar:
grammar BajaPower;
// Gramaticas
programa:PROGRAM ID ';' vars* bloque ;
vars:VAR ((ID|ID',')+ ':' tipo ';')+;
tipo:(INT|FLOAT);
bloque:'{' estatuto+ '}';
estatuto: (asignacion|condicion|escritura);
asignacion: ID '=' expresion ';';
condicion: 'if' '(' expresion ')' bloque (';'|'else' bloque ';');
escritura: 'print' '(' (expresion|STRING ',')* (expresion|STRING) ')' ';';
expresion: exp ('>'|'<'|'<>') exp;
exp: (termino ('+'|'-')*|termino);
termino: (factor ('*'|'/')*|factor);
factor: ('(' expresion ')')|('+'|'-') varcte| varcte;
varcte: (ID|CteI|CteF);
// Tokens
WS: [\t\r\n]+ -> skip;
PROGRAM:'program';
ID:([a-zA-Z]['_'(a-zA-Z0-9)+]*);
VAR:'var';
INT:'int';
FLOAT:'float';
CteI: ([1-9][0-9]*|'0');
CteF: [+-]?([0-9]*[.])?[0-9]+;
STRING:'"' [a-zA-Z0-9]+ '"';
And I'm trying to test it with the following code:
program TestCorrect;
var
x,y:int;
z:float;
{
x = 1;
y = 2;
z = (x+y*3)/4;
if (z > x) {
print("hola mundo",(x+y));
}
}
When I run it it only detects program as an ID and not the PROGRAM token.
There are quite a few things going wrong. In future, I suggest you incrementally create your grammar instead of (trying) to write the entire thing in one go and then coming to the conclusion it doesn't do what you meant it to.
Let's start with the lexer:
WS: [\t\r\n]+ -> skip does not include spaces
ID: ['_'(a-zA-Z0-9)+]* should be ('_'[a-zA-Z0-9]+)*
ID: the first part, [a-zA-Z], should probably be [a-zA-Z]+
VAR, INT, FLOAT are placed after ID, so when ID is properly defined, it will match var, int and float before these tokens
CteF: don't include [+-]?, leave that for the parser to deal with
STRING: [a-zA-Z0-9]+ doe not include spaces, so "hola mundo" will not be matched
Now the parser:
vars: (ID|ID',')+ is wrong because it now always has to end with a comma if you want to match multiple ID's. Do ID (',' ID)* instead
condicion: (';'|'else' bloque ';') mandates a semi-colon should always be present after an if or else block, but in your input, you do not have a semi-colon. Do ('else' bloque)? instead
expresion: exp ('>'|'<'|'<>') exp means an expresion always contains one of the operators >, < or <>, which is not correct (an expression can also just be 1*2). Do exp (('>'|'<'|'<>') exp)? instead
exp: termino ('+'|'-')* is odd: that will match 1++++++++++++. Do termino (('+'|'-') termino)* instead
termino: factor ('*'|'/')* should be factor (('*'|'/') factor)* (same as exp)
varcte: should probably include STRING so that you do not have to do this on multiple places: (expresion|STRING) but can then just do expresion
All in all, this should do the trick:
grammar BajaPower;
programa
: PROGRAM ID ';' vars* bloque
;
vars
: VAR (ID (',' ID)* ':' tipo ';')+
;
tipo
: INT
| FLOAT
;
bloque
:'{' estatuto+ '}'
;
estatuto
: asignacion
| condicion
| escritura
;
asignacion
: ID '=' expresion ';'
;
condicion
: 'if' '(' expresion ')' bloque ('else' bloque)?
;
escritura
: 'print' '(' (expresion ',')* expresion ')' ';'
;
expresion
: exp (('>'|'<'|'<>') exp)?
;
exp
: termino (('+'|'-') termino)*
;
termino
: factor (('*'|'/') factor)*
;
factor
: '(' expresion ')'
| ('+'|'-')? varcte
| STRING
;
varcte
: ID
| CteI
| CteF
;
WS : [ \t\r\n]+ -> skip;
PROGRAM : 'program';
VAR : 'var';
INT : 'int';
FLOAT : 'float';
CteI : [1-9][0-9]* | '0';
CteF : [0-9]* '.' [0-9]+;
ID : [a-zA-Z]+ ('_' [a-zA-Z0-9]+)*;
STRING : '"' .*? '"';

Parsing key / value subparameters

I'm a bit clueless as to how I can parse (more or less) "free form" parameter lists, suppose the syntax allows for
PARM=(VAL1, 'VAL2', VAL3, KEY4=VAL4, KEY5=VAL5(XYZ), PARM=ABC, SOMETHING=ELSE)
I have managed to basically parse combos of positional and key/value parameters, but as soon as I hit a lexer token like PARM= the parser bails out with a "mismatched input", and I can't specifically allow for or expect anything because these parameters passed to a function are completely arbitrary.
So I'd think I'll need to switch to a specific lexer mode but right now I can't see how I would properly switch back to "normal" mode, the delimiters are PARM=( on the left and the closing ) on the right, but as the "data" itself can contain (pairs of) brackets how would I identify the correct closing paren so I don't prematurely end the lexer mode?
TIA - Alex
Edit 1:
Minimal grammar showing the issue with keywords being used where they shouldn't, as this is part of a complex grammar I can't change the order of tokens to put ID in front of everything else, for example, as it would catch too much. So I don't see how this can work short of breaking out into a different lexer mode.
lexer grammar ParmLexer;
SPACE : [ \t\r\n]+ -> channel(HIDDEN) ;
COMMA : ',' ;
EQUALS : '=' ;
LPAREN : '(' ;
RPAREN : ')' ;
PARM : 'PARM=' ;
ID : ID_LITERAL ;
fragment ID_LITERAL : [A-Za-z]+ ;
.
parser grammar ParmParser;
options { tokenVocab=ParmLexer; }
parms : PARM LPAREN parm+ RPAREN ;
parm : (pkey=ID EQUALS)? pval=ID COMMA? ;
Input:
PARM=( TEST, KEY=VAL, PARM=X)
Results in
line 1:22 extraneous input 'PARM=' expecting {')', ID}
So I'd think I'll need to switch to a specific lexer mode but right now I can't see how I would properly switch back to "normal" mode
Instead of switching to modes (with -> mode(...)), you can push your "special" mode on a stack (with -> pushMode(...)) and then when encountering a ) you pop a mode from the stack. That way, you can have multiple nested lists (..(..(..).)..). A quick demo:
lexer grammar ParmLexer;
SPACE : [ \t\r\n]+ -> channel(HIDDEN);
EQUALS : '=' ;
LPAREN : '(' -> pushMode(InList);
PARM : 'PARM';
ID : [A-Za-z] [A-Za-z0-9]*;
mode InList;
LST_LPAREN : '(' -> type(LPAREN), pushMode(InList);
RPAREN : ')' -> popMode;
COMMA : ',';
LST_EQUALS : '=' -> type(EQUALS);
STRING : '\'' ~['\r\n]* '\'';
LST_ID : [A-Za-z] [A-Za-z0-9]* -> type(ID);
LST_SPACE : [ \t\r\n]+ -> channel(HIDDEN);
and:
parser grammar ParmParser;
options { tokenVocab=ParmLexer; }
parse
: PARM EQUALS list EOF
;
list
: LPAREN ( value ( COMMA value )* )? RPAREN
;
value
: ID
| STRING
| key_value
| ID list
;
key_value
: ID EQUALS value
;
which will parse your example input PARM=(VAL1, 'VAL2', VAL3, KEY4=VAL4, KEY5=VAL5(XYZ), PARM=ABC, SOMETHING=ELSE) like this:
You don't have a rule (alternative) that recognizes a PARM token in your parm rule.
Bart has provided an answer using Lexer modes (and assuming that LPAREN and RPAREN always control those modes), but you can also just set up a parser rule that matches all of your keywords:
lexer grammar ParmLexer
;
SPACE: [ \t\r\n]+ -> channel(HIDDEN);
COMMA: ',';
EQUALS: '=';
LPAREN: '(';
RPAREN: ')';
PARM: 'PARM';
KW1: 'KW1';
KW2: 'KW2';
ID: ID_LITERAL;
fragment ID_LITERAL: [A-Za-z]+;
parser grammar ParmParser
;
options {
tokenVocab = ParmLexer;
}
parms: PARM EQUALS LPAREN parm (COMMA parm)* RPAREN;
parm: ((pkey = ID | kwid = kw) EQUALS)? pval = ID;
kw: PARM | KW1 | KW2;
input
"PARM=( TEST, KEY=VAL, KW2=v2, PARM=X)"
yields:
(parms PARM = ( (parm TEST) , (parm KEY = VAL) , (parm (kw KW2) = v) , (parm (kw PARM) = X) ))

Problem matching single digits when integers are defined as tokens

I'm having problem trying to get a grammar working. Here is the simplified version. The language I try to parse has expressions like these:
testing1(2342);
testing2(idfor2);
testing3(4654);
testing4[1..n];
testing5[0..1];
testing6(7);
testing7(1);
testing8(o);
testing9(n);
The problem arises when I introduce the rules for the [1..n] or [0..1] expressions. The grammar file (one of the many variations I've tried):
grammar test;
tests
: test* ;
test
: call
| declaration ;
call
: callName '(' callParameter ')' ';' ;
callName : Identifier ;
callParameter : Identifier | Integer ;
declaration
: declarationName '[' declarationParams ']' ';' ;
declarationName : Identifier ;
declarationParams
: decMin '..' decMax ;
decMin : '0' | '1' ;
decMax : '1' | 'n' ;
Integer : [0-9]+ ;
Identifier : [a-zA-Z_][a-zA-Z0-9_]* ;
WS : [ \t\r\n]+ -> skip ;
When I parse the sample with this grammar, it fails on testing7(1); and testint(9);. It matches as decMin or decMax instead of Integer or Identifier:
line 8:9 mismatched input '1' expecting {Integer, Identifier}
line 10:9 mismatched input 'n' expecting {Integer, Identifier}
I've tried many variations but I can't make it work fine.
I think your problem comes from not using lexer rules clearly defining what you want.
When you added this rule :
decMin : '0' | '1' ;
You in fact created an unnamed lexer rule that matches '0' and another one matching '1' :
UNNAMED_0_RULE : '0';
UNNAMED_1_RULE : '1';
And your parser rule became :
decMin : UNNAMED_0_RULE | UNNAMED_1_RULE ;
Problem : now, when your lexer see
testing7(1);
**it doesn't see **
callName '(' callParameter ')' ';'
anymore, it sees
callName '(' UNNAMED_1_RULE ')' ';'
and it doesn't understand that.
And that is because lexer rules are effective before the parser rules.
To solve your problem, define your lexer rules efficiently, It would probably look like that :
grammar test;
/*---------------- PARSER ----------------*/
tests
: test*
;
test
: call
| declaration
;
call
: callName '(' callParameter ')' ';'
;
callName
: identifier
;
callParameter
: identifier
| integer
;
declaration
: declarationName '[' declarationParams ']' ';'
;
declarationName
: identifier
;
declarationParams
: decMin '..' decMax
;
decMin
: INTEGER_ZERO
| INTEGER_ONE
;
decMax
: INTEGER_ONE
| LETTER_N
;
integer
: (INTEGER_ZERO | INTEGER_ONE | INTEGER_OTHERS)+
;
identifier
: LETTER_N
| IDENTIFIER
;
/*---------------- LEXER ----------------*/
LETTER_N: N;
IDENTIFIER
: [a-zA-Z_][a-zA-Z0-9_]*
;
WS
: [ \t\r\n]+ -> skip
;
INTEGER_ZERO: '0';
INTEGER_ONE: '1';
INTEGER_OTHERS: '2'..'9';
fragment N: [nN];
I just tested this grammar and it works.
The drawback is that it will cut your integers at the lexer step (cutting 1245 into 1 2 4 5 in lexer rules, and the considering the parser rule as uniting 1 2 4 and 5).
I think it would be better to be less precise and simply write :
decMin: integer | identifier;
But then it depends on what you do with your grammar...

Want to parse same structure

I would like to make ANTLR4 parse this:
FSM
name type String
state type State
Relation
name type String
And i am using this grammar :
grammar Generator;
classToGenerate:
name=Name NL
(attributes NL)+
classToGenerate| EOF;
attributes: attribute=Name WS 'type' WS type=Name;
Name: ('A'..'Z' | 'a'..'z')+ ;
WS: (' ' | '\t')+;
NL: '\r'? '\n';
I would like to read successfully, i don't know why, but each time i run my program, i get this error :
line 6:18 no viable alternative at input '<EOF>'
Any fix?
The trailing EOF is messing things up for you. Try creating a separate rule that matches the EOF token, preceded by one or more classToGenerate (the parse rule in my example):
grammar Generator;
parse
: classToGenerate+ EOF
;
classToGenerate
: name=Name NL (attributes NL)+
;
attributes
: attribute=Name WS 'type' WS type=Name
;
Name: ('A'..'Z' | 'a'..'z')+ ;
WS: (' ' | '\t')+;
NL: '\r'? '\n';
And do you really need to keep the spaces and line breaks? You could let the lexer discard them, which makes your grammar a whole lot easier to read:
grammar Generator;
parse
: classToGenerate+ EOF
;
classToGenerate
: name=Name attributes+
;
attributes
: attribute=Name 'type' type=Name
;
Name : [a-zA-Z]+;
Spaces : [ \t\r\n] -> skip;

ANTLR4 tells me: mismatched input 'little' expecting {'big', 'little'}

I have the following simple grammar:
grammar TestG;
p : pDecl+ ;
pDecl : endianDecl
| dTDecl
;
endianType : E_BIG
| E_LITTLE
;
endianDecl : 'endian' '=' endianType ';' ;
dTDecl : 'dT' '[' STRING ']' '=' ID ';' ;
STRING: '"'.*?'"' ; //Embedded quotes?
COMMENT: '#' .*? [\n\r] -> skip ; // Discard comments for now
ID : [a-zA-Z][a-zA-Z0-9_]* ;
WS : [ \t\n\r]+ -> skip ;
INT : ('0x')?[0-9]+ ; // How to handle 0xDD and ensure non zero?
E_BIG : 'big' ;
E_LITTLE : 'little' ;
When I run grun TestG p and input the following:
endian = little;
I get the following:
line 1:9 mismatched input 'little' expecting {'big', 'little'}
What have I done wrong?
Because your lexer rule for ID precedes that for E_LITTLE, your 'little' input is being lexed as an ID.
[#0,0:5='endian',<'endian'>,1:0]
[#1,7:7='=',<'='>,1:7]
[#2,9:14='little',<ID>,1:9] <== see here it's being lexed as an ID
[#3,15:15=';',<';'>,1:15]
[#4,18:17='<EOF>',<EOF>,2:0]
line 1:9 mismatched input 'little' expecting {'big', 'little'}
Moving the these lexer tokens above ID like so:
STRING: '"'.*?'"' ; //Embedded quotes?
COMMENT: '#' .*? [\n\r] -> skip ; // Discard comments for now
E_BIG : 'big' ;
E_LITTLE : 'little' ;
ID : [a-zA-Z][a-zA-Z0-9_]* ;
WS : [ \t\n\r]+ -> skip ;
INT : ('0x')?[0-9]+ ; // How to handle 0xDD and ensure non zero?
yields the correct output from your test input.
[#0,0:5='endian',<'endian'>,1:0]
[#1,7:7='=',<'='>,1:7]
[#2,9:14='little',<'little'>,1:9] <== see here being lexed correctly
[#3,15:15=';',<';'>,1:15]
[#4,18:17='<EOF>',<EOF>,2:0]
Remember, for lexer tokens, the longest match wins, but in the case of a tie, the one that appears FIRST wins. This is why you want your more specific lexer tokens at the top of the lexer token list, and the more general ones (like identifiers, strings, etc.) farther down.

Resources