ANTLR - mismatched input error

Question

I have a grammar which looks like this consisting of comment and control statements of a particular language:

Grammar:

grammar DD;

ddlist: (ddstmt| jclcomment)+;

ddstmt:        dd1 | dd2 | dd3 | dd4 ;

dd1:    JCLBEGIN ddname  DDWORD 'DUMMY';
dd2:    JCLBEGIN ddname  DDWORD 'DYNAM';
dd3:    JCLBEGIN ddname  DDWORD NAME'=' ('*'|NAME);
dd4:    JCLBEGIN ddname  DDWORD '*' inlinerec INLINESTMTEND?;

inlinerec: (INLINEDATA )+ ;
fragment INLINEDATA: (~[\r\n])*;

ddname: NAME;

jclcomment: JCLCOMMENT+;
JCLCOMMENT: COMMENTBEGIN ~[\r\n]*;

DDWORD:     'DD';

JCLBEGIN:       '//'    ;
COMMENTBEGIN:   '//*'   ;
INLINESTMTEND:  '/*'    ;

NAME  : [A-Z#] (ALPHA | NUMBER | SPECIALCHARS)*;

NUMBER: [0-9];
ALPHA: [A-Z];
SPECIALCHARS:   '#' | '@' | '$';

STRING
 : '\'' (~[\r\n"])* '\''
 | '"' (~[\r\n"])* '"'
 ;

WS     : [ \r\n] -> channel(HIDDEN);

My input is this:

//SYSIN    DD  *                                      
SORT FIELDS=COPY
INCLUDE COND
/*                                                    
//SYSPRINT DD  SYSOUT=*        
//* Comment line #1                       
//* Comment line #2
//SYSOUT   DD  SYSOUT=*                               
//CEEDUMP  DD  SYSOUT=*                               
//* Comment line #3
//SYSUDUMP DD  SYSOUT=A

When I use AntlrWorks to run this grammar with the input, I get the following error:

line 2:0 mismatched input 'SORT' expecting INLINEDATA

How can this error be resolved ?

If the parser rule dd4 creates conflicts, does it make sense to write a separate grammar for dd4 alone an then go for the writing the main grammar? Requesting to share best practices if any — ssdimmanuel

BernardK BernardK · Accepted Answer · 2017-02-07T19:01:44

1) The white space rule generates a lot of tokens :

$ echo $CLASSPATH
.:/usr/local/lib/antlr-4.6-complete.jar
$ alias grun
alias grun='java org.antlr.v4.gui.TestRig'
$ grun DD ddlist -tokens jcl.txt
[@6,11:12='DD',<'DD'>,1:11]
[@7,13:13=' ',<WS>,channel=1,1:13]
[@8,14:14=' ',<WS>,channel=1,1:14]
[@9,15:15='*',<'*'>,1:15]
[@10,16:16=' ',<WS>,channel=1,1:16]
[@11,17:17=' ',<WS>,channel=1,1:17]
[@12,18:18=' ',<WS>,channel=1,1:18]
[@13,19:19=' ',<WS>,channel=1,1:19]
[@14,20:20=' ',<WS>,channel=1,1:20]
[@15,21:21=' ',<WS>,channel=1,1:21]
[@16,22:22=' ',<WS>,channel=1,1:22]

You can consume all consecutive spaces in a single token with the + modifier (= one or more) :

WS     : [ \t\r\n]+ -> channel(HIDDEN);

[@3,11:12='DD',<'DD'>,1:11]
[@4,13:14='  ',<WS>,channel=1,1:13]
[@5,15:15='*',<'*'>,1:15]
[@6,16:54='                                      \n',<WS>,channel=1,1:16]
[@7,55:58='SORT',<NAME>,2:0]

.

2) You don't mention an important error message that occurs when compiling the grammar :

warning(125): DD.g4:12:12: implicit definition of token INLINEDATA in parser

Using an undefined token in the parser is as if you had a lexer rule :

INLINEDATA : 'INLINEDATA' ;

that is a string constant. Thus the parser rule

dd4:    JCLBEGIN ddname  DDWORD '*' inlinerec INLINESTMTEND?;

means : I expect the input stream to be :

//{a name} DD * 'INLINEDATA'

but the input is :

//SYSIN    DD  *   SORT

hence the message

line 2:0 mismatched input 'SORT' expecting INLINEDATA

.

3) My grammar for this kind of job control statement :

grammar JCL;

/* Parsing JCL, ignoring inline sysin. */

jcl
    :   jcl_card+   // good old punched cards :-)
    ;

jcl_card
    :   dd_statement
    |   COMMENT
    ;

dd_statement
    :   '//' NAME 'DD' file_type ( NL | EOF )
    ;

file_type
    :   'DUMMY'
    |   'DYNAM'
    |   NAME '=' ( '*' | NAME )
    |   '*' NL inline_sysin
    ;

inline_sysin
    :   NON_JCL_CARD* END_OF_FILE
    ;

NAME          : [A-Z#] ( LETTER | DIGIT | SPECIAL_CHARS )* ;
COMMENT       : '//*' .*? ( NL | EOF ) ;
END_OF_FILE   : '/'  {getCharPositionInLine() == 1}? '*' ;
NON_JCL_CARD  : ~'/' {getCharPositionInLine() == 1}? .*? ( NL | EOF ) ;
STRING        : '\'' .*? '\'' | '"' .*? '"' ;
NL  : [\r\n] ;
WS  : [ \t]+ -> skip ; // or -> channel(HIDDEN) to keep white space tokens

fragment DIGIT  : [0-9] ;
fragment LETTER : [A-Z] ;
fragment SPECIAL_CHARS : '#' | '@' | '$' ;

With the input

//SYSIN    DD  *     
SORT FIELDS=COPY
INCLUDE COND
any other program input @ $ ! & %
/*                  
//SYSPRINT DD  SYSOUT=*
//* Comment line #1           
//* Comment line #2
//SYSOUT   DD  SYSOUT=*     
//SYSOUT   DD  DUMMY
//SYSIN    DD  *     
 /* not end of input    
/*

it gives

$ grun JCL jcl -tokens jcl.txt
[@0,0:1='//',<'//'>,1:0]
[@1,2:6='SYSIN',<NAME>,1:2]
[@2,11:12='DD',<'DD'>,1:11]
[@3,15:15='*',<'*'>,1:15]
[@4,21:21='\n',<NL>,1:21]
[@5,22:38='SORT FIELDS=COPY\n',<NON_JCL_CARD>,2:0]
[@6,39:51='INCLUDE COND\n',<NON_JCL_CARD>,3:0]
[@7,52:85='any other program input @ $ ! & %\n',<NON_JCL_CARD>,4:0]
[@8,86:87='/*',<END_OF_FILE>,5:0]
[@9,106:106='\n',<NL>,5:20]
...
@17,131:161='//* Comment line #1           \n',<COMMENT>,7:0]
...
[@31,232:233='//',<'//'>,11:0]
[@32,234:238='SYSIN',<NAME>,11:2]
[@33,243:244='DD',<'DD'>,11:11]
[@34,247:247='*',<'*'>,11:15]
[@35,253:253='\n',<NL>,11:21]
[@36,254:278=' /* not end of input    \n',<NON_JCL_CARD>,12:0]
[@37,279:280='/*',<END_OF_FILE>,13:0]
[@38,281:280='<EOF>',<EOF>,13:2]

.

Give a try to the -gui option to display the parse tree :

$ grun JCL jcl -gui jcl.txt

ANTLR - mismatched input error

1 Answers