ANTLR Grammar to parse a Asterisk-delimited input - antlr

Question

I am attempting to use ANTLR (v4) to create a parser generator for a asterisk-delimited list encapsulated by START and END markers.
START**na**na**aa*aa*a*asdfaaa*aaDDFdasa*aaaffdda*aa*aassda*ataaaaaaaaa*a*a*aEND
Where a normal input string would be something like:
START*na*na*aa*aa*a*asdfaaa*aaDDFdasa*aaaffdda*aa*aassda*ataaaaaaaaa*a*a*aEND
I would still need to be able to allow spaces, tabs, null/empty fields (basically any character except START, END, * between the asterisks.
that includes things like ** * * *asdf fdsa* * asdf *
Here is my grammar so far:
parseIt: ENTRY ;
ENTRY : 'START*' FIELD_SET 'END' ;
fragment Delim : '*' ;
fragment Data : (ANY | WS)* ;
fragment FIELD_SET : Data (Delim Data|Delim)* ;
I can recognize simple input (like the first example I gave), but am having trouble recognizing tokens that have spaces or special characters between the asterisks.

Answer 1

I’m pretty sure you could handle this with a RegEx and capture groups, but if you really want to use ANTLR…
The following works:
grammar asterisks;
parseIt: 'START' dataItem* 'END' EOF;
dataItem: Delim Data?;
Delim : '*' ;
Data : ~[*]+ {!(
(getText().endsWith("E") && _input.LA(1) == (int) 'N' && _input.LA(2) == (int) 'D') ||
(getText().endsWith("EN") && _input.LA(1) == (int) 'D') ||
(getText().endsWith("END")))}?;
and gives the following parse tree (for you first input) (click on it to view it full size):
Unfortunately for you, the way the lexer works, a simple lexer rule like Data : ~[*]+ will preferentially match aEND over your END implied lexer rule, because the ANTLR lexer uses the rule that matches the longest sequence ion input characters, and Data : ~[*]+ matches aEND while END only matches END (ANTLR also, doesn't look ahead for token matches). As a result the rather tortured semantic predicate is the only way to disallow a token that is a stream of characters that ends with END.
(Note: Semantic predicates a target-language specific, and this predicate is for Java. Other targets would require the equivalent int that target language.)
Another approach would be to check if your input endswith(“END”), and then just remove it prior to parsing using this grammar:
grammar asterisks;
parseIt: 'START' dataItem* 'END' EOF;
dataItem: Delim Data?;
Delim : '*' ;
Data : ~[*]+;
This avoids the END token problem by just removing it from the input stream. Given that it's the very end of the stream, this might be simpler.