How to separate out start and end tags of a block in ANTLR

Question

I am currently using the XML lexer grammar defined here:

https://github.com/antlr/grammars-v4/blob/master/xml/XMLLexer.g4

With the given input, I get the following lexer events generated:

XML input:

<item>
  <![CDATA[
    My CDATA Block
  ]]>
</item>

Lexer output

[@-1,0:0='<',<7>,1:0]
[@-1,1:4='item',<16>,1:1]
[@-1,5:5='>',<10>,1:5]
[@-1,6:8='\n  ',<6>,1:6]
[@-1,9:42='<![CDATA[\n    My CDATA Block\n  ]]>',<2>,2:2]
[@-1,43:43='\n',<6>,4:5]
[@-1,44:44='<',<7>,5:0]
[@-1,45:45='/',<13>,5:1]
[@-1,46:49='item',<16>,5:2]
[@-1,50:50='>',<10>,5:6]

However, I would like to separate the '<[[CDATA' start tag and ']]>' end tag from the CDATA event so that I can have an event that contains only the contents of this block. I have tried the following grammar which almost works except that because the CDATA block uses a non-greedy match, every single character in the block creates a new event:

New grammar:

https://gist.github.com/alexspurling/2e243b1c806a4482697700ea1f686d44

(Diff: https://gist.github.com/alexspurling/2e243b1c806a4482697700ea1f686d44/revisions)

Output:

[@-1,0:0='<',<6>,1:0]
[@-1,1:4='item',<15>,1:1]
[@-1,5:5='>',<9>,1:5]
[@-1,6:8='\n  ',<5>,1:6]
[@-1,9:17='<![CDATA[',<2>,2:2]
[@-1,18:18='\n',<19>,2:11]
[@-1,19:19=' ',<19>,3:0]
[@-1,20:20=' ',<19>,3:1]
[@-1,21:21=' ',<19>,3:2]
[@-1,22:22=' ',<19>,3:3]
[@-1,23:23='M',<19>,3:4]
[@-1,24:24='y',<19>,3:5]
[@-1,25:25=' ',<19>,3:6]
[@-1,26:26='C',<19>,3:7]
[@-1,27:27='D',<19>,3:8]
[@-1,28:28='A',<19>,3:9]
[@-1,29:29='T',<19>,3:10]
[@-1,30:30='A',<19>,3:11]
[@-1,31:31=' ',<19>,3:12]
[@-1,32:32='B',<19>,3:13]
[@-1,33:33='l',<19>,3:14]
[@-1,34:34='o',<19>,3:15]
[@-1,35:35='c',<19>,3:16]
[@-1,36:36='k',<19>,3:17]
[@-1,37:37='\n',<19>,3:18]
[@-1,38:38=' ',<19>,4:0]
[@-1,39:39=' ',<19>,4:1]
[@-1,40:42=']]>',<18>,4:2]
[@-1,43:43='\n',<5>,4:5]
[@-1,44:44='<',<6>,5:0]
[@-1,45:45='/',<12>,5:1]
[@-1,46:49='item',<15>,5:2]
[@-1,50:50='>',<9>,5:6]

My desired output would be:

[@-1,0:0='<',<7>,1:0]
[@-1,1:4='item',<16>,1:1]
[@-1,5:5='>',<10>,1:5]
[@-1,6:8='\n  ',<6>,1:6]
[@-1,9:42='<![CDATA[',<2>,2:2]
[@-1,9:42='\n    My CDATA Block\n  ',<19>,2:2]
[@-1,9:42=']]>',<18>,2:2]
[@-1,43:43='\n',<6>,4:5]
[@-1,44:44='<',<7>,5:0]
[@-1,45:45='/',<13>,5:1]
[@-1,46:49='item',<16>,5:2]
[@-1,50:50='>',<10>,5:6]

How can I change the grammar to achieve this?

Answer 1

You should change

mode IN_CDATA;

CDATA_CLOSE   : ']]>'                   -> popMode;
CDATA_CONTENT : .+? ;

to

mode IN_CDATA;

CDATA_CLOSE   : ']]>'                   -> popMode;
CDATA_CONTENT : (~']'+? | (']'+? (~']' ~']' ~'>')?));