Matching line continuation in shell style comment

[kevemueller]

What is your question?

I am trying to build a grammar for bmake style Makefiles. In my first run I handled line continuation outside of the grammar using a generator and then fed the grammar only single lines. I would like to improve on that by handling line continuation in the grammar and feed the whole file to the grammar.

I am struggling with the most simple thing as the SH_COMMENT:

# this is a comment \
that spans multiple \
lines

I am working now refining this, but can't get the continuation handling right. Continuation is a backslash directly followed by new line. Both need to be consumed and the rest of the line appended to the comment.

start: bmake_comment

bmake_comment: SH_COMMENT ((CONT /[^\n]*\n/)+|NEWLINE)

CONT: "\\" NEWLINE

%import common.SH_COMMENT
%import common.NEWLINE

[erezsh]

The comment should be one terminal. Meaning rename it to BMAKE_COMMENT.

[kevemueller]

Hi Erez,
thank you for helping a beginner like me.

I have updated the grammar to be

start: BMAKE_COMMENT

BMAKE_COMMENT: SH_COMMENT (_CONT /[^\n\\]*/)* NEWLINE

_CONT: "\\" NEWLINE

SH_COMMENT: /#[^\n\\]*/
#%import common.SH_COMMENT
%import common.NEWLINE

But I had to add the \\ into the regexp of both the SH_COMMENT as well as the line coming after the continuation. This is not precise as the comment may have \ anywhere inside it not bearing any special meaning.
Without this, I could not get the CONT match, as any preceding regexp would eat the \. I tried using with (?!\\), but I could not get that right either.

How can I get rid of adding \\ to the regexp and allow \ to appear inside the comment, and treat it special only at the end of the line?

[kevemueller]

Ok. So I got the negative lookbehind in place:

start: BMAKE_COMMENT*

BMAKE_COMMENT: SH_COMMENT CONT* NEWLINE

CONT: /\\/ NEWLINE /[^\n]*(?<!\\)/
SH_COMMENT: /#[^\n]*(?<!\\)/

#%import common.SH_COMMENT
%import common.NEWLINE

This will not treat \ inside the comment any special.

Is this the best way to do it?

Bonus problem, how do I get the '\\n' out of the matched string. I tried _CONT and %ignore CONT, but both left the continuation characters in the parse result.