Parser in Perl

MTK358 · 01-31-2010, 07:24 PM

I can separate x and y perfectly, I just can't separate the green and red parts. I haven't thought of using Text::Balanced for this, I will try and report back.

Sergei Steshenko · 01-31-2010, 07:27 PM

Quote:

Originally Posted by MTK358

I can separate x and y perfectly, I just can't separate the green and red parts. I haven't thought of using Text::Balanced for this, I will try and report back.

???????????????????????????????????????????????????????????????

Then why/what for have we been discussing Text::Balanced in the first place ?

And, of course, just one pair (no nesting) of any type of brackets can be parsed by regular expression, but don't reinvent the wheel - Text::Balanced does this already.

MTK358 · 01-31-2010, 07:33 PM

This seems to work:

Code:

sub process_pstruct() {
	my @parts = extract_bracketed($_[0], '{}', '[^{]*');
	
	$parts[0] =~ s/{([^}]*)}/\1/;
	my @members = split(/\s*;\s*/, $parts[0]);
	
	my @parents = split(/\s+/, $parts[2]);
	shift(@parents);
	my $name = shift(@parents);
	
	print "#define $name\_PSTRUCT_TEMPLATE ";
	for $parent (@parents) {
		print "$parent\_PSTRUCT_TEMPLATE "
	}
	print "\\\n";
	for $member (@members) {
		if($member !~ /^\s*$/) {
			$member =~ s/\s*([^\s]*)\s*/\1/;
			print "\t$member;\\\n";
		}
	}
	print "\n";
	
	print "struct $name {\n";
	print "\t$name\_PSTRUCT_TEMPLATE\n";
	print "}\n";
	print "typedef struct $name $name;\n";
}

MTK358 · 01-31-2010, 07:35 PM

Quote:

Originally Posted by Sergei Steshenko

???????????????????????????????????????????????????????????????

Then why/what for have we been discussing Text::Balanced in the first place ?

And, of course, just one pair (no nesting) of any type of brackets can be parsed by regular expression, but don't reinvent the wheel - Text::Balanced does this already.

This is a different situation here then the one I was discussing earlier. Here I want to split the string at the first left bracket and don't care about nesting, before I was talking about finding the location of the corresponding right bracket.

And I still will need to do that to find the text to feed into process_pstruct().

And then how do I replace the original text with the one retured by process_pstruct()?

Sergei Steshenko · 01-31-2010, 09:05 PM

Quote:

Originally Posted by MTK358

This is a different situation here then the one I was discussing earlier. Here I want to split the string at the first left bracket and don't care about nesting, before I was talking about finding the location of the corresponding right bracket.

And I still will need to do that to find the text to feed into process_pstruct().

And then how do I replace the original text with the one retured by process_pstruct()?

A general answer to the item in bold:

if you use regular expressions, there are predefined Perl variables which hold the part before match, the matching part and, I think, the remainder, but remainder can always be constructed from the whole test, part before match, and match; anyway, see below on 'length' and 'substr';
if you are using Text::Balanced, you definitely have $prefix, $extracted, $remainder - conceptually the same as above;
'length' (perldoc -f length) returns string length;
'substr' (perldoc -f substr) can be used as LHS of assignment, so any part of string can be replaced by something else.

MTK358 · 02-01-2010, 07:16 AM

Are you saying that modifying extract_bracketed()'s output will modify the original string?

And I already made process_pstruct() return a string instead of printing to stdout.

Sergei Steshenko · 02-01-2010, 08:18 AM

Quote:

Originally Posted by MTK358

Are you saying that modifying extract_bracketed()'s output will modify the original string?

...

No, I'm not saying that.

I'm saying that having the original text and extracted part one can then replace the extracted part with something else in original text.

MTK358 · 02-01-2010, 08:43 AM

Don't you need to know the index, too?

A regex that I think would match a pstruct definition would be /pstruct[^}]+}/.

Sergei Steshenko · 02-01-2010, 01:15 PM

Quote:

Originally Posted by MTK358

Don't you need to know the index, too?

A regex that I think would match a pstruct definition would be /pstruct[^}]+}/.

As I said, 'length' can be used, and it returns in this case index - if length of prefix is taken.

Sergei Steshenko · 02-01-2010, 01:16 PM

Quote:

Originally Posted by MTK358

...
A regex that I think would match a pstruct definition would be /pstruct[^}]+}/.

Not quite - think of

Code:

FooBarpstruct { ... }

case.

Sergei Steshenko · 02-01-2010, 01:20 PM

Quote:

Originally Posted by Sergei Steshenko

Not quite - think of

Code:

FooBarpstruct { ... }

case.

And, actually, think also of

Code:

FooBarpstructDooDah { ... }

case.

MTK358 · 02-01-2010, 02:07 PM

How would you fix that?

Sergei Steshenko · 02-01-2010, 02:45 PM

Quote:

Originally Posted by MTK358

How would you fix that?

You know my recipe - read. And here is clue for you: word boundary.

perldoc perlretut
perldoc perlre

are your friends; you need less than 350 lines of patience.

MTK358 · 02-01-2010, 04:00 PM

Good, but how do iterate over all the instances of pstruct { ... } in the file using regex matching?

MTK358 · 02-01-2010, 04:20 PM

I figured that out, but now how do I replace the match with the function's return?