Perl Regex: How to parse string from " to" without \"? - regex

Question

I have to parse current line "abc\",","\"," by regex in Perl,
and get this result "abc\"," and "\","
I do this
while (/(\s*)/gc) {
if (m{\G(["])([^\1]+)\1,}gc){
say $2;
}
}
but it is wrong, because this regexp go to the last ",
My question is, How can I jump over this \" and stop on first ", ?

Answer 1

The following program performs matches according to your specification:
while (<>) {
#arr = ();
while (/("(?:\\"|[^"])*")/) {
push #arr, $1;
$_ = $';
}
print join(' ', #arr), "\n";
}
Input file input.txt:
"abc", "def"
"abc\",","\","
Output:
$ ./test.pl < input.txt
"abc" "def"
"abc\"," "\","
It can be improved to match more strictly because in this form a lot of input is possible that is maybe not desirable, but it serves as a first pointer. Additionally, it is better matching a CSV file with the corresponding module and not with regular expressions, but you have not stated if your input is really a CSV file.

Answer 2

Don't reinvent the wheel. If you have CSV, use a CSV parser.
use Text::CSV_XS qw( );
my $string = '"abc\",","\","';
my $csv = Text::CSV_XS->new ({ binary => 1, auto_diag => 1 });
$csv->parse($_)
my #fields = $csv->fields();

Answer 3

Regexes aren't the best tool for this task. The standard Text::ParseWords module does this easily.
#!/usr/bin/perl
use strict;
use warnings;
use 5.010;
use Text::ParseWords;
my $line = '"abc\",","\","';
my #fields = parse_line(',', 1, $line);
for (0 .. $#fields) {
say "$_: $fields[$_]"
}
The output is:
0: "abc\","
1: "\","

Answer 4

split /(?<!\\)",(?<!\\)"/, $_
(preceded by cleaning the boundary of $_ with s/^"// && s/"$//; because enclosing external quotes didn't need to be in the definition of the input string, but you have them)
returns directly the array you want (without the need of external loop as the loop is inside the core perl function split, you might add \s* surrounding the comma according to how the string might be provided).
..but (actually just a note as you didn't mention) there could be a deeper case
If you have \" meaning " you possibly have also \\ meaning \, so you might have \\\" and \\", the last one (more generally an even number of \ preceding ") is complicate with one line regexp because look-behind is implemented for fixed size, and the unsupported regexp form (?<!\\(?:\\\\)*)" which would potentially get well also a string delimiter after backslash not intending as escape quote \" from the sequence \\", is inapplicable and a less efficient code that mine would be required, but again this marginal consideration is about the case that \\ has to be hypothetically interpreted too.