Thanks! I see you're filtering lines beginning with # when they occur at the beginning of the file; the way I understood the OP's sample data is that the comments can occur anywhere. And my worry was that, even though in the OP's data this is probably not the case, filter-based solutions will remove lines that may actually not be comments, and I wasn't sure if there was a easy solution for this?
use warnings;
use strict;
use Data::Peek;
use Text::CSV_XS qw/csv/;
DDumper csv( in=>*DATA, escape_char=>"\\",
filter => sub { $_[1][0] !~ m/^\s*#/ });
__DATA__
# This is a comment
a,b,c
# Also a comment
x,y,z
"#not",a,comment
\#also,not,"a comment"
Output:
[
[ 'a',
'b',
'c'
],
[ ''
],
[ 'x',
'y',
'z'
],
[ ''
]
]