Parsing string in multiline data with positive lookbehind - perl

Question

I am trying to parse data like:
header1
-------
var1 0
var2 5
var3 9
var6 1
header2
-------
var1 -3
var3 5
var5 0
Now I want to get e.g. var3 for header2. Whats the best way to do this?
So far I was parsing my files line-by-line via
open(FILE,"< $file");
while (my $line = <FILE>){
# do stuff
}
but I guess it's not possible to handle multiline parsing properly.
Now I am thinking to parse the file at once but wasn't successful so far...
my #Input;
open(FILE,"< $file");
while (<FILE>){ #Input = <FILE>; }
if (#Input =~ /header2/){
#...
}

Answer 1

The easier way to handle this is "paragraph mode".
local $/ = "";
while (<>) {
my ($header, $body) =~ /^([^\n]*)\n-+\n(.*)/s
or die("Bad data");
my #data = map [ split ], split /\n/, $body;
# ... Do something with $header and #data ...
}
The same can be achieved without messing with $/ as follows:
my #buf;
while (1) {
my $line = <>;
$line =~ s/\s+\z// if !defined($line);
if (!length($line)) {
if (#buf) {
my $header = shift(#buf);
shift(#buf);
my #data = map [ split ], splice(#buf);
# ... Do something with $header and #data ...
}
last if !defined($line);
next;
}
push #buf, $line;
}
(In fact, the second snippet includes a couple of small improvements over the first.)
Quick comments on your attempt:
The while loop is useless because #Input = <FILE> places the remaining lines of the file in #Input.
#Input =~ /header2/ matches header2 against the stringification of the array, which is the stringification of the number of elements in #Input. If you want to check of an element of #Input contains header2, will you will need to loop over the elements of #Inputs and check them individually.

Answer 2

while (<FILE>){ #Input = <FILE>; }
This doesn't make much sense. "While you can read a record from FILE, read all of the data on FILE into #Input". I think what you actually want is just:
my #Input = <FILE>;
if (#Input =~ /header2/){
This is quite strange too. The binding operator (=~) expects scalar operands, so it evaluates both operands in scalar context. That means #Input will be evaluated as the number of elements in #Input. That's an integer and will never match "header2".
A couple of approaches. Firstly a regex approach.
#!/usr/bin/perl
use strict;
use warnings;
use feature 'say';
my $file = 'file';
open my $fh, '<', $file or die $!;
my $data = join '', <$fh>;
if ($data =~ /header2.+var3 (.+?)\n/s) {
say $1;
} else {
say 'Not found';
}
The key to this is the /s on the m// operator. Without it, the two dots in the regex won't match newlines.
The other approach is more of a line by line parser.
#!/usr/bin/perl
use strict;
use warnings;
use feature 'say';
my $file = 'file';
open my $fh, '<', $file or die $!;
my $section = '';
while (<$fh>) {
chomp;
# if the line all word characters,
# then we've got a section header.
if ($_ !~ /\W/) {
$section = $_;
next;
}
my ($key, $val) = split;
if ($section eq 'header2' and $key eq 'var3') {
say $val;
last;
}
}
We read the file a line at a time and make a note of the section headers. For data lines, we split on whitespace and check to see if we're in the right section and have the right key.
In both cases, I've switched to using a more standard approach (lexical filehandles, 3-arg open(), or die $!) for opening the file.