Parsing plain text database into data structures - regex

Question

I have a 100MB plain text database file which I would like to parse and convert into datastructure for easy access. The environment is perl and cygwin. Since we receive the plain text file with data from third party, I am not able to use any existing parser like xml or google protocol buffers.
Text file looks like below.
Class=Instance1
parameterA = <val>
parameterB = <val>
parameterC = <val>
ref = Instance2
Class=Instance2
parameterA = <val>
parameterB = <val>
parameterC = <val>
The file contains a huge number class variants.
What would be the best option to parse this ? Will yacc/lex help me or should i write my own perl parser ?

Answer 1

This should do the trick. It auto-detects the line ending by checking the first one, and the assumption here is a record is separated by a blank line.
Within each record, key/value pairs are assumed to be joined with an equal sign (=), and maybe some whitespace.
Here's my code:
#!/bin/env perl
use strict;
use warnings;
use Data::Dumper;
use Getopt::Long;
my $db_file;
GetOptions(
'file=s' => \$db_file,
);
sub detect_line_ending {
my ($fh) = #_;
my $line = <$fh>;
# Rewind to the beginning
seek($fh, 0, 0);
my ($ending) = $line =~ m/([\f\n\r]+$)/s;
return $ending;
}
sub process_chunk {
my ($chunk, $line_ending) = #_;
my #lines = split(/$line_ending/, $chunk);
my $section = {};
foreach my $line (#lines) {
my ($key, $value) = split(/[ \t]*=[ \t]*/, $line, 2);
$section->{$key} = $value;
}
return $section;
}
sub read_db_file {
my ($file) = #_;
my $data = [];
open (my $fh, '<', $file) or die $!;
my $line_ending = detect_line_ending($fh);
{
local $/ = $line_ending.$line_ending;
while (my $chunk = <$fh>) {
chomp $chunk;
my $section = process_chunk($chunk, $line_ending);
push #$data, $section;
}
}
close $fh;
return $data;
}
print Dumper read_db_file($db_file);

Answer 2

Is this what you want?
#!/usr/bin/perl
use Data::Dumper;
use Modern::Perl;
my %classes;
my $current;
while(<DATA>) {
chomp;
if (/^Class\s*=\s*(\w+)/) {
$classes{$1} = {};
$current = $1;
} elsif (/^(\w+)\s*=\s*(.+)$/) {
$classes{$current}{$1} = $2;
}
}
say Dumper\%classes;
Output:
$VAR1 = {
'Instance2' => {
'parameterC' => '<val>',
'parameterB' => '<val>',
'parameterA' => '<val>'
},
'Instance1' => {
'parameterC' => '<val>',
'ref' => 'Instance2',
'parameterB' => '<val>',
'parameterA' => '<val>'
}
};