parsing multiline nested tokens from a file in perl - regex

Question

I have a file that looks line this:
Alpha 27600
Beta 1
Charlie true
BEGIN Delta
BEGIN Epsilon Setting High Hook 50 END
BEGIN Foxtrot Corp 71 END
BEGIN "Jelly Bean" Corp 88 END
END
BEGIN Hotel
Height 25
Lawn 85
END
Basically it is several key/value pairs separated by one or more spaces. The tricky part is the BEGIN/END blocks that might be nested and might span multiple lines. I need to go through the file and take some action based on what follows the BEGIN. For example, if it's Delta, i might need to process each of the sub BEGIN lines where as if it is Hotel, i can skip that completely.
I looked at Parse::RecDescent a little bit but wasn't sure how to make it handle the BEGIN/END situation properly. Speed isn't as important as having a an easier to understand and maintain solution.
Any suggestions?
EDIT: I liked Miller's solution, but then looking over the data realized why I didn't just split on whitespace. Some of the labels have whitespace in them. Added "Jelly Bean" label in above data file to reflect that.

Answer 1

Just parse the whole data structure, and filter out sections you don't need after the fact:
use strict;
use warnings;
use Text::ParseWords;
my #tokens = parse_line( qr{\s+}, 0, do { local $/; <DATA> } );
my %hash;
my #levels = \%hash;
while ( defined( my $key = shift #tokens ) ) {
if ( $key eq 'BEGIN' ) {
push #levels, $levels[-1]{ shift #tokens } = {};
} elsif ( $key eq 'END' ) {
pop #levels;
} else {
$levels[-1]{$key} = shift #tokens;
}
}
use Data::Dump;
dd \%hash;
__DATA__
Alpha 27600
Beta 1
Charlie true
BEGIN Delta
BEGIN Epsilon Setting High Hook 50 END
BEGIN Foxtrot Corp 71 END
BEGIN "Jelly Bean" Corp 88 END
END
BEGIN Hotel
Height 25
Lawn 85
END
Outputs:
{
Alpha => 27600,
Beta => 1,
Charlie => "true",
Delta => {
"Epsilon" => { Hook => 50, Setting => "High" },
"Foxtrot" => { Corp => 71 },
"Jelly Bean" => { Corp => 88 },
},
Hotel => { Height => 25, Lawn => 85 },
}

Answer 2

Personally I'd hack something up with Parser::MGC (though perhaps I'm biased because I wrote it).
Using a nested scope of its scope_of method will easily handle those BEGIN/END markers for you.