Search code examples
regexrakumultiset

Perl6: Confused about BagHash/Matching

I'm attempting to count matches of a regex using a BagHash, and getting odd results. 

my $fh = open "versions.txt";
my $versions = BagHash.new();

while (defined my $line = $fh.get) {
    my $last = '';
    if $line ~~ /(\d+)\.?(\d*)/ {
        say 'match ' ~ $/[0];
        if $last !eq  $/[0] {
            say 'not-same: ' ~ $/[0];
            $versions{$/[0]}++
        }
        $last = $/[0];
    }
    else {
        $last = '';
   }

}

say 'count: ' ~ $versions.elems;

Output is: 

match 234
not-same: 234
match 999
not-same 999
count: 1 # I expect 2 here.

The test case I'm working with is: 

version history thingy

version=234.234
version=999

What am I missing?

Solution

  • You are resetting $last with each iteration. Also, don't trust say. It's meant to be used to avoid flooding a terminal or logfile with infinite lists. Use dd (Rakudo internal) or a module to dump debug output. If you would have used dd would would have seen that $/[0] contains a Match, a complex structure that is not suited to generate Hash keys.

    # my @lines = slurp('version.txt');
my @lines = ('version=234.234', 'version=999');
my BagHash $versions.=new;
for @lines {
    ENTER my $last = '';
    if .Str ~~ /(\d+) '.'? (\d*)/ {
        $versions{$0.Str}++ if $last ne $0.Str;
        $last = $0.Str
    }else{
        $last = ''
    }
};

dd $versions;
# OUTPUT«BagHash $versions = ("234"=>1,"999"=>1).BagHash␤»

    The whole point of BagHash is that it's constructor will do the counting for you. If you supply lazy lists all the way down, this can be fairly efficient.

    my @lines = ('version=234.234', 'version=999');
dd BagHash.new(@lines».split('=')».[1]);
# OUTPUT«("234.234"=>1,"999"=>1).BagHash␤»