Quick and dirty script to measure how null-checking prevalence in Java programs.

nullitis.pl

#!/usr/bin/env perl
use strict;
use warnings;

sub locs {
    my $repo = shift;
    `sloccount $repo | grep ^java` =~ /^java:\s+(\d+)/;
    return $1;
}

sub conditionals {
    my $repo = shift;
    my $conditionals = 0;
    my $nulls = 0;

    open(my $lines, '-|', "find $repo -name '*.java' -exec cat {} \\; | egrep '\\<(if|while|for)\\>\\s*\\('")
    or die("Cannot read $repo\n");
    while(my $line = <$lines>) {
	$conditionals++;
	$nulls++ if $line =~ /\bnull\b/;
    }
    close($lines);
    return 100 if $conditionals == 0;
    return 100 * $nulls / $conditionals;
}

sub repo_age {
    my $repo = shift;
    my $timestamp = `cd $repo && git log --reverse --format=%ct . | head -n 1`;
    chomp $timestamp;
    return (time - $timestamp) / (3600 * 24 * 365);
}

sub print_stats {
    my $repo = shift;
    printf "\"%s\",%d,%d,%5.2f\n", $repo,
	locs($repo), repo_age($repo), conditionals($repo);
}

my $base = $ARGV[0];
opendir(my $d, $base) or die "Cannot read dir $base\n";
print "repo,locs,years,nullitis\n";
while(my $name = readdir $d) {
    my $path = "$base/$name";
    next if $name =~ /^\./;
    next unless -d $path;
    print_stats($path);
}
close($d);