'Wide character in subroutine entry\" - UTF-8 encoded cyrillic words as sequence of bytes

Question

I am working on an Android word game with a large dictionary -

The words (over 700 000) are kept as separate lines in a text file (and then put in an SQLite database).

To keep competitors from extracting my dictionary, I'd like to encode all words which are longer than 3 chars with md5. (I don't obfuscate short words and words with rare Russian letters ъ and э, because I'd like to list them in my app).

So here is my script which I try to run with perl v5.18.2 on Mac Yosemite:

#!/usr/bin/perl -w

use strict;
use utf8;
use Digest::MD5 qw(md5_hex);

binmode(STDIN, ":utf8");
#binmode(STDOUT, ":raw");
binmode(STDOUT, ":utf8");

while(<>) {
        chomp;
        next if length($_) < 2; # ignore 1 letter junk
        next if /жы/;           # impossible combination in Russian
        next if /шы/;           # impossible combination in Russian

        s/ё/е/g;

        #print "ORIGINAL WORD $_\tENCODED WORD: ";

        if (length($_) <= 3 || /ъ/ || /э/) { # do not obfuscate short words
                print "$_\n";                # and words with rare letters
                next;
        }

        print md5_hex($_) . "\n";            # this line crashes
}

As you can see, I have to use cyrillic letters in the source code of my Perl script - that is why I've put use utf8; on its top.

However my real problem is that length($_) reports too high values (probably reporting number of bytes instead of number of characters).

So I have tried adding:

binmode(STDOUT, ":raw");

or:

binmode(STDOUT, ":utf8");

But the script then dies with Wide character in subroutine entry at the line with print md5_hex($_).

Please help me to fix my script.

I run it as:

perl ./generate-md5.pl < words.txt > encoded.txt

and here is example words.txt data for your convenience:

а
аб
абв
абвг
абвгд
съемка

Answer 1

md5_hex expects a string of bytes for input, but you're passing a decoded string (a string of Unicode Code Points). Explicitly encode the string.

use strict;
use utf8;
use Digest::MD5;
use Encode;
# ....
# $_ is assumed to be utf8 encoded without check
print Digest::MD5::md5_hex(Encode::encode_utf8($_)),"\n";
# Conversion only when required:
print Digest::MD5::md5_hex(utf8::is_utf8($_) ? Encode::encode_utf8($_) : $_),"\n";

Answer 2

my real problem is that length($_) reports too high values

Yes, you are reading from the ARGV file handle and haven't set its encoding to UTF-8

You can use the open pragma to fix this. Instead of all your binmode statements, use

use open qw/ :std :encoding(utf8) /;

which will change the default open mode for all filehandles, including the standard ones, to :encoding(utf8)