NAME

perlunitut - Perl Unicode Tutorial


DESCRIPTION

The days of just flinging strings around are over. It's well established that modern programs need to be capable of communicating funny accented letters, and things like euro symbols. This means that programmers need new habits. It's easy to program Unicode capable software, but it does require discipline to do it right.

There's a lot to know about character sets, and text encodings. It's probably best to spend a full day learning all this, but the basics can be learned in minutes.

These are not the very basics, though. It is assumed that you already know the difference between bytes and characters, and realise (and accept!) that there are many different character sets and encodings, and that your program has to be explicit about them. Recommended reading is "The Absolute Minimum Every Software Developer Absolutely, Positively Must Know About Unicode and Character Sets (No Excuses!)" by Joel Spolsky, at http://joelonsoftware.com/articles/Unicode.html.

This tutorial speaks in rather absolute terms, and provides only a limited view of the wealth of character string related features that Perl has to offer. For most projects, this information will probably suffice.

Definitions

It's important to set a few things straight first. This is the most important part of this tutorial. This view may conflict with other information that you may have found on the web, but that's mostly because many sources are wrong.

You may have to re-read this entire section a few times...

Unicode

Unicode is a character set with room for lots of characters. The ordinal value of a character is called a code point.

There are many, many code points, but computers work with bytes, and a byte can have only 256 values. Unicode has many more characters, so you need a method to make these accessible.

Unicode is encoded using several competing encodings, of which UTF-8 is the most used. In a Unicode encoding, multiple subsequent bytes can be used to store a single code point, or simply: character.

UTF-8

UTF-8 is a Unicode encoding. Many people think that Unicode and UTF-8 are the same thing, but they're not. There are more Unicode encodings, but much of the world has standardized on UTF-8.

UTF-8 treats the first 128 codepoints, 0..127, the same as ASCII. They take only one byte per character. All other characters are encoded as two or more (up to six) bytes using a complex scheme. Fortunately, Perl handles this for us, so we don't have to worry about this.

Text strings (character strings)

Text strings, or character strings are made of characters. Bytes are irrelevant here, and so are encodings. Each character is just that: the character.

On a text string, you would do things like:

    $text =~ s/foo/bar/;
    if ($string =~ /^\d+$/) { ... }
    $text = ucfirst $text;
    my $character_count = length $text;

The value of a character (ord, chr) is the corresponding Unicode code point.

Binary strings (byte strings)

Binary strings, or byte strings are made of bytes. Here, you don't have characters, just bytes. All communication with the outside world (anything outside of your current Perl process) is done in binary.

On a binary string, you would do things like:

    my (@length_content) = unpack "(V/a)*", $binary;
    $binary =~ s/\x00\x0F/\xFF\xF0/;  # for the brave :)
    print {$fh} $binary;
    my $byte_count = length $binary;

Encoding

Encoding (as a verb) is the conversion from text to binary. To encode, you have to supply the target encoding, for example iso-8859-1 or UTF-8. Some encodings, like the iso-8859 ("latin") range, do not support the full Unicode standard; characters that can't be represented are lost in the conversion.

Decoding

Decoding is the conversion from binary to text. To decode, you have to know what encoding was used during the encoding phase. And most of all, it must be something decodable. It doesn't make much sense to decode a PNG image into a text string.

Internal format

Perl has an internal format, an encoding that it uses to encode text strings so it can store them in memory. All text strings are in this internal format. In fact, text strings are never in any other format!

You shouldn't worry about what this format is, because conversion is automatically done when you decode or encode.

Your new toolkit

Add to your standard heading the following line:

    use Encode qw(encode decode);

Or, if you're lazy, just:

    use Encode;

I/O flow (the actual 5 minute tutorial)

The typical input/output flow of a program is:

    1. Receive and decode
    2. Process
    3. Encode and output

If your input is binary, and is supposed to remain binary, you shouldn't decode it to a text string, of course. But in all other cases, you should decode it.

Decoding can't happen reliably if you don't know how the data was encoded. If you get to choose, it's a good idea to standardize on UTF-8.

    my $foo   = decode('UTF-8', get 'http://example.com/');
    my $bar   = decode('ISO-8859-1', readline STDIN);
    my $xyzzy = decode('Windows-1251', $cgi->param('foo'));

Processing happens as you knew before. The only difference is that you're now using characters instead of bytes. That's very useful if you use things like substr, or length.

It's important to realize that there are no bytes in a text string. Of course, Perl has its internal encoding to store the string in memory, but ignore that. If you have to do anything with the number of bytes, it's probably best to move that part to step 3, just after you've encoded the string. Then you know exactly how many bytes it will be in the destination string.

The syntax for encoding text strings to binary strings is as simple as decoding:

    $body = encode('UTF-8', $body);

If you needed to know the length of the string in bytes, now's the perfect time for that. Because $body is now a byte string, length will report the number of bytes, instead of the number of characters. The number of characters is no longer known, because characters only exist in text strings.

    my $byte_count = length $body;

And if the protocol you're using supports a way of letting the recipient know which character encoding you used, please help the receiving end by using that feature! For example, E-mail and HTTP support MIME headers, so you can use the Content-Type header. They can also have Content-Length to indicate the number of bytes, which is always a good idea to supply if the number is known.

    "Content-Type: text/plain; charset=UTF-8",
    "Content-Length: $byte_count"

Q and A

This isn't really a Unicode tutorial, is it?

No, Perl has an abstracted interface for all supported character encodings, so this is actually a generic Encode tutorial. But many people think that Unicode is special and magical, and I didn't want to disappoint them, so I decided to call this document a Unicode tutorial.

What about binary data, like images?

Well, apart from a bare binmode $fh, you shouldn't treat them specially. (The binmode is needed because otherwise Perl may convert line endings on Win32 systems.)

Be careful, though, to never combine text strings with binary strings. If you need text in a binary stream, encode your text strings first using the appropriate encoding, then join them with binary strings. See also: "What if I don't encode?".

What about the UTF-8 flag?

Please, unless you're hacking the internals, or debugging weirdness, don't think about the UTF-8 flag at all. That means that you very probably shouldn't use is_utf8, _utf8_on or _utf8_off at all.

Perl's internal format happens to be UTF-8. Unfortunately, Perl can't keep a secret, so everyone knows about this. That is the source of much confusion. It's better to pretend that the internal format is some unknown encoding, and that you always have to encode and decode explicitly.

When should I decode or encode?

Whenever you're communicating with anything that is external to your perl process, like a database, a text file, a socket, or another program. Even if the thing you're communicating with is also written in Perl.

What if I don't decode?

Whenever your encoded, binary string is used together with a text string, Perl will assume that your binary string was encoded with ISO-8859-1, also known as latin-1. If it wasn't latin-1, then your data is unpleasantly converted. For example, if it was UTF-8, the individual bytes of multibyte characters are seen as separate characters, and then again converted to UTF-8. Such double encoding can be compared to double HTML encoding (>), or double URI encoding (%253E).

This silent implicit decoding is known as "upgrading". That may sound positive, but it's best to avoid it.

What if I don't encode?

Your text string will be sent using the bytes in Perl's internal format. In some cases, Perl will warn you that you're doing something wrong, with a friendly warning:

    Wide character in print at example.pl line 2.

Because the internal format is often UTF-8, these bugs are hard to spot, because UTF-8 is usually the encoding you wanted! But don't be lazy, and don't use the fact that Perl's internal format is UTF-8 to your advantage. Encode explicitly to avoid weird bugs, and to show to maintenance programmers that you thought this through.

Is there a way to automatically decode or encode?

If all data that comes from a certain handle is encoded in exactly the same way, you can tell the PerlIO system to automatically decode everything, with the encoding layer. If you do this, you can't accidentally forget to decode or encode anymore, on things that use the layered handle.

You can provide this layer when opening the file:

    open my $fh, '>:encoding(UTF-8)', $filename;  # auto encoding on write
    open my $fh, '<:encoding(UTF-8)', $filename;  # auto decoding on read

Or if you already have an open filehandle:

    binmode $fh, ':encoding(UTF-8)';

Some database drivers for DBI can also automatically encode and decode, but that is typically limited to the UTF-8 encoding, because they cheat.

Cheat?! Tell me, how can I cheat?

Well, because Perl's internal format is UTF-8, you can just skip the encoding or decoding step, and manipulate the UTF-8 flag directly.

Instead of :encoding(UTF-8), you can simply use :utf8. This is widely accepted as good behavior.

Instead of decode and encode, you could use _utf8_on and _utf8_off. But this is, contrary to :utf8, considered bad style.

There are some shortcuts for oneliners; see -C in perlrun.

What if I don't know which encoding was used?

Do whatever you can to find out, and if you have to: guess. (Don't forget to document your guess with a comment.)

You could open the document in a web browser, and change the character set or character encoding until you can visually confirm that all characters look the way they should.

There is no way to reliably detect the encoding automatically, so if people keep sending you data without charset indication, you may have to educate them.

Can I use Unicode in my Perl sources?

Yes, you can! If your sources are UTF-8 encoded, you can indicate that with the use utf8 pragma.

    use utf8;

This doesn't do anything to your input, or to your output. It only influences the way your sources are read. You can use Unicode in string literals, in identifiers (but they still have to be "word characters" according to \w), and even in custom delimiters.

Data::Dumper doesn't restore the UTF-8 flag; is it broken?

No, Data::Dumper's Unicode abilities are as they should be. There have been some complaints that it should restore the UTF-8 flag when the data is read again with eval. However, you should really not look at the flag, and nothing indicates that Data::Dumper should break this rule.

Here's what happens: when Perl reads in a string literal, it sticks to 8 bit encoding as long as it can. (But perhaps originally it was internally encoded as UTF-8, when you dumped it.) When it has to give that up because other characters are added to the text string, it silently upgrades the string to UTF-8.

If you properly encode your strings for output, none of this is of your concern, and you can just eval dumped data as always.

How can I determine if a string is a text string or a binary string?

You can't. Some use the UTF-8 flag for this, but that's misuse, and makes well behaved modules like Data::Dumper look bad. The flag is useless for this purpose, because it's off when an 8 bit encoding (by default ISO-8859-1) is used to store the string.

This is something you, the programmer, has to keep track of; sorry. You could consider adopting a kind of "Hungarian notation" to help with this.

How do I convert from encoding FOO to encoding BAR?

By first converting the FOO-encoded byte string to a text string, and then the text string to a BAR-encoded byte string:

    my $text_string = decode('FOO', $foo_string);
    my $bar_string  = encode('BAR', $text_string);

or by skipping the text string part, and going directly from one binary encoding to the other:

    use Encode qw(from_to);
    from_to($string, 'FOO', 'BAR');  # changes contents of $string

or by letting automatic decoding and encoding do all the work:

    open my $foofh, '<:encoding(FOO)', 'example.foo.txt';
    open my $barfh, '>:encoding(BAR)', 'example.bar.txt';
    print { $barfh } $_ while <$foofh>;

What about the use bytes pragma?

Don't use it. It makes no sense to deal with bytes in a text string, and it makes no sense to deal with characters in a byte string. Do the proper conversions (by decoding/encoding), and things will work out well: you get character counts for decoded data, and byte counts for encoded data.

use bytes is usually a failed attempt to do something useful. Just forget about it.

What are decode_utf8 and encode_utf8?

These are alternate syntaxes for decode('utf8', ...) and encode('utf8', ...).

What's the difference between UTF-8 and utf8?

UTF-8 is the official standard. utf8 is Perl's way of being liberal in what it accepts. If you have to communicate with things that aren't so liberal, you may want to consider using UTF-8. If you have to communicate with things that are too liberal, you may have to use utf8. The full explanation is in Encode.

UTF-8 is internally known as utf-8-strict. This tutorial uses UTF-8 consistently, even where utf8 is actually used internally, because the distinction can be hard to make, and is mostly irrelevant.

Okay, if you insist: the "internal format" is utf8, not UTF-8. (When it's not some other encoding.)

I lost track; what encoding is the internal format really?

It's good that you lost track, because you shouldn't depend on the internal format being any specific encoding. But since you asked: by default, the internal format is either ISO-8859-1 (latin-1), or utf8, depending on the history of the string.

Perl knows how it stored the string internally, and will use that knowledge when you encode. In other words: don't try to find out what the internal encoding for a certain string is, but instead just encode it into the encoding that you want.

What character encodings does Perl support?

To find out which character encodings your Perl supports, run:

    perl -MEncode -le "print for Encode->encodings(':all')"

Which version of perl should I use?

Well, if you can, upgrade to the most recent, but certainly 5.8.1 or newer. This tutorial is based on the status quo as of 5.8.7.

You should also check your modules, and upgrade them if necessary. For example, HTML::Entities requires version >= 1.32 to function correctly, even though the changelog is silent about this.


SUMMARY

Decode everything you receive, encode everything you send out. (If it's text data.)


ACKNOWLEDGEMENTS

Thanks to Johan Vromans from Squirrel Consultancy. His UTF-8 rants during the Amsterdam Perl Mongers meetings got me interested and determined to find out how to use character encodings in Perl in ways that don't break easily.

Thanks to Gerard Goossen from TTY. His presentation "UTF-8 in the wild" (Dutch Perl Workshop 2006) inspired me to publish my thoughts and write this tutorial.

Thanks to the people who asked about this kind of stuff in several Perl IRC channels, and have constantly reminded me that a simpler explanation was needed.

Thanks to the people who reviewed this document for me, before it went public. They are: Benjamin Smith, Jan-Pieter Cornet, Johan Vromans, Lukas Mai, Nathan Gray.


AUTHOR

Juerd Waalboer <juerd@cpan.org>


SEE ALSO

perlunicode, perluniintro, Encode