Unicode in Python

A quick run-down of Unicode,

Its use in Python 2 and 3,

and some of the gotchas that arise.

History

A bit about where all this mess came from…

What the heck is Unicode anyway?

  • First there was chaos…
    • Different machines used different encodings – different ways of mapping binary data that the computer stores to letters.
  • Then there was ASCII – and all was good (7 bit), 127 characters
    • (for English speakers, anyway)
  • But each vendor used the top half of 8bit bytes (127-255) for different things.
    • MacRoman, Windows 1252, etc…
    • There is now “latin-1”, a 1-byte encoding suitable for European languages – but still a lot of old files around that use the old ones.
  • Non-Western European languages required totally incompatible 1-byte encodings
  • This means there was no way to mix languages with different alphabets in the same document (web page, etc.)

Enter Unicode

The Unicode idea is pretty simple:
  • One “code point” for all characters in all languages
But how do you express that in bytes?
  • Early days: we can fit all the code points in a two byte integer (65536 characters)
  • Turns out that didn’t work – 65536 is not enough for all languages. So we now need 32 bit integer to hold all of Unicode “raw” (UTC-4).
  • But it’s a waste of space to use 4 full bytes for each character, when so many don’t require that much space.
Enter “encodings”:
  • An encoding is a way to map specific bytes to a code point.
  • Each code point can be represented by one or more bytes.
  • Each encoding is different – if you don’t know the encoding, you don’t know how to interpret the bytes! (though maybe you can guess)

Unicode

A good start:

The Absolute Minimum Every Software Developer Absolutely, Positively Must Know About Unicode and Character Sets (No Excuses!)

http://www.joelonsoftware.com/articles/Unicode.html

Everything is Bytes

  • If it’s on disk or on a network, it’s bytes
  • Python provides some abstractions to make it easier to deal with bytes

Unicode is a Biggie

Actually, dealing with numbers rather than bytes is big

– but we take that for granted

Mechanics

What are strings?

Py2 strings were simply sequences of bytes. When text was one per character that worked fine.

Py3 strings (or Unicode strings in py2) are sequences of “platonic characters”.

It’s almost one code point per character – there are complications with combined characters: accents, etc – but we can mostly ignore those – you will get far thiking of a code point as a character.

Platonic characters cannot be written to disk or network!

(ANSI: one character == one byte – so easy!)

Strings vs Unicode

Python 2 had two types that let you work with text:

  • str
  • unicode

And two ways to work with binary data:

  • str
  • bytes() (and bytearray)

but:

In [86]: str is bytes
Out[86]: True

bytes is there in py2 for py3 compatibility – but it’s good for making your intentions clear, too.

py3 is more clear:

str for text byte for binary data

Unicode

The py3 string (py2 Unicode) object lets you work with characters, instead of bytes.

It has all the same methods you’d expect a string object to have.

Encoding / Decoding

If you need to deal with the actual bytes for some reason, you may need to convert between a string object and a particular set of bytes.

“encoding” is converting from a string object to bytes

“decoding” is converting from bytes to a string object

(sometimes this feels backwards…)

And can get even more confusing with py2 strings being both text and bytes!

This is actually one of the biggest differences between Python 2 and Python 3. As an ordinary user (particularly one that used English…), you may not notice – text is text, and things generally “just work”, but under the hood it is very different, and folks writing libraries for things like Internet protocols struggle with the differences.

Using Unicode in Py2

If you do need to write Python2 code, you really should use Unicode.

Here are the basics:

Built in functions

ord()
chr()
unichr()
str()
unicode()

The codecs module

import codecs
codecs.encode()
codecs.decode()
codecs.open() # better to use ``io.open``

Encoding and Decoding

(Python 2!)

Encoding: text to bytes – you get a bytes (str) object

In [17]: u"this".encode('utf-8')
Out[17]: 'this'

In [18]: u"this".encode('utf-16')
Out[18]: '\xff\xfet\x00h\x00i\x00s\x00'

Decoding bytes to text – you get a unicode object

In [2]: text =  '\xff\xfe."+"x\x00\xb2\x00'.decode('utf-16')

In [3]: type(text)
Out[3]: unicode

In [4]: print text
∮∫x²

Unicode Literals

  1. Use Unicode in your source files:
# -*- coding: utf-8 -*-

(This is only required on Py2 – the UTF-8 encoding is default for Python 3)

  1. Escape the Unicode characters:
print u"The integral sign: \u222B"
print u"The integral sign: \N{integral}"

Lots of tables of code points are available online:

One example: http://inamidst.com/stuff/unidata/

hello_unicode.py.

Using Unicode

Use unicode objects in all your code

Decode on input

Encode on output

Many packages do this for you: XML processing, databases, …

Gotcha:

Python has a default encoding (usually ascii)

In [2]: sys.getdefaultencoding()
Out[2]: 'ascii'

The default encoding will get used in unexpected places!

Using Unicode Everywhere

Python 2.6 and above have a nice feature to make it easier to use Unicode everywhere

from __future__ import unicode_literals

After running that line, the u'' is assumed

In [1]: s = "this is a regular py2 string"
In [2]: print type(s)
<type 'str'>

In [3]: from __future__ import unicode_literals
In [4]: s = "this is now a unicode string"
In [5]: type(s)
Out[5]: unicode

NOTE: You can still get py2 strings from other sources! So you still need to think about str vs unicdode

This is a really good idea if you want to write code compatible with Python2 and 3.

Encodings

What encoding should I use???

There are a lot:

http://en.wikipedia.org/wiki/Comparison_of_Unicode_encodings

But only a couple you are likely to need:

  • utf-8 (*nix)
  • utf-16 (Windows)

and of course, still the one-bytes ones.

  • ASCII
  • Latin-1

UTF-8

Probably the one you’ll use most – most common in Internet protocols (xml, JSON, etc.)

Nice properties:

  • ASCII compatible: First 127 characters are the same as ASCII
  • Any ascii string is a utf-8 string
  • Compact for mostly-English text.

Gotchas:

  • “higher” code points may use more than one byte: up to 4 for one character
  • ASCII compatible means in may work with default encoding in tests – but then blow up with real data…

UTF-16

Kind of like UTF-8, except it uses at least 16bits (2 bytes) for each character: NOT ASCII compatible.

But is still needs more than two bytes for some code points, so you still can’t process it as two bytes per character.

In C/C++ held in a “wide char” or “wide string”.

MS Windows uses UTF-16, as does (I think) Java.

UTF-16 criticism

There is a lot of criticism on the net about UTF-16 – it’s kind of the worst of both worlds:

  • You can’t assume every character is the same number of bytes
  • It takes up more memory than UTF-8

UTF-16 Considered Harmful

But to be fair:

Early versions of Unicode: everything fit into two bytes (65536 code points). MS and Java were fairly early adopters, and it seemed simple enough to just use 2 bytes per character.

When it turned out that 4 bytes were really needed, they were kind of stuck in the middle.

Latin-1

NOT Unicode:

A 1-byte per char encoding.

  • Superset of ASCII suitable for Western European languages.
  • The most common one-byte per char encoding for European text.
  • Nice property – every byte value from 0 to 255 is a valid character ( at least in Python )
  • You will never get an UnicodeDecodeError if you try to decode arbitrary bytes with latin-1.
  • And it can “round-trip” through a unicode object.
  • Useful if you don’t know the encoding – at least it won’t raise an Exception
  • Useful if you need to work with combined text+binary data.

latin1_test.py.

Unicode Docs

Python Docs Unicode HowTo:

http://docs.python.org/howto/unicode.html

“Reading Unicode from a file is therefore simple”

use io.open:

from io import open
io.open('unicode.rst', encoding='utf-8')
for line in f:
    print repr(line)

(https://docs.python.org/2/library/io.html#module-interface)

Encodings Built-in to Python:
http://docs.python.org/2/library/codecs.html#standard-encodings

Gotchas in Python 2

file names, etc:

If you pass in unicode, you get unicode

In [9]: os.listdir('./')
Out[9]: ['hello_unicode.py', 'text.utf16', 'text.utf32']

In [10]: os.listdir(u'./')
Out[10]: [u'hello_unicode.py', u'text.utf16', u'text.utf32']

Python deals with the file system encoding for you…

But: some more obscure calls don’t support unicode filenames:

os.statvfs() (http://bugs.python.org/issue18695)

Exception messages:

  • Py2 Exceptions use str when they print messages.
  • But what if you pass in a unicode object?
    • It is encoded with the default encoding.
  • UnicodeDecodeError Inside an Exception????

NOPE: it swallows it instead.

exception_test.py.

Unicode in Python 3

The “string” object is Unicode (always).

Py3 has two distinct concepts:

  • “text” – uses the str object (which is always Unicode!)
  • “binary data” – uses bytes or bytearray

Everything that’s about text is Unicode.

Everything that requires binary data uses bytes.

It’s all much cleaner.

(by the way, the recent implementations are very efficient…)

So you can pretty much ignore encodings and all that for most basic text processing. If you do find yourself needing to deal with binary data, you ay need to encode/decode stuff yourself. IN which case, Python provides an .encode() method on strings that encode the string to a bytes object with the encoding you select:

In [3]: this_in_utf16 = "this".encode('utf-16')

In [4]: this_in_utf16
Out[4]: b'\xff\xfet\x00h\x00i\x00s\x00'

And bytes objects have a .decode method that decodes the bytes and makes a string object:

In [5]: this_in_utf16.decode(‘utf-16’) Out[5]: ‘this’

It’s all quite simple an robust.

Note

During the long and painful transition from Python2 to Python3, the Unicode-always string type was a major source of complaints. There are many rants and well thought out posts about it still available on the internet. It was enough to think that Python had made a huge mistake.

But there are a couple key points to remember:

  • The primary people struggling were those that wrote (or wored with) libraries that had to deal with protocols that used both binary and text data in the same data stream.
  • As of Python 3.4 or so, the python string object had grown the features it needed to support even those ugly binary+text use cases.

For a typical user, the Python3 text model is MUCH easier to deal with and less error prone.

Exercises

Basic Unicode LAB

  • Find some nifty non-ascii characters you might use.
    • Create a unicode object with them in two different ways.
    • here is one example
  • Read the contents into unicode objects:

and / or

  • write some of the text from the first exercise to file – read that file back in.

Some Help

reference: http://inamidst.com/stuff/unidata/

NOTE: if your terminal does not support unicode – you’ll get an error trying to print. Try a different terminal or IDE, or google for a solution.

Challenge Unicode LAB

Here is an error in Python2:

In [38]: u'to \N{INFINITY} and beyond!'.decode('utf-8')
---------------------------------------------------------------------------
UnicodeEncodeError                        Traceback (most recent call last)
<ipython-input-38-7f87d44dfcfa> in <module>()
----> 1 u'to \N{INFINITY} and beyond!'.decode('utf-8')

/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/encodings/utf_8.pyc in decode(input, errors)
     14
     15 def decode(input, errors='strict'):
---> 16     return codecs.utf_8_decode(input, errors, True)
     17
     18 class IncrementalEncoder(codecs.IncrementalEncoder):

UnicodeEncodeError: 'ascii' codec can't encode character u'\u221e' in position 3: ordinal not in range(128)

But why would you decode a unicode object?

And it should be a no-op – why the exception?

And why ‘ascii’? I specified ‘utf-8’!

It’s there for backward compatibility

What’s happening under the hood

u'to \N{INFINITY} and beyond!'.encode().decode('utf-8')

It encodes with the default encoding (ascii), then decodes

In this case, it barfs on attempting to encode to ‘ascii’

So never call decode on a unicode object!

But what if someone passes one into a function of yours that’s expecting a py2 string?

Type checking and converting – yeach!

Read:

http://axialcorps.com/2014/03/20/unicode-str/

See if you can figure out the decorators:

unicodify.py.

(This is advanced Python JuJu: Aren’t you glad I didn’t ask you to write that yourself?)