homepage

The following is a minimal example which does not work under Python
3.1.1 but functions as expected on Pyhton 2.6 and 3.0.
Python 3.1.1 believes the single UTF-8 glyph is two entirely different
(and illegal) unicode characters:
Traceback (most recent call last):
 File "t.py", line 2, in <module>
 print('𐑛')
UnicodeEncodeError: 'utf-8' codec can't encode character '\ud801' in
position 0: surrogates not allowed
Test system is Ubuntu 9.10-beta 32-bit

While t.py only bugs on 3.1, the following happens with 3.0 as well:
>>> line = '𐑑𐑧𐑕𐑑𐑦𐑙'
>>> first = '𐑑'
>>> first
'𐑑'
>>> line[0]
'\ud801'
>>> line[0] == first
False
And with 2.6:
>>> line = u'𐑑𐑧𐑕𐑑𐑦𐑙'
>>> first = u'𐑑'
>>> first
u'\ud801\udc51'

I can't reproduce that; it prints fine for me.
Notice that it is perfectly fine for Python to represent this as two
code points in UCS-2 mode (so that len(s)==2); this is called UTF-16.

I can't reproduce it either on Ubuntu 9.04 32-bit. I tried both from the
terminal and from the file, using Py3.2a0.
As Martin said, the fact that in narrow builds of Python the codepoints
outside the BMP are represented with two surrogate pairs is a known
"issue". This is how UTF-16 works, even if it has some problematic
side-effects.
In your example 'line[0]' is not equal to 'first' because line[0] is the
codepoint of the first surrogate and 'first' is a scalar value that
represents the SHAVIAN LETTER TOT (U+010451).
Regarding the traceback you pasted in the first post, have you used
print('𐑑') or print(line[0])?
This is what I get using line[0]:
>>> line = '𐑑𐑧𐑕𐑑𐑦𐑙'
>>> first = '𐑑'
>>> print(line[0])
Traceback (most recent call last):
 File "<stdin>", line 1, in <module>
UnicodeEncodeError: 'utf-8' codec can't encode character '\ud801' in
position 0: surrogates not allowed
In this case you are getting an error because lone surrogates are
invalid and they can't be encoded. If you use line[:2] instead it works
because it takes both the surrogates:
>>> print(line[0:2])
𐑑
>>> first == line[0:2]
True
If you really got that error with print('𐑛'), then #3297 could be related.
Can you also try this and see what it prints?
>>> import sys
>>> sys.maxunicode

Python 3.1.1 (r311:74480, Sep 13 2009, 22:19:17)
[GCC 4.4.1] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import sys
>>> sys.maxunicode
1114111
>>> u = '𐑑'
>>> print(u)
Traceback (most recent call last):
 File "<stdin>", line 1, in <module>
UnicodeEncodeError: 'utf-8' codec can't encode character '\ud801' in
position 0: surrogates not allowed

The page:
http://www.fileformat.info/info/unicode/char/d801/index.htm
has a big warning saying that "U+D801 is not a valid unicode character."
The problem is similar to issue6697, and lead to the same question:
should python validate utf-8 input, and refuse invalid unicode characters?

Amaury, you are absolutely correct, \ud801 is not a valid unicode glyph,
however I am not giving Python \ud801, I am giving Python '𐑑' (==
'\U00010451').
I am attaching a different short example that demonstrates that Python
is mishandling UTF-8 on both the interactive terminal and in scripts, u.py
The output should be the same, but on Python 3.1.1 compiled for wide
unicode it reports two different values. As someone on #python-dev
found '𐑑'.encode('utf-16').decode('utf-16') outputs the correct value.

I believe this is a duplicate of issue #3297. When given a high unicode
scalar value directly in the source (rather than in escaped form) python
will split it into surrogates, even on a UTF-32 build where those
surrogates are nonsensical and ill-formed.
Patches for Issue #3672 probably made this more visible.

This behavior is identical whether u.py or u.pyc is run on my systems,
where that previous ticket concerns differing behavior.
Though it is obviously related.

This is a duplicate of #3297, and Adam's patch there fixes it.