CODEWITHSUNDEEP
pythonunicodefilenames
jack
jack

Reputation: 17881

UnicodeEncodeError on joining file name

It throws out "UnicodeDecodeError: 'ascii' codec can't decode byte 0xc2 in position 2: ordinal not in range(128)" when executing following code:

filename = 'Spywaj.ttf'
print repr(filename)
>> 'Sp\xc2\x88ywaj.ttf'
filepath = os.path.join('/dirname', filename)

But the file is valid and existed on disk. Filename was extracted from "unzip -l" command. How can join filenames like this?

OS and filesystem

Filesystem: ext3    relatime,errors=remount-ro 0       0
Locale: en_US.UTF-8

Alex's suggestion os.path.join works now but I still cannot access the file on disk with the filename it joined.

filename = filename.decode('utf-8')
filepath = os.path.join('/dirname', filename)
print filepath
>> /dirname/u'Sp\xc2\x88ywaj.ttf'
print os.path.isfile(filepath)
>> False

new_filepath = filepath.encode('Latin-1').encode('utf-8')
print new_filepath
>> /dirname/u'Sp\xc2\x88ywaj.ttf'
print type(filepath)
>> <type 'unicode'>
print os.path.isfile(new_filepath)
>> False

valid_filepath = glob.glob('/dirname/*.ttf')[0]
print valid_filepath
>> /dirname/Spywaj.ttf (SO cannot display the chars in filename)
print type(valid_filepath)
>> <type 'str'>
print os.path.isfile(valid_filepath)
>> True

Upvotes: 9

Views: 17594

Answers (4)

Don Grem
Don Grem

Reputation: 1267

I have fixed the UnicodeDecodeError by adding these lines to /etc/apache2/envvars and restarting Apache.

export LANG='en_US.UTF-8'
export LC_ALL='en_US.UTF-8'

as described here: https://docs.djangoproject.com/en/dev/howto/deployment/wsgi/modwsgi/#if-you-get-a-unicodeencodeerror

I have spent some time debugging this.

Upvotes: 7

John Machin
John Machin

Reputation: 82924

Evidence problem 1 ###

It throws out "UnicodeDecodeError: 'ascii' codec can't decode byte 0xc2 in position 2: ordinal not in range(128)" when executing following code:

filename = 'Spywaj.ttf'
print repr(filename)
>> 'Sp\xc2\x88ywaj.ttf'
filepath = os.path.join('/dirname', filename)

I can't see how it is possible to get that exception -- both args of os.path.join are str objects. There is no reason to try converting anything to unicode. Are you sure that the above code is exactly what you ran?

Evidence problem 2

Alex's suggestion os.path.join works now but I still cannot access the file on disk with the filename it joined.

filename = filename.decode('utf-8')
filepath = os.path.join('/dirname', filename)
print filepath
>> /dirname/u'Sp\xc2\x88ywaj.ttf'

Sorry, assuming that filename has not changed from the previous snippet, that's definitely impossible. It looks like the result of os.path.join('/dirname', repr(filename)) ... please ensure that you publish the code that you actually ran, together with actual output (and actual traceback, if any).

Confusion

new_filepath = filepath.encode('Latin-1').encode('utf-8')

Alex meant to try twice, each time with one of those encodings -- not try once with both encodings! As all the characters in filepath were in the ASCII range (see evidence problem 2) the effect was simply filepath.encode('ascii')

Simple solution

You know how to find the name of the file that you are interested in:

valid_filepath = glob.glob('/dirname/*.ttf')[0]

If you must hard-code that name in your script, you can use the repr() function to get the representation that you can type into your script without worrying about utf8, unicode, encode, decode and all that noise:

print repr(valid_filepath)

Let's suppose that it prints '/dirname/Sp\xc2\x88ywaj.ttf' ... then all you need to do is carefully copy that and paste it into your script:

file_path = '/dirname/Sp\xc2\x88ywaj.ttf'

Upvotes: 0

YOU
YOU

Reputation: 123821

filename = filename.decode('utf-8').encode("latin-1")

works for me with the file from Splywaj.zip

>>> os.path.isfile(filename.decode("utf8").encode("latin-1"))
True
>>>

Upvotes: 2

Alex Martelli
Alex Martelli

Reputation: 881575

In both Latin-1 (ISO-8859-1) and Windows-1252, 0xc2 would a capital A with a circumflex accent... doesn't seem to be anywhere in the code you show! Can you please add a

print repr(filename)

before the os.path.join call (and also put the '/dirname' in a variable and print its repr for completeness?). I'm thinking that maybe that stray character is there but you're not seeing it for some reason -- the repr will reveal it.

If you do have a Latin-1 (or Win-1252) non-Ascii character in your filename, you have to use Unicode -- and/or, depending on your OS and filesystem, some specific encoding thereof.

Edit: the OP confirms, thanks to repr, that there are actually two bytes that can't possibly be ASCII -- 0xc2 then 0x88, corresponding to what the OP thinks is one lowercase L. Well, that sequence would be a Unicode uppercase A with caret (codepoint 0x88) in the justly popular UTF-8 encoding - how that could look like a lowercase L to the OP beggars explanation, but I imagine some fonts could be graphically crazy enough to afford such confusion.

So I would first try filename = filename.decode('utf-8') -- that should allow the os.path.join to work. If open then balks at the resulting Unicode string (it might work, depending on the filesystem and OS), next attempt is to try using that Unicode object's .encode('Latin-1') and .encode('utf-8'). If none of the encodings work, information on the OS and filesystem in use, which the OP, I believe, hasn't given yet, becomes crucial.

Upvotes: 8

Related Questions