Write to UTF-8 file in Python

Question

I'm really confused with the codecs.open function. When I do:

file = codecs.open("temp", "w", "utf-8")
file.write(codecs.BOM_UTF8)
file.close()

It gives me the error

UnicodeDecodeError: 'ascii' codec can't decode byte 0xef in position 0: ordinal not in range(128)

If I do:

file = open("temp", "w")
file.write(codecs.BOM_UTF8)
file.close()

It works fine.

Question is why does the first method fail? And how do I insert the bom?

If the second method is the correct way of doing it, what the point of using codecs.open(filename, "w", "utf-8")?

Solution 1

I believe the problem is that codecs.BOM_UTF8 is a byte string, not a Unicode string. I suspect the file handler is trying to guess what you really mean based on "I'm meant to be writing Unicode as UTF-8-encoded text, but you've given me a byte string!"

Try writing the Unicode string for the byte order mark (i.e. Unicode U+FEFF) directly, so that the file just encodes that as UTF-8:

import codecs

file = codecs.open("lol", "w", "utf-8")
file.write(u'\ufeff')
file.close()

(That seems to give the right answer - a file with bytes EF BB BF.)

EDIT: S. Lott's suggestion of using "utf-8-sig" as the encoding is a better one than explicitly writing the BOM yourself, but I'll leave this answer here as it explains what was going wrong before.

Solution 2

Read the following: http://docs.python.org/library/codecs.html#module-encodings.utf_8_sig

Do this

with codecs.open("test_output", "w", "utf-8-sig") as temp:
    temp.write("hi mom\n")
    temp.write(u"This has ♭")

The resulting file is UTF-8 with the expected BOM.

Solution 3

It is very simple just use this. Not any library needed.

with open('text.txt', 'w', encoding='utf-8') as f:
    f.write(text)

Solution 4

@S-Lott gives the right procedure, but expanding on the Unicode issues, the Python interpreter can provide more insights.

Jon Skeet is right (unusual) about the codecs module - it contains byte strings:

>>> import codecs
>>> codecs.BOM
'\xff\xfe'
>>> codecs.BOM_UTF8
'\xef\xbb\xbf'
>>> 

Picking another nit, the BOM has a standard Unicode name, and it can be entered as:

>>> bom= u"\N{ZERO WIDTH NO-BREAK SPACE}"
>>> bom
u'\ufeff'

It is also accessible via unicodedata:

>>> import unicodedata
>>> unicodedata.lookup('ZERO WIDTH NO-BREAK SPACE')
u'\ufeff'
>>> 

Solution 5

I use the file *nix command to convert a unknown charset file in a utf-8 file

# -*- encoding: utf-8 -*-

# converting a unknown formatting file in utf-8

import codecs
import commands

file_location = "jumper.sub"
file_encoding = commands.getoutput('file -b --mime-encoding %s' % file_location)

file_stream = codecs.open(file_location, 'r', file_encoding)
file_output = codecs.open(file_location+"b", 'w', 'utf-8')

for l in file_stream:
    file_output.write(l)

file_stream.close()
file_output.close()

View more solutions

Comments

[tchrist]

Don’t use a BOM in UTF-8. Please.

[Salman von Abbas]

@tchrist Huh? Why not?

[Alois Mahdal]

@SalmanPK BOM is not needed in UTF-8 and only adds complexity (e.g. you can't just concatenate BOM'd files and result with valid text). See this Q&A; don't miss the big comment under Q

[Apache]

Warning: open and open is not the same. If you do "from codecs import open", it will NOT be the same as you would simply type "open".

[Mohamad Fakih]

Thanks. That worked (Windows 7 x64, Python 2.7.5 x64). This solution works well when you open the file in mode "a" (append).

[beta-closed]

you can also use codecs.open('test.txt', 'w', 'utf-8-sig') instead

[show0k]

Use # coding: utf8 instead of # -*- coding: utf-8 -*-which is far easier to remember.

[Dustin Andrews]

This didn't work for me, Python 3 on Windows. I had to do this instead with open(file_name, 'wb') as bomfile: bomfile.write(codecs.BOM_UTF8) then re-open the file for append.

[Mugen]

I'm getting "TypeError: an integer is required (got type str)". I don't understand what we're doing here. Can someone please help? I need to append a string (paragraph) to a text file. Do I need to convert that into an integer first before writing?

[Jon Skeet]

@Mugen: The exact code I've written works fine as far as I can see. I suggest you ask a new question showing exactly what code you've got, and where the error occurs.

[northben]

@Mugen you need to call codecs.open instead of just open

[user2905353]

Maybe add temp.close() ?

[matheburg]

@user2905353: not required; this is handled by context management of open.

[paradox]

I am really interested in seing something like that working on windows