ConfigParser and Unicode

Python has a very convenient module for saving and reading ini-like configuration files called ConfigParser .

When using it, I had a problem related to saving Unicode strings to a file. In some subtle cases (for example, it showed me when the application was running under Windows XP), when reading or writing such parameters, a string conversion error pops up.

I could not find ready-made solutions on the Internet, although there are quite a lot of questions about “how to make it always work” - they usually answer in the spirit of “ask the author of the module to fix it”.

I want to offer my solution for those who use Python 2.X - it is quite simple and helps to solve this problem.



First, you need to inherit the RawConfigParser class by overriding the write () method - namely, replacing all str () calls with unicode () calls :

Copy Source | Copy HTML

1. class UnicodeConfigParser(ConfigParser.RawConfigParser):
2.  
3.     def __init__(self, *args, **kwargs):
4.         ConfigParser.RawConfigParser.__init__(self, *args, **kwargs)
5.  
6.     def write(self, fp):
7.         """Fixed for Unicode output"""
8.         if self._defaults:
9.             fp.write("[%s]\n" % DEFAULTSECT)
10.             for (key, value) in self._defaults.items():
11.                 fp.write("%s = %s\n" % (key, unicode(value).replace('\n', '\n\t')))
12.             fp.write("\n")
13.         for section in self._sections:
14.             fp.write("[%s]\n" % section)
15.             for (key, value) in self._sections[section].items():
16.                 if key != "__name__":
17.                     fp.write("%s = %s\n" %
18.                              (key, unicode(value).replace('\n','\n\t')))
19.             fp.write("\n")
20.  
21.     # This function is needed to override default lower-case conversion
22.     # of the parameter's names. They will be saved 'as is'.
23.     def optionxform(self, strOut):
24.         return strOut
25.  

Secondly, writing and reading the configuration file must be done with a wrapper for open () from the codecs module , which must be specified with utf-8 as the encoding. In the case of loading, this can be done if you use readfp () rather than read () for reading :

Copy Source | Copy HTML

1. import codecs
2.  
3. # Saving
4.  
5. confFile = codecs.open('myConfig.ini', 'w', 'utf-8')
6. config = UnicodeConfigParser()
7. # ...
8. config.write(confFile)
9. confFile.close()
10.  
11. # Loading
12.  
13. config = UnicodeConfigParser()
14. config.readfp(codecs.open('myConfig.ini', "r", "utf-8"))

I hope someone comes in handy. If you have a more beautiful and successful solution, I will be glad to hear it.