Trouble importing / using a custom text codec

[Jurie Horneman]

I'm having an irritating problem with a text codec I've written.

- The codec is in site-packages/MBCS_codecs/cp932.py.

- site-packages/MBCS_codecs contains an empty __init__.py.

- from MBCS_codecs import cp932 works fine.

- If I put cp932.py in the same place as a script that decodes a text
using encoding "cp932", there is no problem.

However, I want to have all my encodings somewhere else...

- With cp932.py in site-packages as described above, if I decode a
text using encoding "MBCS_codecs.cp932", I get the error message
"LookupError: unknown encoding". This happens even if I import cp932
by hand before decoding the text.

- The code that fails is in encodings/__init__.py :

    modname = encoding.replace('-', '_')
    modname = aliases.aliases.get(modname,modname)
    try:
        mod = __import__(modname,globals(),locals(),'*')
    except ImportError,why:
        # cache misses
        _cache[encoding] = None
        return None

- Executing the following code in one of my own scripts works:

    modname = "MBCS_codecs.cp932"
    modname = modname.replace('-', '_')
    try:
        mod = __import__(modname,globals(),locals(),'*')
    except ImportError,why:
        print "FAILED"

i.e. the module is imported.

- The aliases in encoding/__init__.py do not modify the module name.
In other words, it appears that the same code works or does not work
depending on where it is executed.

I cannot explain this behavior. Any help is appreciated.

Jurie Horneman