modifying a codec

[Tim Arnold]

Hi, I'm using the codecs module to read in utf8 and write out cp1252
encodings. For some characters I'd like to override the default behavior.
For example, the mdash character comes out as the code point \227 and I'd
like to translate it as — instead.
Example: the file myutf8.txt contains this string:
'factor one - initially'
====================
import codecs

fd0 = codecs.open('myutf8.txt', 'rb', encoding='utf8')
line = fd0.read()
fd0.close()

fd1 = codecs.open('my1252.txt', 'wb', encoding='cp1252')
fd1.write(line)
fd1.close()
====================

The codec is doing its job, but I want to override the codepoint for this
character (plus others) to use the html entity instead (from \227 to
— in this case).

I see hints writing your own codec and updating the decoding_map, but I
could use some more detail.

Is that the best way to solve the problem?

thanks,
--Tim Arnold

 

[Martin v. Löwis]

The codec is doing its job, but I want to override the codepoint for this

character (plus others) to use the html entity instead (from \227 to
— in this case).

I see hints writing your own codec and updating the decoding_map, but I
could use some more detail.

Is that the best way to solve the problem?

I would say so, yes. Look at the source code of cp1252, and it should be
fairly obvious how a charmap codec works. Make a copy of it, and remove
the EM DASH line. This will give you a codec that just won't encode the
character at all anymore.

Then write an error handler that returns u"—" for \227, but
otherwise continues to raise errors. See PEP 293 for code examples
of error handlers.

Notice that this approach only works for encoding; for decoding, your
scheme can't work, because you would need to specify how —
occurring in the input should get decoded -
as u"—" or as u"\u2014"? Most likely, decoding that output
is of no concern to you, in which case the approach with the error
handler is the best way (IMO).

Regards,
Martin