unicode.isnumeric() and unicodedata.numeric() fails to
recognize a bunch of numeric unicode characters.

The patch fixes this.

Logged In: YES 
user_id=21627

Which version of the Unicode database is this based on?

Logged In: YES 
user_id=621306

The patch makes it match version 4.1.0. Though it didn't match
version 3.2.0 either.

Logged In: YES 
user_id=38388

Rather than creating a patch for every new version, how
about extracting the relevant data from the Unicode database
using a script and putting that into Tools/unicode/ ?!

Note that the original version was also generated from the
database. Unfortunately, I can't find that script anymore.

One nit with the patch: it should put non-BMP Unicode code
points into #ifdef Py_UNICODE_WIDE ... #endif clauses.

Logged In: YES 
user_id=21627

I agree it should be possible to regenerate that easily (or
perhaps entirely merge it into unicodedata/unicodectype).

andersch, how did you create the patch?

Logged In: YES 
user_id=621306

I got the differenced through comparing with _numeric in
http://codespeak.net/svn/pypy/dist/pypy/module/unicodedata/unicodedb.py
which we have generated with
ssh://codespeak.net/svn/pypy/dist/pypy/module/unicodedata/generate_unicodedb.py

If You are looking into generating this from the Unicode
database You might want to fix _PyUnicode_IsLinebreak too.

Logged In: YES 
user_id=21627

Thanks for the patch, committed as 46432. I conditionalized
the non-BMP characters on Py_UNICODE_WIDE, and updated
PyUnicode_IsNumeric to recognize U+0F33 as a numeric character.

If anybody wants to contribute a generator for these
functions (or perhaps generate a table in the first place),
please go ahead.