[Xapian-tickets] [Xapian] #346: Python 3 support

[Xapian]

#346: Python 3 support
--------------------------------------+-------------------------------------
 Reporter:  olly                      |       Owner:  richard  
     Type:  defect                    |      Status:  assigned 
 Priority:  highest                   |   Milestone:  1.3.2    
Component:  Xapian-bindings (Python)  |     Version:  SVN trunk
 Severity:  normal                    |    Keywords:           
Blockedby:                            |    Platform:  All      
 Blocking:                            |  
--------------------------------------+-------------------------------------

Comment(by barry):

 I've gotten back to this and spent more time on the swig-based bindings.
 It's not going well.  One of the big problems I have is that I'm not sure
 swig's Python 3 mappings for std::string are entirely right, but maybe I
 don't know swig or c++ that well.  An example: iiuc, std::string is just a
 container for bytes not unicodes, but swig wants to convert to and from
 unicodes to std::strings.  E.g. SWIG_AsCharPtrAndSize() only checks its
 first arg for PyUnicode-ness, not PyBytes-ness, but I don't see any reason
 why it shouldn't accept bytes, or maybe *only* bytes.  In Python 3, it
 then tries to decode it as a utf8 string, but there's no reason why it
 must be utf8 encoded data.

 As another example, look at smoketest3.py.  There's code that does the
 equivalent of xapian.Query(xapian.Query.OP_OR, ('foo, 'bar\xa3')).  A
 strict translation of that to Python 3 ought to be
 xapian.Query(xapian.Query.OP_OR, (b'foo', b'bar\xa3')) which incidentally
 will work exactly the same in Python >= 2.6.  With some hacker to accept
 the bytes object, I can make this work, but then swig once again gets in
 the way because it wants to convert the Query's .get_description()
 std::string to a Python 3 unicode using utf-8.  It's the moral equivalent
 of b'bar\xa3'.decode('utf-8') which isn't value utf-8.

 Suggestions welcome.

-- 
Ticket URL: <http://trac.xapian.org/ticket/346#comment:46>
Xapian <http://xapian.org/>
Xapian