[Python-Dev] Any PEP about 2.6 -> 3000 code transition? (original) (raw)
Stefan Behnel stefan_ml at behnel.de
Sun May 25 09:59:55 CEST 2008
- Previous message: [Python-Dev] Any PEP about 2.6 -> 3000 code transition?
- Next message: [Python-Dev] Stabilizing the C API of 2.6 and 3.0
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Gregory P. Smith wrote:
Of particular interest to bsddb is the pep3118 buffer API. I updated the existing bsddb module found in the py3k branch to use it last fall; you may want to use the changes made to it as an example.
I agree, a single C code base is the only sane way forward for bsddb. The interesting part is that you also want to support older python versions (2.3-2.5?) as well so bsddb.c & bsddb.h become a nice example of how to write code that spans multiple python versions. I expect it is less difficult than it sounds.
I would say so. Most things can be done with preprocessor conditionals and some redefinitions, e.g. "#define PyLong_Type PyInt_Type", etc.
The method "generate_module_preamble()" in Cython's ModuleNode.py source file (from line 360 onwards) has a code block of things we found so far.
http://hg.cython.org/cython-devel/file/tip/Cython/Compiler/ModuleNode.py
I believe the buffer api is supposed to be backported to 2.6? I haven't followed the status of that. If so, bsddb should use it in 2.6 as well.
The API is there (in addition to the old buffer API), but it doesn't seem to get used.
Stefan
- Previous message: [Python-Dev] Any PEP about 2.6 -> 3000 code transition?
- Next message: [Python-Dev] Stabilizing the C API of 2.6 and 3.0
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]