What to do with packages that includes a python module?

[synocommunity]

What to do with packages that includes a python module?

[zebulon501]

It seems to me that it is quite easy to embed Python in any spk, simply add a dependency on it, and it will get built. Then add your own additional package. This should work, but is not my preferred solution.

Being able to use the standard python spk while embedding additional module would be far better. That would imply to have access to the (a part of) the build tree of the corresponding python spk to be able to invoke the module's setup.py, and to extract only the additional part. A python shell trampoline might also be needed to add this additional site-package directory to the PYTHON_PATH variable.

PS: your workaround is not that good, as is mk/spksrc.python-module.mk. I see potential troubles there.

[Diaoul]

The problem is that some scripts may use python without the required PYTHONPATH. The idea would be to add a custom lib dir to the python sys.path rather than at execution time in PYTHONPATH.

Don't know if this is possible, AFAIK sitecustomize.py allow some customization but may be hard to manipulate. This would imply regexps to edit it...
Can't find many examples on this subject. http://docs.python.org/library/site.html

[Diaoul]

I went for the easiest solution: embed python in the SPK as a dependency.
If we want to embed say, speechd python module, in the default Python SPK, this will be of no use without the actual speech-dispatcher binary and include speech-dispatcher in the python SPK is not a good idea as it has nothing to do with python (well, except a dedicated module).

However, I go for including all common python modules with C extensions in the python SPK (pyaudio, lxml, and others)

Pure python modules should not even be in cross and should be bundled with the dedicated SPK. pip bundle can do that and I'm working on it so a new SPK should be out soon.

[zebulon501]

I agree with you, keeping more generic python modules in the main python spk, and embedding python for very specific needs. The bad thing is that it will consume more RAM on our NAS, as the two python executable are different files and the code section can't be saved.