Re: [IBMiOSS] Pandas on IBM i

then i have compiled from source (not present on http://www.perzl.org/aix/ ):

libffi-3.2.1
cyrus-sasl-2.1.25

( autogen.sh configure make and so on ...)

maybe i can do a tar.gz and put somewhere
for check diff from original (because i don't remember but i make some mods ...)

than i compiled manually the python pack : (python setup.py install for g++ linker problem )
python-sasl-0.2.1
pandas-0.18.1

for numpy and cpython i don't remember problem

"pip install superset" install all other prereq. and run ok .

PS . ALL ON PYTHON 2.7 !!

So now for who what to try tell me where you stop and i can see the difference on my system .


Il 08/03/2017 19:38, John Yeung ha scritto:

On Wed, Mar 8, 2017 at 1:09 PM, Jim Steil <jim@xxxxxxxxxxxxxxx> wrote:

As a follow-on to this discussion...

https://archive.midrange.com/opensource/201703/msg00012.html

Based on what Massimo said in the referenced post, Is there something we
(IBM) can do to get pandas pip-installable?

I hope Massimo comes to this thread.

His apparent success with pandas strongly implies that he was also
able to build NumPy and Cython in PASE, which would be instrumental in
paving the way for other Python packages.

If IBM isn't going to provide those as PTFs (which I still believe
they should, if they're going to continue using the PTF distribution
model at all), then we ought to at least very carefully and precisely
document what was needed to get them to work, and post the
instructions to one or more of the wikis for i floating around. By
precise, careful documentation, I mean idiot-proof, concrete steps
that anyone can follow, even if they don't understand Unix-style build
processes.

John Y.