Miserlou
Added pandas for python 2.7
Nice.
Before I merge, can I ask - why did you make this? I was under the impression that pandas had a manylinux wheel on PyPI, but it looks like not for this version (and not for Python 3.6).
Actually, now that I look again, version 20.1 has manylinux wheels for 2.7 and 3.6: https://pypi.python.org/pypi/pandas/
v0.20.1 was not released when I built this, but even the old version already had a manylinux wheel... I'm sorry, I am new to python and devopsing in python and was not aware such a system existed. Feel free to close this PR.
Not your fault! This should be documented better.
I'm going to leave this open as I'm working on some lp stuff today anyway, and maybe people would want the old version? Not sure. Have to make sure the manylinxu wheels work as well.
Are you using lp with Zappa?
No, we use pandas to ease some data processing on air quality measurement data. Nothing fancy, but it saves us a couple (hundred) lines of code. And since we run the code on Lambda, this repo came in quite handy (also for numpy and psycopg2, big thanks for those btw! I also really enjoyed your talk at 33c3).
Hi @Tobi042 . I am trying to figure out the best way to build pandas as numpy myself so that it is ready to use by lambda. I would like to use virtualenv and pip installs but that does not seem to build the necessary C-extensions. If I download the source from GitHub and then build them locally myself, the repos seem to get very large (too big for lambda). I am hoping you can clue me in to how to build these libraries but keep the code small (how you did this)? Thank you!!
@adamwlev What I did is build this in an EC2 VM with Amazon Linux AMI using the https://github.com/Miserlou/lambda-packages/blob/master/lambda_packages/cryptography/build.sh script by calling ../cryptography/build.sh pandas 0.19.2 in the lambda_packages/pandas directory on the VM.
But the reason this PR wasn't approved is because there is apparently a manylinux wheel available for pandas. I still have no clue how to use that (since this whole thing was a one-off for me and my build sufficed), but I believe that's what is supposed to be used instead of what I built.
I see - I was struggling when trying to build it from my local machine. Then I just rented an Amazon Linux AMI Ec2, following these instructions: https://stackoverflow.com/questions/34749806/using-moviepy-scipy-and-numpy-in-amazon-lambda and it worked great. I appreciate the response..