Forums

My program is complaining about PySide. So I did: pip install PySide

But is still running after many minutes.... and it is eating my cpu time

23% used (706.08s of your 3,000 second CPU allowance) Allowance resets in 23 hours, 47 minutes

is that normal? I had 0% of CPU just before this command.

At the end, I have used 53% used (1612.37s of your 3,000 second CPU allowance). But running locally (at my own machine) I realized that it takes a long time.

So I guess this is a normal behaviour.

please, consider this questions solved.

I've got to confess that I broke this record : 97% used (2935.04s of your 3,000 second CPU allowance) Allowance resets in 23 hours, 3 minutes

Ouch! Did you just get that by installing PySide? Or something else?

Indeed, this was for the installation of PySide only.

Like I said in a different post (about flask.ext.wtf), I had to use a virtualenv with flask.ext.wtf in it (to create an interactive web app). But because this virtualenv starts from scratch, it was mandatory to reinstall all the other packages necessary for running the initial web app I wanted to put on here (eg: numpy, pandas, matplotlip, and PySide as well). In comparison, to get flask.ext.wtf installed with python 3.4 (the default python version for running and reloading a web app on PythonAnywhere) would propbably take 10 or 15 second CPU allowance at most, I guess.. :-)

Ah, I see. numpy and pandas are pretty huge, so maybe they took up a lot of that. We're looking into ways of speeding things up by using "wheels" -- basically pre-prepared cached binary installations of modules, which can be installed more quickly and without sucking up quite so much CPU. There is a wheel cache available in your sandbox right now at /usr/share/pip-wheels/, but it's not enabled by default (not fully tested yet). In our next release, I think it will be the default, so hopefully things will be faster.

Actually, I had the numpy and pandas libraries installed separetely from PySide (they've been installed one day before, and surprisingly it was pretty quick). So I don't understand what has sucked up quite so much CPU (compared with the 1612s or so for the same kind of installation completed by somebody else).

How to use the pip-wheel to install new libs, should it arise? Is it just "pip-wheels install PySide" for instance?

It's an extra couple of command-line flag for pip:

pip install --no-index --find-links= /usr/share/pip-wheels/ SomePackage

...plus the --user flag if you're not using a virtualenv.