PA under (python 3.4 at least) is still using pandas 0.14.1 which is almost two years old. How about upgrading to the latest version? pandas would have to be one of the more frequently used libraries.
Quite by accident I found out that a major new major version was released today: 0.18.0
You can do this for yourself in a Bash shell with:
pip3.4 install --upgrade --user pandas
(Type it exactly as above - don't substitute anything for 'user'!)
I just tried this and it successfully installed pandas 0.18.0 and its prerequisites.
Thanks for the reply. I know how to do it for myself but why dig into my storage quota.. I think it's an important enough library to merit a system-wide upgrade. A two year old version is way too outdated.
Let's see what PA say then. I thought you might know about the command above, but hopefully it'll help someone else!
We're currently constrained by the fact that sytem-wide upgrades are likely to break things for existing users. We're working on some infrastructure changes that will let different users run different system images though...
Harry, does a pip install --user definitely count against a user's disk quota?
If so, couldn't you allow users to share common installs using symlinks or something? just a thought..,.
Jim
It does count, but unfortunately we can't let people share them via symlinks easily without breaking our sandbox mechanism. The right answer is definitely for us to sort out the switchable system images, which is proving tricky (basically, we were planning to use Docker but that appears to have system performance issues, at least in the way we're using it).
I second the vote to upgrade Pandas as soon as possible. The old version is causing me trouble. I will try the pip install trick in the meantime.
Mark
For anyone interested, I just ran the command 'pip3.4 install --upgrade --user pandas' in a Bash shell, and it installed:
harry
|
2710
posts
|
PythonAnywhere staff
|
pandas 0.19.2
numpy 1.10.4
Jim
@markhmoulton -- this is quite an old thread, the "switchable system images" solution that I referred to in my previous post has been live for four months or so now. It looks like you're on the old "classic" image -- the new "dangermouse" image has pandas 0.18.1.
If you want us to switch you over to dangermouse, we can do that easily from our side -- just let us know. Be aware, though, that this will upgrade almost all of the pre-installed Python modules to newer versions, so any code you have that depends on the older versions may break.
You can see a full list of pre-installed modules, and compare the versions installed in different system images, on the batteries included page.
