Install Numpy/Pandas in VirtualEnv - VS2012

Apr 19, 2013 at 10:13 AM
Edited Apr 19, 2013 at 11:27 AM
Hi
I'm pretty new to all things python and am just trying out PTVS (win7 32 bit). I am trying to install pandas and numpy into a virtual env within PTVS ... and am struggling.

Q1. I've read that there are problems with numpy and a decent open source fortran compiler and hence you need to do an install of the binaries. So I can install numpy via a windows installation and then copy the numpy folder into the site-packages folder onthe virutal env. Is that good enough or could it lead to problems down the line. (I wasn't sure whether as part of the installation there are certain registry settings which get configured and I'd be abusing this)?

Q2. I am trying to install pandas via the install python package in VS2012 but when installing it is obviously looking for gcc. What is the way around this? (I have set the distutils.cfg build setting to mingw and sitnalled mingw), so it doesn't look like this is being taken into account.

EDIT: So I did an easy_install whilst in a command prompt in the virtualenv folder for both numpy and pandas and they seemed to work ok in that, I can see them in the list of dependencies/references in the virtual env and when I run some test programs it seems to work ok. However, I do not have intellisense for these components.

Q3. Was that a reasonable way to install these packages?
Q4. What do I need to do to make intellisense work on them?

Many thx

S
Jun 22 at 12:54 AM
hmm you kinda git it now...
Jul 17 at 8:10 PM
For completeness, here are the step-by-step instructions:
  1. Download the compiled egg package of the library you want to include (numpy, pandas, scipy,...). At the moment of this writing, many are available from here. The package file name will look like this numpy-1.9.0b1-win32-superpack-python3.4.exe. Note the .exe extension.
  2. Start a command prompt. Go to your virtualenv folder, in the Scripts folder.
  3. Type easy_install numpy-1.9.0b1-win32-superpack-python3.4.exe at the prompt.
Works well for me. My web site with on-the-fly differential equation solver using scipy and numpy is faster when served by the free Azure Web Site than when served by my localhost.