In preparation for some posts on analytics and visualization, I was inspired by this video of Wes McKinney introducing a PyCon audience to Pandas. I needed a low-friction tool to interactively work with tabular data and for now Pandas looks like the right choice.
I also wanted to take this opportunity to learn how to use PyPI to install packages instead of relying on binary installers. Unfortunately getting this done end-to-end isn't entirely clear so I've put up my steps below.
State of My System Before I started
- Operating System: Windows 7 SP1
- Python Components: Python 2.7.3 and PyWin32–217.
Step 1 Install SetupTools
Go here: http://pypi.python.org/pypi/setuptools and find the binary installer for Python 2.7
I downloaded this version: setuptools-0.6c11.win32-py2.7.exe
Yes, this is a binary installer. But once we have setuptools everything else is easy and will not require an installer.
Look in the C:\Python27\Scripts folder. You'll see that it contains an EXE called easy_install.
Step 2 Install NumPy
Start a cmd shell are this simple command
C:\>Python27\Scripts\easy_install.exe numpy
This command will start and eventually after spitting out a lot of text numpy will be installed.
Step 3 Install dateutil
This package is actually called “python-dateutil” so we need to use that name instead with easy_install
C:\>Python27\Scripts\easy_install.exe python-dateutil
Step 4 Install Pandas
You can see where this is going …
C:\>Python27\Scripts\easy_install.exe pandas
Step 5: Verify Pandas is working
We are done with the installation, let's make sure it actually works by running a quick script.
import numpy
import pandasdf = pandas.DataFrame( { "A": [11, 12, 13], "B": [34,78,109] } )
df.sum()
Parting Thoughts
- easy_install versus pip: It's generally recommended that one use pip instead of easy_install. I chose not to in this example because I was having problems with pip.
- What next: Sometime soon I'll go through some simple pandas scenarios, focusing getting good analytics out of a bug tracking database.