When I first started, installing packages which required compiling native code on either my work Windows machine and the old Unix servers was not easy.
So I largely stuck to the Python standard library where I could, and most of the operations I had at the time did not require data analysis on a server, that was mostly done in a database. Often the job was validating and transforming the data to then insert it into a database.
As the Python packaging ecosystem matured and I found I could easily use Pandas everywhere it just wasn't my first thing I'd reach to. And occasionally it'd be very helpful to iterate through them with the csv module, only taking a few MBs of memory, vs. loading the entire dataset into memory with Pandas.
