pandas looks really interesting. Surprised I've never heard about it. Can it be used as a in-memory relational database?
It's really a replacement for R's data.frame. The emphasis is on convenience for data manipulation tasks on heterogeneous rectangular data. I don't think it's intended to be used as a database even if that's possible.
You wouldn't want to use it as a database if you are looking for hardened ACID properties. But if you are looking to work with relational or labeled data sets and/or read & write them between memory & disk, it is an extremely powerful tool.