I think that asking questions about your task is totally normal. I
believe that discussion serves as a great source of ideas, and I
absolutely dislike working solo. If you want things done well, do it
yourself --- that's not my way. So when I was asked to help
colleagues from the team which now hosts me for a short while during
this recommender system project, I didn't know for sure but hoped
that it could be a good idea. And yes, today I discovered different
types of data that Pandas can work with. I spent my previous years
most using Hadoop as a primary data source, so I didn't really need
to bother with Pandas, and my knowledge of it remained pretty
superficial. So I usually saved to or read from CSV. I tried to use
HDF5 couple of times, but still, there was no particular gain over
Parquet. And now Pandas supports the Apache Arrow based `feather`
format and even reading from SAS files. If you have already
streamlined infrastructure of getting data from Teradata to SAS why
not read then from SAS files after all? So helping others with your
experience sometimes means not finding all the answers there, and
ultimately finding new solutions to your own puzzles. E.g. today I
got a several times speed-up using `feather` for my own dataset after
researching the theme a bit helping a colleague who had troubles
choosing a better format for storing her data.
