Open brynpickering opened 2 years ago
@timtroendle: would you recommend that at least the intermediate .csv
workflow files should now be .feather
files?
I would wait a little to gather more experience, but in general: yes, I think we should do that.
What can be improved?
Timeseries output can be very large, especially with the addition of multiple years of data (#182). The easiest way to reduce the footprint of the model on disk is to compress the CSVs by default. This can reduce CSV size by 80-90% with minimal additional read/write time.
Pandas can infer this on both read and write. I.e., if a file to be saved ends in
.zip
, then pandas will know to compress it. Likewise if it reads a file that ends in.zip
.Version
1.2.0.dev