undefined | Better HN

0 pointsiskander14y ago0 comments

I'm totally ignorant of mongodb: what does it do for you (in the way of data aggregation) that's not easy in numpy?

0 comments

2 comments · 1 top-level

hogu14y ago· 1 in thread

if your data can fit into arrays, then there's no advantage in terms of the types of aggregations.

however mongo allows you to store complex structures, think nested dictionaries/lists, and query on those nested structures, even allowing you to reach inside of nested structures to do the querying.

I guess you could do nested structured arrays in numpy, I've never done that before.

iskanderOP14y ago

I use h5py's datasets (which are organized hierarchically and stored in compressed chunks) to do basic filtering and then load a fraction of my data into memory as numpy arrays.

j / k navigate · click thread line to collapse