Re:Pandas v2.0 Released
by alfalfasprout on Monday April 03, 2023 @02:27PM
(#35428937)
A lot of libraries depend on numpy directly. Unless polars is a drop in replacement (which I could be wrong but it doesn't seem like it is) then ultimately there's no avoiding pandas in many cases.
[ Parent
| on HN ]
You can just call to_numpy() which is zero copy.
Re:Pandas v2.0 Released
by infinite8s on Friday April 07, 2023 @10:47AM
(#35482099)
Numpy is mostly just a thin metadata wrapper around memory buffers. If pola.rs stores data in that format it should be able to give you zero-copy numpy objects pointing to the underlying pola.rs data.