This could initially start with not assuming that the panda's in memory buffer type is numpy (but also allowing arrow), and then converting it to numpy then continuing as expected.
Eventually it could move to utilizing the DataFrame api.
Then lastly it could perhaps default using polars instead of pandas as its in memory representation? (haha i know this might be wishful)
This could initially start with not assuming that the panda's in memory buffer type is numpy (but also allowing arrow), and then converting it to numpy then continuing as expected.
Eventually it could move to utilizing the DataFrame api.
Then lastly it could perhaps default using polars instead of pandas as its in memory representation? (haha i know this might be wishful)