Closed Illviljan closed 7 months ago
I would like to see the methods get_channel_dtype and get_channel_shape added to the api to avoid loading all the data to memory.
get_channel_dtype
get_channel_shape
To be clear, they would correspond to:
import numpy as np arr = np.array([3,6,8], dtype=float) arr.shape Out[33]: (3,) arr.dtype Out[34]: dtype('float64')
This allows you to reason about the arrays without loading the data with for example dask.
You can have a look at arrow-rs-instead-of-arrow2 branch, it contains a new method .get_channel_dtype(), returning a struct with fields kind and shape
latest release should fix this issue
Cool, thanks!
I would like to see the methods
get_channel_dtype
andget_channel_shape
added to the api to avoid loading all the data to memory.To be clear, they would correspond to:
This allows you to reason about the arrays without loading the data with for example dask.