You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository has been archived by the owner on Feb 18, 2024. It is now read-only.
Step2: use the indices to copy sorted values into new arrays via take
While the method works quite well, it sometimes causes unnecessary memory allocation in step 2. For instance, when references to the columns are unique (not shared), we can do in-place sorting without allocation.
To avoid allocation, maybe we can have a reorder function (within a feature like compute-reorder) to do it.
The idea has been implemented for a single slice here (source:
If I understand it correctly, multiple-column sorting in arrow2 is done in two steps:
While the method works quite well, it sometimes causes unnecessary memory allocation in step 2. For instance, when references to the columns are unique (not shared), we can do in-place sorting without allocation.
To avoid allocation, maybe we can have a
reorder
function (within a feature likecompute-reorder
) to do it.The idea has been implemented for a single slice here (source:
This idea can be easily expanded to multiple columns or chunks.
Hope this makes sense. If not please comment on how in-place sorting can be done with the current version of
arrow2
. Thanks!!The text was updated successfully, but these errors were encountered: