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 }}
Repository navigation
Follow-on work for median aggregate function #3040
Is your feature request related to a problem or challenge? Please describe what you are trying to do.
Follow-on issue to address feedback in #3009
In particular:
Run some performance test and see if it would be better to use builders in the aggregate state rather than arrays. Should we just use ScalarValue::List instead?
Should we have a configurable limit on the number of values that can be collected to prevent OOM?
Can we leverage the concat and take kernels?
Should we add an average kernel to arrow-rs?
Can we reduce memory overhead with some form of compression (such as run-length encoding or maintaining a map of unique values with counts)
Another optimization is that you do not need to to sort all values, only the value at the middle index needs to be at the right place. Vec::select_nth_unstable does exactly that in O(n) using a quickselect algorithm.
Activity
jhorstmann commentedon Aug 9, 2022
Another optimization is that you do not need to to sort all values, only the value at the middle index needs to be at the right place. Vec::select_nth_unstable does exactly that in O(n) using a quickselect algorithm.
Dandandan commentedon Aug 23, 2023
#7376 solves most of these