-
Notifications
You must be signed in to change notification settings - Fork 107
Open
Labels
epicfeatureRelease label indicating a new feature or requestRelease label indicating a new feature or request
Description
As of right now, we have these compute functions defined in vortex-compute, which is how we share logic for computation in the new operator world:
-
arithmetic: Binary Arithmetic -
arrow: convert to Arrow -
cast: Cast to a givenDType -
comparison: Binary comparison -
expand: Expands aBufferto true positions of aMask -
filter: Filter data by a selection -
logical: Boolean and Kleene logic -
take: Take/gather data according to indices
Note that unticked items above have yet to be fully implemented (as in there are known missing implementations on different types for a given compute function).
There are other compute functions we still need to add (names TBD):
-
patches: Apply patches to data (this is the only one we don't already have as a compute function in the old world) -
fill_null: Fill null values with valid data -
invert: We probably don't need this since we havelogical? -
is_constant: Data is constant -
is_sorted: Data is sorted -
like: We probably don't need this since we havecomparison? -
list_contains: We probably don't need this since we havecomparison? -
mask: Replace values with null. This feels very similar tofill_null? -
min_max: The min and max of data -
nan_count: The number of NaN -
sum: The sum of all data -
zip: Element-wise conditional selection between two arrays based on aMask.
I also think it would be nice to have a reduce compute function that would allow us to generically write the compute functions for is_constant, is_sorted, min_max, nan_count, and sum.
Metadata
Metadata
Assignees
Labels
epicfeatureRelease label indicating a new feature or requestRelease label indicating a new feature or request