edponce commented on code in PR #12460:
URL: https://github.com/apache/arrow/pull/12460#discussion_r846175739
##########
docs/source/python/api/compute.rst:
##########
@@ -45,6 +45,21 @@ Aggregations
tdigest
variance
+Cumulative Functions
+--------------------
+
+Cumulative functions are vector functions that perform a running total on its
+input and outputs an array containing the corresponding intermediate running
values.
Review Comment:
We captured the behavior of cumsum for Pandas, R, numpy in
[JIRA](https://issues.apache.org/jira/browse/ARROW-13530) and there are
differences among them. Current implementation default behavior is consistent
with Python and R.
Also, the only inputs supported in this PR are `Array` and `ChunkedArray`.
Support for `RecordBatch` and `Table` inputs can be added in this PR or as a
follow-up one. I agree that if inputs are limited, it should be stated explicit
in the wordings.
@ianmcook You are correct, cumulative operations require window functions
for them to be usable in the query engine. This PR provides a kernel that can
be called directly to perform cumulative operations on arrays. Also, it
sketches a solution for cumulative operations and shows a kernel internally
using another kernel.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
