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
Describe the bug
In cuDF, explode_outer_position will mark the position values of empty rows with 0. Meanwhile, the position values of empty rows are marked as null in Spark.
Steps/Code to reproduce bug
For input data like:
[[5,null,15], 100]
[null, 200]
[[], 300]
cuDF returns
[0, 5, 100]
[1, null, 100]
[2, 15, 100]
[0, null, 200]
[0, null, 300]
But Spark returns
[0, 5, 100]
[1, null, 100]
[2, 15, 100]
[null, null, 200]
[null, null, 300]
The text was updated successfully, but these errors were encountered:
… of null rows (#7754)
`explode_outer` supports writing a position column, but if the row was null it would incorrectly set the position to 0 and the row valid. Instead, it should null that position row as well. Luckily the null column matches 100% with the null column of the exploded column, so we can just copy it after it is created.
Fixes#7721
Authors:
- Mike Wilson (@hyperbolic2346)
Approvers:
- Conor Hoekstra (@codereport)
- Jake Hemstad (@jrhemstad)
URL: #7754
Describe the bug
In cuDF,
explode_outer_position
will mark the position values of empty rows with0
. Meanwhile, the position values of empty rows are marked asnull
in Spark.Steps/Code to reproduce bug
For input data like:
cuDF returns
But Spark returns
The text was updated successfully, but these errors were encountered: