Robert (Bobby) Evans

To be clear I don't think that the issue has anything to do with caching. https://github.com/NVIDIA/spark-rapids/blob/7e4324c6f117e9555a8bb427205562f296591003/sql-plugin/src/main/scala/com/nvidia/spark/rapids/aggregate.scala#L517-L522 is the area of the code that is failing. This code assumes that calling...

@lagarantie I have put up a [patch](https://github.com/NVIDIA/spark-rapids/pull/6767) and merged it into the 22.10 release that should fix the issue you are seeing. It fixes all of the places where we...

@lagarantie have you had a chance to test it? Can I close this issue?

So do you want us to document this somewhere? I am just trying to understand what the result of this is. If it is more awareness, then okay we can...

> Not familiar with JNI, but I wonder could we move the implementation to https://github.com/NVIDIA/spark-rapids-jni ? Min/Max for a float/double reduction is something we could look into, but a group...

@res-life Yes this is a known issue with JSON parsing. Spark distinguishes between quoted fields and non-quoted fields. For now we are just documenting this https://github.com/NVIDIA/spark-rapids/blob/branch-22.06/docs/compatibility.md#json-floating-point > Another limitation of...

I really dislike regular expression use in casts, but it is a good first step. It would be nice to file a follow on issue to write a custom kernel...

Just like the test prints out. ``` WARNING : Unsupported version of Java (17). You may encounted unexpected test failures ``` Although we should fix the spelling of "encounted" in...

From looking at the patch and the code for GPuCratedNamedStruct I don't think we will be impacted. The name of each field is passed into the constructor. But this is...

We also need to add in more tests to cover the cases that https://github.com/rapidsai/cudf/pull/11792 now supports and to cover aggregations that were missed but require sort based operations.