MithunR

Sorry for the delay. It took a while to get my head around it. Thank you, this is an impressive speedup.

Attached [herewith](https://github.com/NVIDIA/spark-rapids/files/13444584/decimals_avg.parquet.zip) is a zipped Parquet file with 102 rows in a single `Decimal(8,3)` column. Taking the window functions out of the equation, one sees that running `AVG()` produces slightly...

I have filed https://github.com/rapidsai/cudf/issues/14507 to track the CUDF side of this. I was able to repro this on CUDF by writing the input as `DECIMAL(12,7)` to Parquet, and then running...

A couple of other findings. I tried querying `SUM`, `COUNT`, `AVG`, etc. as follows: ```sql select sum(c), count(c), sum(c)/count(c), avg(c), cast(avg(c) as DECIMAL(12,8)) , cast(sum(c)/count(c) as decimal(12,7)) from foobar ```...

There were some red herrings in investigating this bug. First off, I have closed the CUDF bug (rapidsai/cudf#14507) I raised for this. CUDF is not at fault; it consistently truncates...

> know what the original input long was before the divide happened and what the double was that we are dividing? I am assuming that it was `(352761953125/ 10 ^...

I can confirm here that [`GpuCast::castFloatsToDecimal()`](https://github.com/NVIDIA/spark-rapids/blob/fcad2279c31aa1d53e0c132cf90432169e212446/sql-plugin/src/main/scala/com/nvidia/spark/rapids/GpuCast.scala#L1620) seems to be the one producing the differing output: ```scala // Approach to minimize difference between CPUCast and GPUCast: // step 1. cast input...

I've relinquished ownership on this bug. I'm not actively working on this one.

All these tests pass with ANSI mode disabled, except for: 1. get_json_test.py::test_get_json_object_quoted_question 2. orc_write_test.py::test_orc_do_not_lowercase_columns 3. logic_test.py::test_logical_with_side_effect (Seems off in the formatting.)

> For example @mythrocks mentioned the need for offset windows. @mythrocks can you please elaborate? An offset window is one where the two ends of the window do not straddle...