David Stryker

This test shows that MERGE erroneously allows non-null columns to be set to null: ``` @Test public void testMergeNonNullableColumns() { String targetTable = "merge_non_nullable_target_" + randomTableSuffix(); assertUpdate(createTableForWrites(format("CREATE TABLE %s (nation_name...

I submitted PR #13535 which changes the two implementations of MergeRowChangeProcessor to throw an exception if a null value is assigned to a non-null column as a result of an...

> We should to this. This would allow the optimizer to eliminate these checks when it's able to prove the values being written are non-NULL. Interesting. My guess was the...

> I'm fine doing this on top of https://github.com/trinodb/trino/pull/13535 It makes sense to me to merge #13535 and if we decide the checks should be moved to the query plan,...

> Alternatively we can merge this PR and @homar or someone else will follow up. I thought I'd responded to this but I don't see the response. Yes, since setting...

I really think this PR should be merged, since allowing null to be assigned to a non-null column is a big-time correctness issue. Can someone who agrees and has the...

I've updated the PR as suggested by @electrum and @martint, and all tests pass. This PR should be merged.

> Delete the overridden testDelete and testUpdate methods from BaseIcebergFailureRecoveryTest. We can use the base ones nows. Removed

There still are a bunch of test failures in the delta lake connector, and I think they reflect problems in the engine implementation, revolving around partition columns and values, and...

All the fault-tolerant delete and update tests are failing for the TASK type because a partition-related exception is raised before the expected injected exception. @electrum pinged the fault-tolerant folks for...