Inconsistent handling of NULL values

[derekjn]

Given a stream s1 that contains a NULL value:

CREATE STREAM s1 (x INTEGER) ...
INSERT INTO s1 (x) VALUES (null);

If that NULL expression is used in an arithmetic operation, the result of the operation will be NULL:

SELECT x + 1 FROM s1 EMIT CHANGES;

However, if a NULL literal is used in the same way, an NPE will be thrown:

ksql> SELECT NULL + 1 FROM s1 EMIT CHANGES;
java.lang.NullPointerException

We should make NULL handling consistent regardless of context, as some users will evaluate ksqlDB's NULL handling by testing a simple example using a NULL literal like the one above. Any assumptions they derive from such tests will currently not be valid for actual queries.

[stevenpyzhang]

select ARRAY[value0+5, 1+5], ARRAY[1+5], value0 + 5 from TEST emit changes;
If value0 is null in a record, then the output is null, [6], null. This seems a bit strange since it seems like the first array should be [null, 6]

select ARRAY[ucase(value0), 'test'] from TEST1 emit changes; returns [null, 'test']

It seems like if there's a NPE thrown in the expression when the query is processing, the corresponding select column gets a null value. The NPE thrown before the query is created/executed is because of our type checking of the expression before we create the query which fails.

[cprasad1]

We got rid of the NPE, and we now give a new error message in such cases. See the description of #7554 for more details. I am closing this issue, and I have created this issue #7602 instead since the scope of feature that we want is much larger.