I am using materializations and my materialization table has nullable
columns. Calcite correctly determines that the underlying query can
never produce null for these columns, but I didn't encode that into the
table definitino. I was just wondering why the casts are generated in
the SQL that is sent to the owner of the materialization table.
Thanks for the explanation, I suppose the cast causes proper exceptions
to be thrown when null is encountered instead of simply a NPE? Or is the
cast in such a case really unnecessary?
Mit freundlichen Grüßen,
------------------------------------------------------------------------
*Christian Beikov*
Am 23.08.2017 um 20:01 schrieb Julian Hyde:
I presume you’re talking about RexNodes. It makes a lot of things easier if the
argument to function call are EXACTLY the type required by that function. The
system inserts implicit casts where necessary.
If you’re generating Java code (in Enumerable convention) there is a lot of
difference between INTEGER NOT NULL and INTEGER. The former can be represented
by int, the latter only by java.lang.Integer.
Julian
On Aug 23, 2017, at 2:08 AM, Christian Beikov <[email protected]>
wrote:
Hello,
I came to notice that the use of a nullable column in a context where a
non-nullable is expected, causes a cast that is essentially useless. Am I
missing something and is there reason to why that is done?
--
Mit freundlichen Grüßen,
------------------------------------------------------------------------
*Christian Beikov*
