I’m doing a little series on some of the nice features/capabilities in Snowflake (the cloud data warehouse). In each part, I’ll highlight something that I think it’s interesting enough to share. It might be some SQL function that I’d really like to be in SQL Server, it might be something else.
Often you want to check the values of a column for NULL values. Because NULL is evil, that’s way. Or you want to do a join and because NULL = NULL does not make sense, you want to make sure those rows can match as well if you have a nullable column in your key. In SQL Server, this will leave you open to the villainy of NULL if the key2 column is nullable:
FROM tableA A JOIN tableB B ON A.key1 = B.key1 A.key2 = B.key2
Typically you resolve this with ISNULL:
FROM tableA A JOIN tableB B ON A.key1 = B.key1 ISNULL(A.key2,'') = ISNULL(B.key2,'')
If the values are NULL, they will be replaced with an empty string and you can actually compare those with each other. But what about integers? You can use -1, but what if this value is actually present in the data? And what about dates? Furthermore, this solution might not use existing indexes because there’s now a function around key2 which prevents indexes from being used. Check out Sargable predicates. In a WHERE clause, you can fix it like this:
WHERE A.key1 = B.key1 AND ( (A.key2 IS NULL AND B.key2 IS NULL) OR (A.key2 = B.key2) )
Which of course leads to a more lengthy expression. In Snowflake, they have this interesting conditional expression: IS [NOT] DISTINCT FROM. Our join becomes:
FROM tableA A JOIN tableB B ON A.key1 IS NOT DISTINCT FROM B.key1 AND A.key2 IS NOT DISTINCT FROM B.key2
This single expression both checks for the equality of its members, but also checks the nullability of both columns. Awesome. A good habit would be to use IS [NOT] DISTINCT FROM instead of every = or <> in every expression (join clauses, WHERE clauses etc.) and you’ll never get burned by those pesky NULLs again!
p.s.: want this in SQL Server? Vote here!