Add support for NullsDistinct() / NullsNotDistinct() for Postgre uinque indexes

We have a migration for Postgres that creates an index over two columns, one of them is non-nullable. We want to mimic MsSql behavior (nulls are treated as the same value) and to forbid duplicates like this:

 Name     OwnerId
------- ----------
 User1       NULL
 User1       NULL

There’re NullsDistinct() / NullsNotDistinct() extension methods for MsSql server (actually, it does in reverse, allowing nulls to be treated as a distinct values), but there are no such helpers for Postgres.

There are two well-known approaches In PgSql

1. Create a secondary index with WHERE OwnerId IS NULL filter. Generally it is not a good solution as its nontrivial to apply if you have more than one nullable column in unique index.
2. Wrap column with Coalesce(OwnerId, '00000000-0000-0000-0000-000000000000') expression.

I’d prefer to have a helper for the second approach, or, at least, a way to set a custom expression for .OnColumn() calls.

Is this a feature you’d like to submit a PR for? No 😦

[jzabroski]

@lillo42 See previous reply. I assigned it to you. Feel free to un-assign yourself if you’re not that interested.

[jzabroski]

@fairking , the syntax is:

Create.Index
                .OnTable(GeneratorTestHelper.TestTableName1)
                .OnColumn(GeneratorTestHelper.TestColumnName1).Ascending().NullsDistinct()
                .OnColumn(GeneratorTestHelper.TestColumnName2).Descending()
                .WithOptions().UniqueNullsNotDistinct();