Use non-nullable columns when WHERE IS NOT NULL is used.

[tom5760]

Hello!

I have a query like this playground link:

-- Example queries for sqlc
CREATE TABLE authors (
  id   BIGSERIAL PRIMARY KEY,
  name text      NOT NULL,
  bio  text
);

-- name: GetAuthor :one
SELECT name, bio FROM authors
WHERE bio IS NOT NULL;

This currently generates the following return type for the GetAuthor method:

type GetAuthorRow struct {
	Name string
	Bio  sql.NullString
}

Since I'm filtering for bio IS NOT NULL, is there any way to get sqlc to make Bio just be a string?

[blacob]

There may be other ways, but perhaps an override would work: https://docs.sqlc.dev/en/stable/reference/config.html?highlight=override#overrides

Adding the following to your sqlc.yaml might work, though you might have to play around with it a little to get what you want.

overrides:
- column: "authors.bio"
  go_type: "string"
  nullable: true

[seanlaff]

I'm running into a use case for this as well, I had a parameter sqlc.arg('foo') that Im using in two places in my WHERE statement. In one table the column is and in the other table it's non-nullable. Sqlc throws a

error generating code: named param foo has incompatible types: int32, sql.NullInt32

To allow compilation, I have to make them both nullable through use of sqlc.narg(), which is regrettable since I don't want to support queries where this param is null.

It seems like maybe we need three functions

sqlc.arg() <- inferred type
sqlc.narg() <- force nullable
sqlc.nnarg() <- force non-nullable

[seanlaff]

Actually on second thought, sqlc.arg() and sqlc.narg() should be enough. In scenarios where sqlc.arg() is being used in two places (int32, and NullInt32), sqlc.arg() should just consider the arg as non-nullable, since the intersection of int32 and NullInt32 is int32.