python - SQLAlchemy: how to filter on PgArray column types?

Question

In pure postgres we can write:

SELECT * FROM my_table WHERE 10000 = ANY (array_field);

or

SELECT * FROM my_table WHERE 10000 = ALL (array_field);

How to do the same with the help of sqlalchemy without raw sql?

Answer 1

a = ANY(b_array) is equivalent to aIN(elements_of_b_array)¹.

Therefore you can use the in_() method.

I can't remember ever having used a = ALL(b_array) in all my years with PostgreSQL. Have you?

---

If you are dealing with an array column and want to test whether it contains a given element (or all elements of a given array) in that column, then you can utilize PostgreSQL array operators @> (contains) or more appropriately the inverse sibling <@ (is contained by).

Array operators carry the advantage that they can be supported with a GIN index on the array column (unlike the ANY construct).

Your SQL statement:

SELECT * FROM my_table WHERE 10000 = ANY (array_field);

is (almost)¹ equivalent to

SELECT * FROM my_table WHERE 10000 <@ array_field;

I am no expert with SQLAlchemy, but according to the tutorial in the SQLAlchemy manual, you can use any operator:

If you have come across an operator which really isn’t available, you can always use the op() method; this generates whatever operator you need:

>>> print users.c.name.op('tiddlywinks')('foo') users.name tiddlywinks :name_1

Bold emphasis mine. Your statement could look like this in SQLA:

s = select([my_table], array_field.op('@>')('ARRAY[10000]'))

Or with alternative input syntax for PostgreSQL array values:

s = select([my_table], array_field.op('@>') (cast('{10000}', int[])))

---

¹ There is a subtle difference with NULL handling:

SELECT '{NULL}'::int[] <@ ... -- that's an array with a single NULL element

always yields FALSE.

SELECT NULL IN (...)
SELECT NULL = ANY (...)
SELECT NULL::int[] <@ ...

always yield NULL.

If you are not going to query for NULL values, you can ignore this.