Incorrect behavior of select().eq("column", None)

[st-pasha]

Bug report

Describe the bug

If a table contains a nullable column, then trying to use that column in a filter via .eq(column, None) produces rows where the column has text "None" (instead of matching NULL values); and using .is_(column, None) produces an error "failed to parse filter (is.None)".

To Reproduce

Create a simple table "tmp" with a nullable column "text" of type text:

create table public.tmp (
    id bigint generated by default as identity,
    text text null,
    constraint tmp_pkey primary key (id)
  ) tablespace pg_default;

Insert records:

insert into tmp(text) values("None");
insert into tmp(text) values(NULL);

Query the table via the python client:

import supabase
client = supabase.create_client(...)

response = client.table("tmp").select("*").eq("text", None)
print(response.data)

Expected behavior

Actual behavior: Row 1 is returned (where the text is "None"),
Expected behavior: Row 2 is returned (where the text is NULL).

System information

supabase.__version__: 2.4.1

Additional context

Can be fixed by redefining the function eq() in BaseFilterRequestBuilder as follows:

    def eq(self, column: str, value: Any) -> Self:
        if value is None:
            value = "NULL"
            op = Filters.IS
        else:
            op = Filters.EQ
        return self.filter(column, op, value)

[sapphire008]

According to the Python doc, this is how null checking is achieved: https://supabase.com/docs/reference/python/is

data, count = supabase.table('countries')
  .select('*')
  .is_('name', 'null')
  .execute()

[silentworks]

As @sapphire008 has stated you should use the is_ method with null. I have opened a PR to add support for Python's None type with the is_ method. #446