Does it work with SQL Alchemy?

[NathanAP]

Hi!
Looking through packages we found yours and we found it very interesting for what we are building right now! Our only concern is if it can also replace datetime when using it with SQL Alchemy (I did not found anything related in docs), so here is a small example:

class MyTable

  my_date: Mapped[date] = mapped_column(nullable=False)
  my_time: Mapped[time] = mapped_column(nullable=False)
  my_datetime: Mapped[datetime] = mapped_column(nullable=False)

With this small code we can use alembic to create migrations and manage our database freely and easily. Also, when we recover it from the database, we can freely do:

my_table_instance = MyTableSchema.model_validate(my_select_result)
my_table_instance.my_date # -> 2025-01-01
my_table_instance.my_time # -> 12:00:00
my_table_instance.my_datetime # -> 2025-01-01T12:00:00Z
...

Does it work the same way with whenever or I would need to convert from datetime to whenever after the recovery?

edit: fixed my second example return to ISO 8601 just to be clearer.

[ariebovenberg]

I haven't thought of this yet, but it makes sense to implement 👍 . I've found some docs on implementing custom types in sqlalchemy.

Similar to pydantic support (#175), I expect I'll need a few preview releases to iron out the kinks. Until I get to this though, you'll have to use (from_)py_datetime() to go through the standard library unfortunately.

Another note: your second code snippet shows Pydantic, I assume. Are you using these formats as well, or are they just illustrative? This because whenever's pydantic pupport is currently limited to the ISO-like format.

[NathanAP]

Thats great! I believe that a "translation" to Alchemy (and allow Alembic to create automatic migrations) would be an awesome feature for this package because we would only depend on it!

As for you question: it was my bad, its just an illustrative result that I made quickly, I'll fix it😅

[aquarhead]

You can add a TypeDecorator as documented here https://docs.sqlalchemy.org/en/20/core/custom_types.html

class SATypeWheneverDateDelta(TypeDecorator):
    impl = Unicode

    cache_ok = True

    def process_bind_param(self, value: DateDelta | None, dialect: Dialect) -> str | None:
        if value is None:
            return None
        return value.format_common_iso()

    def process_result_value(self, value: str | None, dialect: Dialect) -> DateDelta | None:
        if value is None:
            return None
        return DateDelta.parse_common_iso(value)

Then if you have a base class you can make it automatic with:

class BaseDB(MappedAsDataclass, DeclarativeBase):
    # ref: https://docs.sqlalchemy.org/en/20/orm/declarative_tables.html#customizing-the-type-map
    type_annotation_map = {
        DateDelta: SATypeWheneverDateDelta,
    }

Which allows a simpler mapped_column:

class Thing(BaseDB, kw_only=True):
    ...

    a: Mapped[DateDelta | None] = mapped_column(default=None)

[ariebovenberg]

Thanks @aquarhead, this certainly looks doable.

It'll take a while before I get to this though, as I'm currently focussing on getting the main API ready for a 1.0 release. THe code above shows there's at least a fairly straightforward workaround