Hello Mike, thank you very much for the in-depth reply and providing a
solution :)
There is no specific reason that I posted in google groups, I did not visit
the support page, this was just something I decided after thinking where to
post first between googlegroups and stackoverflow, I will use github
discussions when I open a new discussion in the future!
Your workaround will help me continue so many many thanks. I thought I
tried all variations, but I can only concede after seeing this work that I
did not attempt adding the selectinload to the root options().
It might warrant a new thread, but can you also tell me if there is a way
to control the options() that sqlalchemy uses for the recursion of
next_step ?
Ie. I included this bit in the original post:
```
.options(
selectinload(StepModel.next_step, recursion_depth=-1),
...
)
```
and it would be helpful to know if I can chain the same options to
next_step, so that step also has its actions and their relationships()
eagerly available etc.
PS
FWIW the reason I included `raiseload("*")` in options() is because I am
running async queries, and personally the error I will be confronted with
trying to access lazy attributes is more helpful, so I've come to add it by
default. Without raisedload(*) I would see:
```
sqlalchemy.exc.MissingGreenlet: greenlet_spawn has not been called; can't
call await_only() here. Was IO attempted in an unexpected place?
(Background on this error at: https://sqlalche.me/e/20/xd2s)
```
With the help of raisedload(*) I get to see:
```
sqlalchemy.exc.InvalidRequestError: 'ServiceActionModel.service' is not
available due to lazy='raise```
```
This helps me tackle those cases more easily one-by-one.
On Thursday, 14 September 2023 at 15:30:23 UTC+2 Mike Bayer wrote:
>
>
> On Thu, Sep 14, 2023, at 7:36 AM, Cornelis Poppema wrote:
>
> Hi all,
>
>
> I am new to sqlalchemy, I think the idea of what I am trying to achieve is
> relatively simple, but I can't seem to figure out how to retrieve
> `.service` in the same query. I failed to find an example in the 2.0
> documentation for exactly this.
>
> My attempts have been to simply chain a .selectinload after the
> .selectin_polymorphic, ie.:
>
> ```python
> .options(
> selectinload(StepModel.next_step, recursion_depth=-1),
> selectinload(StepModel.actionbases).selectin_polymorphic(
> [
> ServiceActionModel,
> ],
> )
> .selectinload(ServiceActionModel.service),
> raiseload("*"),
> )
> ```
>
> This gives the error:
>
> ```
> /usr/local/lib/python3.11/site-packages/sqlalchemy/orm/strategy_options.py:2442:
>
> in _raise_for_does_not_link
> raise sa_exc.ArgumentError(
> E sqlalchemy.exc.ArgumentError: ORM mapped entity or attribute
> "ServiceActionModel.service" does not link from relationship
> "StepModel.actionbases". Did you mean to use
> "StepModel.actionbases.of_type(ServiceActionModel)"?
> ```
>
>
> Hi -
>
> A few up front things, is it possible you can post these questions that
> have a lot of code on github discussions? That's at
> https://github.com/sqlalchemy/sqlalchemy/discussions . I'm sort of
> wondering how new users are still arriving here at the mailing list, did
> you find this list via the support page at
> https://www.sqlalchemy.org/support.html ? I would want to change the
> verbiage there to please refer people to GH discussions instead.
> Especially with these advanced inheritance eager loading problems, which in
> the majority of cases end up being real bugs in SQLAlchemy, as seems to be
> the case here (at least, there is an inconsistency in the API that somehow
> needs to be documented, or something).
>
> As for the question, first off this is really advanced usage and I've
> hardly ever seen people using selectin_polymorphic(), much less deep within
> a chain of loaders like this.
>
> The correct form for this load would follow from how it's described at
> https://docs.sqlalchemy.org/en/20/orm/queryguide/inheritance.html#combining-additional-loader-options-with-selectin-polymorphic-subclass-loads
>
> , where the ORM allows the selectin_polymorphic(Target, [TargetSubclassA])
> to be a sibling to the appropriate relationship load,
> selectinload(TargetSubclassA.elements). The example there places both of
> these options comma-separated within select().options(). This is the
> "inconsistent" part because I'm already surprised the ORM is allowing the
> selectinload() to be present against TargetSubclassA when that's not one of
> the primary entities in the select().
>
> However in your case, you are coming off of a parent loader option
> already. So following from this, the correct form based on a direct
> reading of those docs would, *in theory*, be:
>
> select().options(
> selectinload(Parent.target).options(
> selectin_polymorphic(Target, [TargetSubclassA]),
> selectinload(TargetSubclassA.elements)
> )
> )
>
> that is, you can build up sibling options from a parent loader option
> using another call to .options().
>
> however, this doesn't work; the validation of the loader chain
> unsurprisingly notes that TargetSubclass is not linked from Parent.target,
> and they'd like you to use of_type() instead. So I've made a bug for
> this here: https://github.com/sqlalchemy/sqlalchemy/issues/10348 as
> something has to change here, either the docs, or the usage pattern for
> selectin_polymorphic(), or the error checks have to get a lot smarter to
> figure this out and let this case pass, since it works fine the way you are
> expecting if I just have it skip the error checking.
>
> What you can do now is use with_polymorphic() instead that is more
> configurable for this kind of loader chain:
>
> TT = with_polymorphic(Target, [TargetSubclassA])
>
> select().options(selectinload(Parent.target.of_type(TT).selectinload(TT.TargetSubclassA.elements)))
>
> discussion can continue at the issue above or a new github discussion,
> thanks!
>
>
>
--
SQLAlchemy -
The Python SQL Toolkit and Object Relational Mapper
http://www.sqlalchemy.org/
To post example code, please provide an MCVE: Minimal, Complete, and Verifiable
Example. See http://stackoverflow.com/help/mcve for a full description.
---
You received this message because you are subscribed to the Google Groups
"sqlalchemy" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To view this discussion on the web visit
https://groups.google.com/d/msgid/sqlalchemy/3e4b053f-e583-4a20-98b5-abfd08479b60n%40googlegroups.com.
