Added Repo.embedded_load/3 function

[narrowtux]

as described in #3269

[josevalim]

Oh, interesting. Given this function doesn't use the adapter, maybe it doesn't belong on Repo but elsewhere. The question is where? Ecto.Schema? Ecto?

[josevalim]

Btw, really good job. The question is only where to place the code.

[narrowtux]

Yeah that's true. I think it belongs in Schema then.

[josevalim]

@wojtekmach @michalmuskala any thoughts?

[wojtekmach]

maybe it doesn't belong on Repo but elsewhere. The question is where? Ecto.Schema? Ecto?

Yeah, neither option seems optimal to me. Given embedded_load/3 is implemented and documented almost identically to Repo.load/2, I think I'd keep it on the Repo regardless of the caveat you mentioned (which I totally agree with). I think putting it on Ecto.Schema would be fine too but then let's mention both functions in their respective docs imho.

[michalmuskala]

Yes, I agree. We'll need at least a lot of cross-linking between them since they are related. I think there's another question and that is loading data that is not from database into the schema. Since, as far as I remember, for embeds load and cast is exactly the same this is generally already possible with Changeset.cast followed by Changeset.apply_changes. What should be the interaction there? It seems to me this might be a bit confusing as to what function you're supposed to use in what situation.

[narrowtux]

for embeds load and cast is exactly the same this is generally already possible with Changeset.cast followed by Changeset.apply_changes. What should be the interaction there? It seems to me this might be a bit confusing as to what function you're supposed to use in what situation.

One advantage over the changeset calls is that you don't need to repeat the list of fields.

The changeset approach offers more flexibility of course.

[josevalim]

I think there's another question and that is loading data that is not from database into the schema.

Good call. In those cases we may not have a repo, so putting it on a repo is a no-go. I looked at Ecto.Schema and it only has the schema definition functionality, functions that work on the schema at runtime are all in Ecto, so I propose to add it as Ecto.embedded_load and work on cross-linking them in the docs.

Can you please amend it acccordingly @narrowtux ? Thanks everyone!

[narrowtux]

Okay

[narrowtux]

@josevalim I moved the embedded_load/3 function to the Ecto module, and moved the test to embedded_test.exs.

[josevalim]

💚 💙 💜 💛 ❤️

[josevalim]

It feels like Ecto.embedded_dump/3 would really complement this feature, but this is good enough for getting started. Thanks @narrowtux !

[narrowtux]

Yeah, an embedded_dump makes sense! It probably also works with the Ecto.Repo.Schema.do_load function.

[narrowtux]

Also, I noticed that this function doesn't load the data into recursive schemas. I'll continue to work on this a bit