🔀 MERGE: Refactor/rename storage modules

[chrisjsewell]

This is a follow-up to #5330

Previously, I believe this may not have been possible,
due to django loading the database connection on install.
But now, it makes sense to consolidate the full psql_dos
storage backend implementation in a central place.
This also leaves aiida/orm fully storage backend agnostic.

[chrisjsewell]

tiny diff lol

[chrisjsewell]

P.S. obviously, the next thing will be to rename sqlalchemy to psql_dos

[chrisjsewell]

the next thing will be to rename sqlalchemy to psql_dos

Was too impatient, done

[chrisjsewell]

Note, the docs require #5365 to fix

[sphuber]

While we are in the process of renaming things as they should be named, should aiida/backends become aiida/storage?

[chrisjsewell]

While we are in the process of renaming things as they should be named, should aiida/backends become aiida/storage?

Yeh was thinking that, can do it in another commit here?

Then, there are other uses of "backend":

• should the aiida/orm/implementation/backends.py::Backend class be re-named?
• you have things like Entity.backend, but that's a bit more difficult to change

[sphuber]

Yeh was thinking that, can do it in another commit here?

Fine by me to do it in this PR, but not sure if these need to be separate commits. Well, at least the two you have now can really be squashed because it is the same folder being renamed twice in sequence:

1. aiida.orm.implementation.sqlalchemy - > aiida.backends.sqlalchemy.orm
2. aiida.backends.sqlalchemy.orm -> aiida.backends.psql_dos.orm

Is really the same as

1. aiida.orm.implementation.sqlalchemy - > aiida.backends.psql_dos.orm

The next move of aiida.backends -> aiida.storage maybe should indeed better be a separate commit, since it involves the renaming of more modules.

should the aiida/orm/implementation/backends.py::Backend class be re-named?

I don't think the name backend is problematic in itself, but for consistency it should probably be storage_backend/StorageBackend.

you have things like Entity.backend, but that's a bit more difficult to change

Why is this more difficult? From a backwards-incompatible perspective? Because if it is that, users are not supposed to be touching this anyway, right?

[chrisjsewell]

Well, at least the two you have now can really be squashed because it is the same folder being renamed twice in sequence
Is really the same as aiida.orm.implementation.sqlalchemy - > aiida.backends.psql_dos.orm

I would not say that is equivalent; it's a movement of some code between modules, then a name change of an entire module.

[chrisjsewell]

Why is this more difficult? From a backwards-incompatible perspective? Because if it is that, users are not supposed to be touching this anyway, right?

Yeh, although we don't "intend" users to touch it, Entity.backend is still a public attribute.
and just making sure everything is correctly updated, e.g. things like orm.QueryBuilder(backend=self.backend)

[chrisjsewell]

I don't think the name backend is problematic in itself

it mainly just not particularly descriptive and/or immediately obvious to new users/developers what it means

[sphuber]

it mainly just not particularly descriptive and/or immediately obvious to new users/developers what it means

I meant that I don't think the word "backend" is problematic, however, when it is isolated, it loses meaning. But when combined with a word indicating what it is actually the backend of, such as storage_backend, then I think it is fine and conveys the correct message.

[chrisjsewell]

But when combined with a word indicating what it is actually the backend ... then I think it is fine and conveys the correct message.

yep 👍

[chrisjsewell]

Ok @sphuber, done what I'm going to do in this PR

[chrisjsewell]

Ok I'm going to be merging this soon, since I have other PRs to go on top