fix(docs): ensure class member order follows source

[Umpire2018]

Description

At first glance, other modules works fine. e.g. https://docs.litestar.dev/main/reference/middleware/logging.html#litestar.middleware.logging.LoggingMiddlewareConfig

After digging in, i found that if a class is imported into __init__.py via from ... import ... and added to __all__, Sphinx may lose the original source order of its members so i changed to document classes directly from their original module, rather than through indirect imports in __init__.py.

Before	After

Closes

close #4393

[github-actions]

Documentation preview will be available shortly at https://litestar-org.github.io/litestar-docs-preview/4436

[provinzkraut]

Interesting find! Does this replicate in a fresh project? If so, I think you should submit this as a bug to https://github.com/sphinx-doc/sphinx/.

As for the proposed fix on our side, I'm not entirely sure this is fine to do, since it would mean that only members of litestar.middleware.base get documented, and things that don't originate there but are imported into litestar.middleware are not. We could of course explicitly document litestar.middleware.base, but then I think it would be confusing if the import is from litestar.middleware import ASGIMiddleware, but the docs are for litestar.middleware.base.ASGIMiddleware.

@euri10, got some thoughts on this? I know how much you like to consult documentation :)

[euri10]

vey interesting finding yeah.

i agree with you it would be confusing, and if i had to choose i'd sacrifice the order for the "correct" import.

now the order is already weird to me in other cases, not just class member, if you look at https://docs.litestar.dev/main/reference/middleware/constraints.html for instance i find it weird to get exceptions before classes for instance,

[provinzkraut]

Alright, then let's merge this and see if can can sort out the imports again once the bug has been fixed upstream.

[euri10]

Alright, then let's merge this and see if can can sort out the imports again once the bug has been fixed upstream.

any issue to follow ?

[Umpire2018]

I think it would be confusing if the import is from litestar.middleware import ASGIMiddleware, but the docs are for litestar.middleware.base.ASGIMiddleware.

1. I agree with you on that. After investigating more on that, i think it may be better to leave it to what it is nowadays because i don't think upstream will fix this autodoc related issue considering there are lots of ongoing tickets right now.

2. It seems like class members are in alphabetical sequence elsewhere instead of by source. See here.

Maybe we can consider alternatives like https://docusaurus.io/ ?

[euri10]

idk https://docusaurus.io/ to be fair so I cant comment on this
I had good success with the autoapi extension but I have to admitt I didnt dwelve into all the details, but overall I found it to be easier to use and have a nicer look than autodoc, but that is just a very uninformed impression

[Umpire2018]

now the order is already weird to me in other cases, not just class member, if you look at https://docs.litestar.dev/main/reference/middleware/constraints.html for instance i find it weird to get exceptions before classes for instance,

@euri10 Maybe this is expected behavior ? Because this is the sequence in source code and this PR haven't introduce relevant changes.

litestar/litestar/middleware/constraints.py

Lines 27 to 40 in 94073d3

	class MiddlewareConstraintError(LitestarException):
	pass
	class ConstraintViolationError(MiddlewareConstraintError):
	pass
	class CycleError(MiddlewareConstraintError):
	pass
	@dataclasses.dataclass(frozen=True)
	class MiddlewareForwardRef:

[euri10]

now the order is already weird to me in other cases, not just class member, if you look at https://docs.litestar.dev/main/reference/middleware/constraints.html for instance i find it weird to get exceptions before classes for instance,

@euri10 Maybe this is expected behavior ? Because this is the sequence in source code and this PR haven't introduce relevant changes.

litestar/litestar/middleware/constraints.py

Lines 27 to 40 in 94073d3

	class MiddlewareConstraintError(LitestarException):
	pass
	class ConstraintViolationError(MiddlewareConstraintError):
	pass
	class CycleError(MiddlewareConstraintError):
	pass
	@dataclasses.dataclass(frozen=True)
	class MiddlewareForwardRef:

yep sorry for the confusion i was speaking in general, its expected behaviour as you rightully say, with autoapi this would be ordered differently I think without thinking about how it's organized in source,