Feature #18841: Proposal: autoload_relative

• Status: Open
• Assignee: -
• Target version: -

[fxn (Xavier Noria)]

In my experience, autoloads often reflect an existing hierarchical structure.

If a project does not use Zeitwerk, and the user declares autoloads for a class or module, chances are they are for child constants. As an example, see the ActiveRecord module. (Those ones do not have a second argument because we define wrapper that derives it by convention, here).

I think it would be convenient to have an autoload_relative in the line of Kernel#require_relative. It would make existing patterns more concise, and as a practical consequence, you skip $LOAD_PATH lookups too.

[ioquatix (Samuel Williams)]

Yes, great idea, and I can confirm I've run into issues with autoload loading the wrong files if you combine autoload with require_relative.

Also, since this does not exist yet, we can be more strict on how this works, as in autoload_relative can raise an exception if the constant is not defined correctly.

[marcandre (Marc-Andre Lafortune)]

Note that this is a duplicate of https://bugs.ruby-lang.org/issues/15330, which includes some real world usage stats.

Making it strict seems like a good idea too.

[fxn (Xavier Noria)]

Oh, I should have searched before! Can't close this one myself.

[ioquatix (Samuel Williams)]

Great minds think alike! Thanks @marcandre (Marc-Andre Lafortune).

@fxn (Xavier Noria) I think the next step is to make a PR and then get Matz to review it.

[byroot (Jean Boussier)]

• Related to Feature #15330: autoload_relative added

[rubyFeedback (robert heiler)]

I am not sure we should add more methods from within that
family (autoload-family). Personally I would even remove
require_relative, but I guess too much code depends on it
nowadays.