There's lots to love here, but let me explain my reasoning behind previous_selector.

In today's CR, selector is a mandatory option in about 2/3rds of the operations, and it's been my experience that often a chain of operations ends up selecting the same element multiple times. By giving the chain a bit of memory, you can opt-in to the behaviour by writing less code. The simplest way to not by implicit is to just continue to do what you already do, today. I love this because it means there's no additional syntax to learn and it's actually discoverable.

There's also the relatively undiscussed stimulusreflex#110 which I recognize isn't as sexy but it comes with a very interesting capability, which is the opportunity to stagger individual operations. This provides developers with a chance to implement chains of functionality that fire off in a sequence. Being able to send a time-staggered set of operations to the same element with very little code is exciting - it's the sort of thing I use to build interfaces that feel alive. (I know this doesn't directly contradict a selector method but I'm trying to paint a picture of a really clean and sexy API.)

The main issue I have with this you probably saw coming because it's the same issue I had with stimulusreflex#91, which is that it introduces something that looks like an operation but isn't. I'm actually really concerned that this will be the straw that breaks the camel's back when it comes to people wrapping their heads around the library. So, while I can be honest and say that I don't love the proposed change, the API syntax is a big enough deal that I would be trying to find another way to do this even if I did love it.

Code feedback!

Only after typing all of the above, I realize/think that you're proposing the selector method in addition to the implicit selector, not instead of, right? I'm still 👎 for all of the reasons above, but I want to fully understand.

Why do you need to require active_record in operation_builder?

Why did you remove the prefix from the dom_id call? Prefixes are important, I think!

Actually, I just had that 🤯 feeling because I think you didn't remove prefixes, you just saw a higher-level abstraction for all three cases that I didn't previously perceive. So: can you verify that you didn't change the functionality, just improved the implementation? If so, I can get used to that!

I think that's everything. Thanks again, Marco!

As discussed on Discord:

• I removed the #selector method
• I removed the hash and delimiter params fromdom_id