Remove merge options covered by idiomorph

[potatotech]

Use cases for inner, outer, and upsert_attributes can all be handled by morph and signals.

[bencroker]

Can you please explain?

[delaneyj]

@potatotech upsert_attributes isn't covered by idiomorph

[potatotech]

Restored upsert_attributes

[delaneyj]

Needs to be removed from datastar.go and any examples as well but getting close!

[bencroker]

I think I’ve changed all instances in all examples now.

[bencroker]

main has been merged, fyi.

[delaneyj]

LGTM will put off until v20 release

[bencroker]

Use cases for inner, outer, and upsert_attributes can all be handled by morph and signals.

@potatotech can you please help me understand how inner can be handled when the default morph style used is outerHTML, which is not currently configurable in Datastar?

[bencroker]

If it turns out that the only redundant merge strategy is outer, then I’d opt for leaving it in, as it still offers the benefit of allowing you to target an element by selector without requiring matching IDs.

[bencroker]

I’ve come to the strong conclusion that there is no upside whatsoever in removing merge options. Even outer can be useful considering that the back-end should have full control over deciding what DOM elements to swap. Without it, an ID must already exist in the front-end. I vote not to merge.

[potatotech]

The backend is the frontend with hypermedia. If there's a missing id, add it in the template. When swapping fragments, datastar needs to target a specific element, and the only unique selector is id. Morph is sufficient because all other selectors are less specific*.

* There are things like #id[attribute] or #id.class, but id is already unique, so these are not necessary. You definitely should not be using them to store state either.

[delaneyj]

I'm going to leave it @potatotech, while I :100: agree you should always use the idiomorph I can see value in the effiency of inner/outerhtml for sheer speed in certain situatins. Thank you for your effort!