fix issue with nested cardinality n tiles not doing a clean delete

[apeters]

No description provided.

[jacobtylerwalls]

Howdy howdy. Thought you wouldn't mind having a couple notes, hope it helps!

[jacobtylerwalls]

tl;dr this should already be stored as self.partial (should have been self.is_partial_update or self.leave_missing_children_alone, I guess).

The goal was always to push requests out the model layer entirely. It's here as a concession to support legacy post tile save functions. If you need this in a hurry, hopefully we can capture a follow-up to remove it?

If you're writing your own views, then you can parse args exactly how you like. If you're using the framework, here is where the view layer passes down arguments to the save method's partial= arg.

arches-querysets/arches_querysets/rest_framework/view_mixins.py

Lines 166 to 168 in ff194ef

	# Freeze some keyword arguments to the model save() method.
	is_partial_update = self.request.method == "PATCH"
	ret.save = partial(ret.save, request=self.request, partial=is_partial_update)

Looking at this now, partial=... is a terrible name given it's rhyming with functools.partial here.

(Is that spooky action at a distance? I guess. The tradeoff being action at a distance versus having to remember to call save() different ways in different views. And the idea being you could always write your own view if you got tired of doing it the framework-way, but otherwise try to provide reasons not to write your own view.)

[jacobtylerwalls]

Also, I'm not familiar with what request.GET looks like on put/patch requests, does it still give you querystring args? So it does!

[apeters]

thanks @jacobtylerwalls I'll look into the self.partial property. I think we'll still need to have the abilty to pass in a querystring param (which btw, works on PUT and PATCH requests just like it does for GETs), to allow users to opt in/out of tile deletes.

[jacobtylerwalls]

I wasn't clear on how much of an error condition this was to have missing keys; I thought we had a lot of knockout code that errored out pretty hard on them. Is the idea to gate this for inserts only, so that we take a "light touch" with existing tiles?

[apeters]

I'll have to look into this again. This check may be rendered useless now that a more final fix is in place. I think at first this was causing an issue.

[jacobtylerwalls]

Ouch. I'd love to hear a little bit more about what you found here.

[apeters]

missing pairs of tile id's from _pair_tiles cause this issue becuase they are considered "deleted" when in fact on a subsequent loop through the code those "deleted" tiles are in fact being updated.

Honestly, there is probably a deeper refactor required here that our budget just can't afford.

[chrabyrd]

code looks good, haven't given it a spin yet. 1 nit 😁

[chrabyrd]

nice 😄