Clojurians Log v2

after a git bisect the first bad commit: [5392b484517216abb6d26f9fb17f326b135d0f24] Fixed a bug in remove-entity. It was not removing nested idents that were not part of the top-level normalized graph.

might be somewhat related to https://clojurians.slack.com/archives/C68M60S4F/p1724270092902999

Created an issue https://github.com/fulcrologic/fulcro/issues/554

Right, but did you update to the latest? I thought I already re-fixed this

and is this a RAD app, or are you using remove-entity somewhere else?

Try 3.7.10-SNAPSHOT. I went back and did what I should have done to start: made it an opt-in change to prevent these regressions. Sorry about that. I’m pretty sure that new version will fix it, and I can go into RAD and adopt the new thing in reports, which is what I was trying to fix.

I am using git deps and had the issue with bb0b834a2c98996fa38a735e60293894c1cc27a4

Which was the latest commit yesterday

Even as an option this seems like a bug (I am using rag indeed, this happens in subforms)

The latest should no longer be removing them unless requested, and I didn’t update forms to do so. So, now I’m puzzled

if you mean this one then maybe fd9148b0ed8d89587db1d09dcff0a8723bebe2e6 but yesterday it wasn't there

and you’re right, a delete child in a form needs the old idents as data. Which is one of the reasons why the original implementation didn’t remove them. But, the new implementation is opt-in for that.

I told you a snapshot version. The commit is fd9148b0ed8d89587db1d09dcff0a8723bebe2e6

I understand that the latest commit makes the new remove cihld behavior optional, but I'm saying the behavior feels wrong, it turns vectors of identities in subforms into vectors of nil as soon as you "touch" the form, even unlrelated fields, not just when you delete a child. It doesn't do that just for one particular field but for all subform fields, it deletes everything. and it's not reversible, cancelling the form doesn't restore the children in that case

yes, I agree that in the form case it makes no sense, but in other cases, completely removing an ident from state Does make sense

now, I would agree that turning the ident to nil in a vector is wrong

GC in general is a major pain, which is why I generally don’t do it much 😄

I have not had the time to debug yet to really understand what's going on, but in my case I do not delete anything so there should not be any GC at all. A simple example would be: I have a form with :user/name and :user/friends, friend is a subform with :user/name as well. I have a user with 2 friends, I change :user/name, now :user/friends becomes [nil nil] instead of [[:user/id 2] [:user/id 3]], without me touching any of those

I understand what you’re saying in the context of forms. But there ARE cases (such as when working with a report that has a stale value in its parameters) where removing the idents makes perfect sense. The recent change I just made fixes your case to no longer be an accident

I had a quick look and seems like the only place where that remove-entity is used is in use-gc hook, which I only have in the pickers

interesting so it is when one of these pickers appear of the screen that this happens

but the new version doesn’t have problems, right?