preview 2022.5-419
2022.4.12 working fine
preview 2022.5-419
2022.4.12 working fine
We canβt really determine what the exact problem is from the screenshots alone, they also show different patches - could you upload a patch showing the issue please?
aaand I recreated β<=β node and all start to work perfect
but I should mention that I decided to write at the forum because I encountered the same problem in third time. Very hard to reproduce, I usually solve it by adding conversion nodes
!:
Try to help the type inference by
<=
node and choosing the integer version orLength
node and choosing the IHasMemory
versionPlease keep the original problematic patch and upload here, so that we can tell whether we have to consider it a bug.
@gregsn
Iβve lost that part for the moment, also Iβm not sure if it would show up after a restart.
But if I run into it again, Iβll be definitely bringing it here.
As I remember the integer version didnβt want to connect and compare nodes, and everything was in the βIHasMemoryβ version.
on last screenshot β<=β absolutley the same
switching to the IHasMemory version helps, seems to me.
Ok I think I understand your confusion. You should use the node Count
to get the number for items in a spread and not Length
. Length
is meant for vectors and such and therefor outputs a float, while Count
is meant for collections and should always output an integer.
Thanks! I will only use βCountβ from now on
But βLengthβ is working!)
So there is ambiguity that cause unpredictable error β sometimes working, sometimes not. Or did I misunderstand something?
It feels like at one moment something breaks right inside the node
The Length
node is adaptive, our system tries to find an implementation based on the value connected to it. If you place it without connecting it, youβll see it outputs a Float32
. Here we see its original intent, that the input should be some sort of vector. Unfortunately in case of a Spread
it finds a βfittingβ implementation through the IHasMemory
interface the spread implements. To make things worse that selected implementation returns an integer and NOT a float, leading to issues youβre experiencing that it seems to behave different from patch to patch.
To address this properly we will need to fine-tune some rules how adaptive candidates are found / matched.
Thanks so much for the explanation!