Question Transform, matrix and direction manipulation

I feel like I'm running out of tools again. I am going to describe what I try to do, but I would like to use it as example for the wider context to reforge it into feedback.



The goal is to spawn burst of particles towards certain direction with some randomization. To do this I used cone, however I want to change cone direction every time spawn event fires. This can be done mainly with two methods:

1. Send event attributes (position + direction), inherit them and use to construct transform with LookAt.
2. Set some 'Transform' graph property and connect it as cone transform.

As I pointed in the other thread the LookAt has some issues when using with attributes or something like that, but let's ignore that for now. The first method could work, but the cone is not looking at Z axis, but Y and in result particle source is rotated by 90 deg on X. If there were quaternions or easy to use matrixes this could be solved by adding quaternion with shiften rotation, but there are no quaternions or transform multiplication or anything like that (or I just missed something).
I must go with the second method, but it requires the C# side to "know" that direction must be pre-rotated. This is not big deal to implement, but it breaks the interface between script and graph, script should not care what happens inside the effect, it just wants to play something here in that direction.
If I decide to create other graph for the same "interface", everything crumbles - I get the same problem but this time rotation has additonal unneeded shift.

If I haven't missed any simple and obvious solution, then my feedback is:

• Implement transfrom multiplication to combine them (not sure if it's possible, haven't checked the implementation) as adding angles produces gimal lock and undesired results
• I can't find multiplication of matrixes, multiplication has no such function
• Add node to create transform from matrix, even if I pass matrix to the graph I can't feed it to the transforms
• If possible add matrix * transform multiplication, but with transform as output
• Fix LookAt
• I have feeling that adding quaternion type would not be bad idea. This is big task and most of the time not required, but it would be nice addition.
• Official custom node please

Let me know if I mixed something!

 

Hi, thank you for the answer. I am on 2021.3.10 and connecting exposed property or attribute to LookAt produces error mentioned in the other thread, otherwise swapping axes with cross product would be viable solution as roll does not matter for cone.

I did not try to use it, because it's not possible to use it. The output of this block is matrix and it can't be plugged to transform. Well, I guess it's simple TRS matrix, so position and rotation could be extracted from columns, but there is no tool to do that - there is no way to read columns, rows or anything. This is why I asked for "the same" node, but with transform output if possible.

The angle is user friendly and as long as we operate in world / local space there is no problem with that. The problem starts when you need some transformation and angles are not driven by user. Transform component solves this problem quite well, user operates on friendly angles, but there is quaternion underneath.
It looks like it should not be needed as there LookAt in theory, I think it would be enough if matrix/transform operations were more polished (but I am not math expert too ).

 

Correction, this is weird. If I understand this correctly it throws error when retriving angles, but I am pretty sure I tried to use whole transform and it was the same. Now I am not sure if this is some ghost bug showing up in certain situation or I just did mistake.

 

OH. It's embarrassing. Again I am not sure if I bugged the UI in some way and it was not possible, or I was so tired and haven't even tried to connect them like this. For now let's assume it's my mistake - if I somehow manage to bug it like that I will let you know.

By the way what is the reason angles cannot be accesed from LookAt? @PaulDemeulenaere

 

No no - that's the thing xD
Right now this one compiled and works:

 

Ah, that actually would make sense as converting to euler angles probably requires bunch of if's and atans if I am correct.
Anyway thank you for help, now when I know transform-matrix interactions are possible and LookAt works it should be a lot easier