2024-11-26 Nodes & Physics Module Meeting

Check the overview thread for more information about the meeting.

Present

  • Jacques Lucke
  • Simon Thommes
  • Hans Goudey
  • Drew Medina
  • Dalai Felinto
  • Lukas Tönne
  • Sean Christofferson
  • Nrd Kraft
  • Mattias F
  • Iliya Katueshenock

Since the Last Meeting

Meeting Topics

  • Hair dynamics prototyping schedule
    • Generally we want to find the right tradeoff between prototyping and built-in C++ nodes compared to building features with building block nodes and node groups.
    • Lukas has been working on a node-based solver demo on the side. He gave a demo in the meeting.
    • As a next step for the project we’ll plan to implement a prototype with built-in nodes.
    • Simon will share files we can use for more realistic use cases, one with geometry nodes hair and another with the old particle system
  • String Key Count and LineSelect nodes
    • The building block nodes seem reasonable for various use cases.
    • It’s not a good solution for complex parsing like building an AST. It doesn’t have to solve those problems though.
    • The node and socket naming needs a pass for consistency.
    • The implementation needs a fair amount of work, for example to not reimplement UTF8 parsing.
    • The problem of linear time complexity scanning string character code points can be solved more generally later if necessary.
    • Find in String should output a count, rather than having a separate node for counts.
      • To be clear this node is handling non-overlapping matches.
      • Rename “Token” to Search.
  • Versioning for deprecation of nodes (e.g. Align Euler to Vector)
    • We didn’t do versioning in the past to improve forward compatibility.
    • List them on the compatibility page in the release notes.
    • In the future, when deprecating nodes and other features, we should be clear when they will be removed.
  • Bundle/Closure experimental feature
    • Jacques wants to ship the bundle & closure branch as an experimental feature sooner rather than later.
  • Nodes: support optionally deriving reroute names from origin sockets
    • The property should be displayed in the sidebar too
    • It would be nice to remove the “Node” icon to the left of the label string button if possible.
    • Simon questions whether the options need to be displayed in the node label popup.
  • Bug report: Field sockets are not drawing in Mac
    • Hans mentioned that the drawing of sockets was separated from keyframes. The bug was probably caused by this.

Need Help

  • Auto-label for the Warning node based on the type (similar to the Math node).
    • Sean seemed interested on this.
11 Likes

Regarding my PR, it seems that a few of us still have disagreements over the naming of this node.

I believe “find token” is more reasonable because script language compilers perform non-overlapping matching when parsing “tokens,” and the core part of this node is C++'s “find”.

The term “search” might imply “overlapping matching” in some contexts.

If we were to translate the node name into Chinese, I would also think that using “搜索” (search) is inappropriate; instead, “寻令词” (find token) should be used.