Stella4D vs. Miratope for 4D element types

[H A M]

Generally, Miratope tends to be more accurate in distinguishing element types than Stella4D for non-compounds. While spittid has 3 element types according to Miratope, Stella4D counts only 2. Also, Miratope can fake out several pseudo-uniforms that Stella4D can't.
There's one edge case that Miratope gets it wrong: SMALL RHOMBIHEXAHEDRON!

[robertw]

Can you provide examples other than spittid? Need examples in order to fix things.

[H A M]

There's obviously its conjugate, spiddit, and several idtessids.
Stella obviously fails at J37 prism. You can build it in Miratope using this OFF file: https://polytope.miraheze.org/wiki/File ... cupola.off

EDIT: Here's the link to my fork of Miratope: https://github.com/H-A-M-G-E-R/miratope-rs (you have to compile it yourself)

[robertw]

J37 is easy enough to build in Stella4D. Load J37 then use "4D->Create 4D Prism on Current Polyhedron".

Maybe your instinct is better, but it's not immediately obvious to me when it gets types wrong, eg in spittid. Takes me some playing around to see the pentagonal prisms come in two types. So unless someone tells me which ones are wrong, I won't know. I've been playing around with this and have improved it, but will only be testing cases I know about (you'll need to be more specific than "some idtessids" if I'm going to test them, and Stella4D doesn't have the idtessids yet anyway).

I have it working better with spittid now: 3 cells types (up from 2), 9 face types (up from 5), edges still 3 types. I think that's right except there should be 4 edge types?, so not quite there yet.

J37 prism looks like it already had 5 cell types correctly. Not sure what the others should be but I now get 9 face types (same as before), 6 edge types (up from 3), and 2 vertex types which seems right, up from 1. How's that sound?

[H A M]

Yes, siggissido has 5 cell types, 9 face types, 6 edge types, and 1 vertex type according to miratope but only 7 face types and 4 edge types according to Stella4D. Here's the link for OFF files for some of the idtessids: https://drive.google.com/drive/folders/ ... JUcZj4MX4D

[H A M]

J37 is easy enough to build in Stella4D. Load J37 then use "4D->Create 4D Prism on Current Polyhedron".

Maybe your instinct is better, but it's not immediately obvious to me when it gets types wrong, eg in spittid. Takes me some playing around to see the pentagonal prisms come in two types. So unless someone tells me which ones are wrong, I won't know. I've been playing around with this and have improved it, but will only be testing cases I know about (you'll need to be more specific than "some idtessids" if I'm going to test them, and Stella4D doesn't have the idtessids yet anyway).

I have it working better with spittid now: 3 cells types (up from 2), 9 face types (up from 5), edges still 3 types. I think that's right except there should be 4 edge types?, so not quite there yet.

J37 prism looks like it already had 5 cell types correctly. Not sure what the others should be but I now get 9 face types (same as before), 6 edge types (up from 3), and 2 vertex types which seems right, up from 1. How's that sound?

Did you borrow Miratope's algorithm to improve 4D element type calculation?

[robertw]

Did you borrow Miratope's algorithm to improve 4D element type calculation?

No, I have no intention of looking at how Miratope does anything. Frankly it would be much harder trying to understand someone else's code than just doing it myself. This has been on my to-do list for many years along with ideas of how to improve it.

Thanks for the siggissido example. Currently it's still coming up same as before, but I can see that the next step will improve that (and hopefully fix it completely), and will also separate the final face type for spittid. Hopefully it will work for any polytope, but still won't currently work for compounds.

[robertw]

I've got it working for all the test cases you've given now. Spittid and spiddit are 3, 9, 4, 1 (cell types, face types, edge types, vert types). J37 prism is 5, 9, 7, 2. Siggissido is 5, 9, 6, 1.

It still can't distinguish chiral pairs or types in a compound. Those will probably have to wait.