Good one McNister.

Howdy ho ho ho,

I'm going to tackle this now:

poly-subdivide - no option to connect the corner points;

Upon further testing, the statement seems to underestimate the issue I initially perceived. I can't figure how to do more than that.

Here's what I'm after:


I cannot figure out how to do this in H. The “Subdivide” node is a type of Sub-D dicing which comes with its well known algorithms (Catmull-Clark, etc.) but it's not what I'm after. I often need polys to be simply subdivided, not “Sub-D divided”. Maybe I'm missing something, so I'll postpone filing an RFE for the moment, hoping someone will shed some light.

edit: I thought I tried all algorithms initially, but apparently one escaped me or didn't notice the change. Anyway, seems like “opensubdiv bilinear” algorithm produces the “+” type of division, although there's no way to not disconnect the edges.

edgedivide

Here's a statement from this SOP's doc that irks me a bit:
Close Path
Toggles whether or not the path traced out by the connected points should be closed. When a path is closed, the point on the last edge is connected to the point on the first edge. Since the connections are done with a minimum distance criteria, to get the desired results, the length of the path from the first edge to last edge should be smaller than the length of the rest of the path.

I think “close path” is a feature that does more harm than good. It obviously doesn't have any say for open meshes (1st attached image) and for “closed” selected edge rings it has the potential to mess things royally (2nd image) when that doc. rule is broken, as it basically forces you to mind the order of your edge selection, which is a huge mental overhead. Added risk due to complexity with no potential benefits.
We always need to consider practical cases, these tests on primitives tend to make thinks appear more simple than they will actually be in everyday work situations.

If someone thinks this feature is ever useful, I don't mind being left there, but there should be another option with an algorithm that doesn't care about selection order, which automatically closes the loop. ON by default.
I have never found myself thinking that it be cool if I had a feature that requires me an extra step to close the loops for full ring selections. XSI's simple approach (no “close path” feature) simplifies work with no practical compromises (there might some extremely rare cases which I never encountered).

Speaking of defaults, “share new points” and “connect points” should be ON by default.
Yes, I know I can create my own defaults, but I'm talking about sensible factory defaults. Sensible, as in not forcing all modelers to create these defaults, because I assure you, they will want these ON and they're the ones calling these modeling SOPs the most. I'm speculating that most non-modelers won't care either way about these defaults.

Lastly, I wish to see a “rounding” feature, like in the last image. The example might not be very convincing regarding its usefulness, but again, think about a more complex geometry where the edge ring selection goes around the model.

Here's another feature proposal for the edgedivide I just dreamt up, very “Houdini like” I think: a ramp distribution for the newly created edges.

edit: the dream spurred from my subconsciousness apparently, as I'm sure I knew the bevel tool in its newest incarnation has this feature too (Shape > Profile) in some form, although it does not control the edge distribution between the end edges (see attached), but rather the shape of the bevel's profile. So this should be a natural path to follow when edgedivide will get a facelift.

is maybe the polybridge helpy?

matthias_k
is maybe the polybridge helpy?

Regarding what issue?

https://sidefx.com/forum/topic/31040/?page=31#post-239508 [sidefx.com]

I have this one in mind :-)

I'm not 100% sure, but I don't think we're on the same plane of thought.

Then I have to say sorry for wasting time :-)

Thought you need something to connect 2 edges with an adjustable profile.
I've seen that you really want to make modeling tools better,
and my try was to say, that there is maybe nearly what you're looking for.

No worries…