What if we could make it harder to model...

[Rodney]

I had an odd thought today...

 

What if we had a mode of modeling that wouldn't allow more than two splines to connect to the same Control Point?

 

I can think of one particular downside to this (more on that later), and surely there must be more problems waiting in the wings but the thought is/was that while most folks are looking for easier ways to model another solution might actually be to make it a little bit harder to model in that... while in that particular mode... modeling with less than optimal splinage would be harder if not impossible.

 

The downside would be where we want to connect discontinuous splines to another spline.

But in this case we (the modelers) may be attempting to introduce an other-than-optimal spline topology into our model.

 

Now, there would be a special case where continuous splines would/should be allowed and that is if the spline is one of the terminal ends of a continuous spline.

 

This then does beg the question of how we could better differentiate the terminal CPs from the internal CPs of a given spline.

 

These thought occurred to me while considering that many folks spend a great deal of effort trying to shore up and improve the bridges to the polygon world but this is often at the expense of the further development and benefits of spline-patch technology and rarely an aid to it. In other words... building bridges to inferior technology.

 

So I guess primary questions might be, "Should the user be prevented from making errors in ideal spline topology and if so what might success look like?

 

A tentative answer might be that A:M might initially refuse to allow the connection but the user could easily override that 'safety' (via holding down the Alt Key or some other equivalent).

 

From a programmatic view the modeling process (in that mode) would be streamlined considerably because after two connections that particular connection/vertex would be locked or closed. This could lead to a number of other possibilities for modeling with splines. As I see it A:M wouldn't even have to process these because although the user might try to attach additional splines/CPs they won't be able to stick to the model. Why won't it stick? Because it's not going to yield optimal topology.

 

Aside: This is a little off topic but relates to spline continuity.

Have you ever considered that a spline circle isn't (i.e. cannot be) perfectly continuous?

This is a similar problem to the real world conundrum of creating a perfect sphere... technically we can't as there is always going to be a start and stop point. The classic remedy/solution is that of scale. If you can make the 'error' so small as to not be noticeable it is an equivalent to not being there. Other classic remedies include 'polishing' or masking with other textures or shapes. But none of these change the fact that the starting and stopping points in the original form are still intrinsically there. They simply obscure/hide that evidence.

 

When considering the plight of the circular spline another potential solution presents itself *if* the start and end points of the spline can transfer to another place on that spline. The ideal place for the terminal points to be with respect to a single point of view would be out of site... on the back side of the circle, sphere, etc. And because these terminal points could be told to stay in that relative place on the back side of any 3D shape these terminal points would effectively disappear. They'd also be a lot easier to find at that location should those that know they exist look for them there. A continuous spline (circle) that can adjust/transfer it's terminal points to another segment along the length of it's spline might be useful.

[robcat2075]

Interesting idea. Maybe there could just be a siren that goes off every time you cross more than two splines.

[largento]

Never in my life have I ever wanted something to be more difficult to do than it already is.

[Roger]

It sounds like an interesting idea.

 

I actually think it is more difficult to go back and fix something you screwed up, so a mode that didn't allow non-optimal splines might not be a bad idea. I'm better than I was but I still make mistakes when modeling, and it can be hard to find where I've made the error or sort out how to fix it.

[Fuchur]

Why not indicate a "bad" connection instead of blocking it?

Sometimes a "bad" connection is all you need and if I see people struggeling with 5 pointers for instance it is just something they get upset about if they select 5 CPs and it can not be changed into a 5pointer but they just do not know why...

 

See you

*Fuchur*

[Rodney]

I had thought of suggesting various colors... an alternative to preventing error would certainly be to tag the error with a different color, larger (or smaller) graphic, etc.

But while nice in it's own way, it must be noted that such an approach doesn't get to the root of the problem which is what an error-free workflow would (theoretically) be designed to address.

 

Sometimes a "bad" connection is all you need and if I see people struggeling with 5 pointers for instance it is just something they get upset about if they select 5 CPs and it can not be changed into a 5pointer but they just do not know why...

 

This is why I suggest it would be a separate modeling mode that one would have to enter into because I agree that sometimes the lazy/sub-optimal way is good enough.

Heck, the whole world is a slave to that approach with tris... and lots of folks are content for it to remain that way.

But if a better way is possible...

 

An example of this is Simcloth which when first implemented was following a polygonal structure that hindered the implementation in many ways.

The programmer (Bob if I recall correctly) started again and pushed the tech into the realm of splines and patches to very good effect... with us as the beneficiaries.

 

As for 5 point patches... that's a bit of another topic and the short answer is; hit the period key twice to deselect and then reselect the 5 CPs.

I'm not entirely sure what that does behind the scenes but I've always thought it reordered the CPs so they get recognized correctly as an eligible 5 point patch candidate.

I have a vague recollection of someone working toward an implementation of an automatic 5 point patch methodology (here I assume that would entail A:M creating a 5 point patch automatically out of any and all eligible 5 point patch locations. I would think that a precursor to that would be that a plugin could be created that would traverse through a model and identify (and group?) all 5 point patch candidates. A subset of that plugin would then be to turn each of those into actual 5 point patches (i.e. a deselect/reselect operation followed by a 5 point patch assignment of the CPs).

 

But back to the original premise of the topic...

 

In order to even consider implementation one must first consider what benefit could be realized.

If there is no benefit or very little benefit then there is very little reason to implement the thing.

Programatically I can think of one benefit and that would be as follows:

- Let's say that we start modeling in the 'ideal splinage' mode and take that model to the point where no other optimal spline connections can be made.

- We then save the model and exit that mode into the general mode that everyone uses today.

- When the model is complete (or at any point thereafter) we save the model under a different name

- Now we run a comparison of the first and second models and note the differences (via this comparison we can identify where special cases have been implemented or may yet be made).

- At this point the system (and users) can 'learn' what ideal topology actually is (without starting from scratch every time if that is also optimal or ideal).

 

Now we could go farther and suggest additional benefits but should we even speculate?

One might be the implementation of a Subdiv routine that could mirror and likely outperform similar implementation in other programs.

With ideal topology this and more can be developed but if A:M has to constantly deal with infinite implementations of bad topology the task becomes infinitely more complex.

 

[Fuchur]

As for 5 point patches... that's a bit of another topic and the short answer is; hit the period key twice to deselect and then reselect the 5 CPs.

I'm not entirely sure what that does behind the scenes but I've always thought it reordered the CPs so they get recognized correctly as an eligible 5 point patch candidate.

I have a vague recollection of someone working toward an implementation of an automatic 5 point patch methodology (here I assume that would entail A:M creating a 5 point patch automatically out of any and all eligible 5 point patch locations. I would think that a precursor to that would be that a plugin could be created that would traverse through a model and identify (and group?) all 5 point patch candidates. A subset of that plugin would then be to turn each of those into actual 5 point patches (i.e. a deselect/reselect operation followed by a 5 point patch assignment of the CPs).

 

It is not that easy. If the surrounding structure is not well formed, you can hit the period-key as many time as you wish (I do it differently > hiding everything else and reselecting again with the rectangel-tool, but that is not important... actually it only makes a difference if you have selected the first controlpoint with shift-pressed or if you selected the first one without it pressed and the second with shift pressed... why is that? I am not sure, but it is the problem ).

 

That is the problem people have with that if you ask me and that is what makes it hard to solve, if you ask me... but anyway I get the point but still think an indication (just give the CPs another color or the splines which are not well formed) would be better than an own "harder to model"-mode. I just would wonder all the time why there is a mode that makes it harder to model as a new user... It just sounds very odd .

 

But a button like "Show endpoints" or "Show non-continous splines" would be a good idea. It would give you a much better idea what is wrong for instance if there are two points very close to eachother without a connection, etc.

See you

*Fuchur*

[Rodney]

Thanks for the response Gerald... I appreciate the time you take to provide input.

It certainly helps.

 

It is not that easy. If the surrounding structure is not well formed, you can hit the period-key as many time as you wish (I do it differently > hiding everything else and reselecting again with the rectangel-tool, but that is not important... actually it only makes a difference if you have selected the first controlpoint with shift-pressed or if you selected the first one without it pressed and the second with shift pressed... why is that? I am not sure, but it is the problem ).

 

I can only speak for myself but I'd say that hitting the period key twice to deselect reselect a five point patch is successful over 95% of the time.

I want to say 99% but I'd be hard pressed to prove that. I will say that out of the time that it doesn't work almost invariably I am dealing with a collection of splines that is not sufficiently constructed to make a 5 point patch in the first place. In other words, it's splines layout prevents it from being eligible to be closed as a 5 point patch. Note that I'd have to take a good look at this to see if having the suggested 'optimal spline layout' would still allow areas where the illusion of a legit 5 point patch could be created exists. At a guess I would say that it would... and if so... optimal spline layout (in the sense that only two spline could go through any given CP) throughout a mesh might be problematic for closing 5 point patches.

 

Aside: This is as good a time as any to remind folks that a single spline can be a five point patch. Not all versions of A:M allow this but most do. I believe the lastest releases allow this more than those just a few releases ago but almost all versions will allow a single spline to be closed as a 5 point patch *if lathed*. In other words, if A:M created the 5 CPs then it can almost always be closed as a 5 point patch (because A:M tends to only create legitimate/optimal splines/patches. As I say though, currently in A:M this is working also without lathing. (See attached video). I know that in the past I haven't always been able to draw a single 5 CP spline and close that as a five point patch.

5 Point Patches.mp4

[Rodney]

I just would wonder all the time why there is a mode that makes it harder to model as a new user... It just sounds very odd .

 

It sounds odd because it would be odd.

It would be harder *for A:M* to model. Easier for us.

We don't generally see what goes on behind the scenes we just want it to work for us (athough A:M also lets us dig deeper than the surface as we turn additional options on... and with the SDK even deeper if'n we want).

 

A:M is one of the best programs on the market at shielding the user from arcane stuff.

It's like that old saying about taking your car in to a mechanic to have the engine worked on and watching as the technician rips (seemingly random) parts out of it.

The typical reaction might be to react a bit emotionally, "What the hell are you doing to my car!"

Which is why we generally don't want to see that stuff.

[robcat2075]

I guess A:M does have some "expert mode" circumstances, like "show advanced properties" which is not on by default.

[Rodney]

I guess A:M does have some "expert mode" circumstances, like "show advanced properties" which is not on by default.

 

There are many rich veins to plunder in related areas. I consider direct editing of A:M's files in a text editor and development of additional features/utilities through the SDK as elements of 'expert mode' as well. The fact that the text editor or programming application launches outside of A:M shouldn't be too much of an issue for someone with that particular expertise.

[pixelplucker]

Going to the poly.world from AM isn't too bad but the bigger issue is if someone wants to convert the model to a sub d surface.

What would be nice it to have matched edges on surfaces which could allow separate groups to act as one much like surfaces in nurbs, maybe expand on the shift click.

 

Trying to make a continuos mesh especially with mechanical models can be tricky. Something as simple as punching out a hole in a cube is harder in AM.

 

I would like to see is a freeze option so bias handles don't move and don't auto adjust. I found it annoying to have to go back over a model and have to straighten them out after I had them fixed just because I moved a cp.

 

Lastly a auto cap option when exporting out stl files for all those that print thier models out.

[Fuchur]

use the peak mode. it will preserve the bias handlers(yes you can set biases in peak mode) and you can move cps around without changing the biases.