Some thoughts after 100+ hours usage

[nowakf]

First, it's worthwhile to mention that the reason why I've used this program so much (certainly more than 100 hours) is because it's very good.

That said, I wanted to compile a list of the most common problems I have, as a user, and further, to sound out how you guys feel about me going and writing some patches to fix them.

1. Invalid NURBS surfaces. My intuition is that most of the invalid surfaces arise when you have two solids, and the edge of one solid rests exactly on the edge of another solid. The fix here is usually to offset the constraint so the solids slightly extend into eachother, which is a pain.

Obviously, I'd like to be running on more than intuition here, so I'd like to try and characterize and collate what geometry leads to invalid NURBs,

2. It's often hard to see the geometry you're interested in. Allowing users to hide selected objects would be really nice here.
3. Variadic selections (I'm aware this has been mentioned before) would be a huge timesaver.

Any thoughts, ideas, stuff I've overlooked?

[phkahler]

1. Invalid NURBS surfaces. My intuition is that most of the invalid surfaces arise when you have two solids, and the edge of one solid rests exactly on the edge of another solid. The fix here is usually to offset the constraint so the solids slightly extend into eachother, which is a pain.

@nowakf The master NURBS issue tracker is #738
Having said that, you are correct about most of the problems involving coincident geometry (see #540 for an example that is now "fixed") but more specifically when surfaces are tangent at the intersection as in #1291. In that issues, @ruevs has highlighted a specific line of code that seems to be "wrong", but it's not clear what the "right" solution is. The edge should be found by the #540 fix, but seems to be discarded in the linked code when the 3 surfaces are tangent at that edge.

There are some issues like #1091 that fail depending on the length or depth of the geometry rather than the topology. In my mind I call these "proximity" issues, where only distances seem to affect weather it fails. Also in that issue is an example here: #1091 (comment) that I looked into a bit and it seems to involve failure to split trim curves properly. My test "fix" kind of verified that but wasn't a real solution.

You'll need more understanding of the general approach to boolean operations in solvespace to make progress on this. We definitely welcome anyone to try to improve things here, so please ask more specific questions in any of the open issues where you'd like to understand better.

2. It's often hard to see the geometry you're interested in. Allowing users to hide selected objects would be really nice here.

Have you tried "Suppress this groups solid model" in the property browser? When you use this, the group name should appear greyed out in the main list of groups so its easy to find again.

3. Variadic selections (I'm aware this has been mentioned before) would be a huge timesaver.

This is IMHO a high priority usability issue. There are 2 open issues on it and 2 or 3 PRs but we need to get the implementation right, for some definition of right. Look around at the PRs and the issues in the 4.0 milestone for some discussion.

Any thoughts, ideas, stuff I've overlooked?

TBH I thought your post was going to be a LOT longer, since there are many ways solvespace could be improved.

[ruevs]

@nowakf welcome to SolveSpace I'm glad you like it :-)

On NURBS - @phkahler already gave you a very good overview. Here is one more "simple" and interesting case #1312 (comment) and a very unscientific classification of failures #1291 (comment) from me.

Any thoughts, ideas, stuff I've overlooked?

I too thought your post was going to be longer ;-)

The pinned issue has a general list of commonly requested things/bugs/features that so far no one has had time/desire/skill/inclination to implement/fix ;-)

[nowakf]

Well, I was trying to keep it somewhat conservative: obviously, there are many features that would be nice to have. I was trying to keep the list contained to things that were either a serious and reoccurring pain, or a relatively simple change.

Regarding 'suppress solid model', there are a couple of problems: first, the ergonomics are bad; you must go up the list of features and select a check box on each one. It's not always obvious what group is responsible for what solid. Second, you're often still confronted with an awful lot of geometry. If you're making an involute gear, for example, where you rotate a fairly complex construction, you can often end up with a fairly hairy set of lines and points, only some of which you are interested in.

Anyway, thanks for all the pointers. I'll start poking around: has anybody tried writing some tests for determining what geometry produces invalid NURBS? If not, that is probably where I would start.

[ruevs]

By the way there is an issue on #2 already #1007.
And "Variadic" constraints were just merged (#1314) so I will close this.

[ruevs]

#1328 just for cross reference ;-)