[Plugin] Manifold v2.2

driven · 24 Jan 2010, 20:49

fantastic,

go out on a family errand and come back to a whole new playing field, anything I can do to help?
a 'test' skp, if you need something more taxing,
or I'll try repairing it myself when you've got the next one up...
or need a tester... whatever you need...

about 25% of the poly count I would normally use for this sort of thing, and not triangulated.

john

TIG · 24 Jan 2010, 22:10

driven

I'll look at your problem - TT has already come up with one set of problems to fix !
With yours I get a non-planar points error - since there's only 3 of them it seems a bit strange...
We're really at 'beta' testing now - I only outlined the ideas some hours ago and issued this as my first go - it WILL get better

pilou · 24 Jan 2010, 22:15

Very useful for boats builders

driven · 24 Jan 2010, 22:50

@unknownuser said:

Very useful for boats builders

and don't forget the humble Air Bed makers of the world...

TIG,
I did this simply to save you having to draw-up or look for something relevant to test on

the last 3 letters in watertight might contain none planer points, one had to be over-drawn with a poly-line and two unsoftened with eraser on the bottoms, before I could JPP them, oh the bottom of that A I made a real hash of.. missed it out and went back... I was in a rush...

I made the ring quite well, two joints (triangulated) near you name, I chose a ragged font in an attempt to create tiny voids and was surprised how well it came out. Fredo's Joint PushPull is remarkably clean, I have another test piece on file that is pilfered with bad seams if that's more useful.

john

I'll give anything a spin, it's great that your on the case...

TIG · 25 Jan 2010, 01:07

Here's v1.2 http://forums.sketchucation.com/viewtopic.php?p=218691#p218691
Basic Ticker added. Better processing for inner faces.
Three-faced-edged external-faces now usually kept. [fixed TT's ball-breaker boxy mess ]
FR lingvo updated by Pilou.

Still struggling with driven's monster...

More feedback and ideas please...

driven · 25 Jan 2010, 01:20

@tig said:

Here's v1.2 http://forums.sketchucation.com/viewtopic.php?p=218691#p218691
Still struggling with driven's monster...

More feedback and ideas please... :ugeek:

that's the simple version, so far it's taken 2hrs 14 min and 49seconds just to explode all the groups so I can then re-group to test the plug on it (and it's only 600,000+ poly's)

I just fired up the other computer so I'll do some simple tests on it, while I wait

TIG · 25 Jan 2010, 01:24

@driven said:

@tig said:

Here's v1.2 http://forums.sketchucation.com/viewtopic.php?p=218691#p218691
Still struggling with driven's monster...
More feedback and ideas please...

that's the simple version, so far it's taken 2hrs 14 min and 49seconds just to explode all the groups so I can then re-group to test the plug on it (and it's only 600,000+ poly's)
I just fired up the other computer so I'll do some simple tests on it, while I wait

I've added some more error reporting in the console to my own version so I can at least see what these errors are - seems related almost exclusively now to healing very tiny holes ?
I'm thinking perhaps to move these vertices to heal these tiny gaps rather than fail making edges that are too small inside a 'rescue' ? It would change the model's form, but so slightly you'd never notice...

Watch this space

driven · 25 Jan 2010, 01:38

Did you scale it up even more?

TIG · 25 Jan 2010, 01:51

@driven said:

Did you scale it up even more?

Scaling it up x100 or so might make the tiny gaps 'face' - but I looking at pinching the nearly closed verices together so that the gap goes... without much success it has to be said !

driven · 25 Jan 2010, 02:53

works on Mac

no Jade highlighting on second test, this model was coloured, 1st wasn't

hole repair goes from every other vert on torus?

messages

2 then
manifold, orienting faces then
6
no option dialogs, etc...(should there be?)

CadSpan Style shows up no reversed faces after, GOOD

I'll try more

driven · 25 Jan 2010, 02:58

Sorry. shows up when I click on that repair, the first was there from the top, I'm sure....
EDIT: and it ungrouped separate geometry?

driven · 25 Jan 2010, 04:05

OK, am I following this a bit better now.

I got a Manifold dialog-box on the comp_fish when no repairs were needed and the new group is moved to one-side, [url]/viewtopic.php?f=10&t=25473&p=218811#p218811[/url]

on the other, after the repairs were done, all 3 two manifold bits stayed in place so I get the 6 meaning six manifold selection? and the box points just stay to one-side.

driven · 25 Jan 2010, 04:13

EDIT: CRASHED OUT SU 20 seconds no crash report, and SU restarted on a new drawing, now that has [bnever happened before, and I mean everever SU doesn't restart....[/b]]

how odd, so I think I'll play some more

TIG · 25 Jan 2010, 09:36

Run with the Ruby Console open for error reports - a new version will be available soon that I hope will be more robust...

thomthom · 25 Jan 2010, 09:56

@tig said:

Three-faced-edged external-faces now usually kept. [fixed TT's ball-breaker boxy mess ]

Very interesting. I was stomped on this when I some time ago tried to make a remove_inner_faces script. Will have to snoop around the code. I'll probably dig up some more fun geometry - thing I got a batch of trouble-samples...

thomthom · 25 Jan 2010, 09:58

@tig said:

I'm thinking perhaps to move these vertices to heal these tiny gaps rather than fail making edges that are too small inside a 'rescue' ?

You need to rescue? Doesn't add_face return nil if it fails?

TIG · 25 Jan 2010, 10:43

@thomthom said:

@tig said:

I'm thinking perhaps to move these vertices to heal these tiny gaps rather than fail making edges that are too small inside a 'rescue' ?

You need to rescue? Doesn't add_face return nil if it fails?

Yes BUT then I do things with the face that fail if it's nil - if the points are non-planar - which I can't see how but it rarely says they are - perhaps co-linear there's an error - therefore a rescue seems appropriate - I need to look at how I get these points.
At the moment to 'heal' a gap we find an edge with only one face and if it has a shared vertex with another edge then we try to use three points from their vertices' points to make a face - however it's possible for the points to be co-linear or virtually so and no face gets made... I'm looking at a more robust algorithm that ensure such edges are 'faceable' before trying it...

TIG · 25 Jan 2010, 10:45

@thomthom said:

@tig said:

Three-faced-edged external-faces now usually kept. [fixed TT's ball-breaker boxy mess ]

Very interesting. I was stomped on this when I some time ago tried to make a remove_inner_faces script. Will have to snoop around the code. I'll probably dig up some more fun geometry - thing I got a batch of trouble-samples...

That was a quick fix BUT it will fail with very convoluted shapes that have inner faces like driven's pretzel ! I looking at an even better fix...

EarthMover · 25 Jan 2010, 19:12

I'm still a little confused about the use of this plugin. Would it be of any value for anyone who is not doing 3D printing?

TIG · 25 Jan 2010, 19:33

@earthmover said:

I'm still a little confused about the use of this plugin. Would it be of any value for anyone who is not doing 3D printing?

Probably not... but it does remove internal partition faces in a volume and fixes any 'holes'...