Stuffing _7.dxf with _12.dxf

[9swampy]

Curious why you've added the C part to the DXF samples. Have you done any work to understand why the following happens (using some more of the sample shapes you've included):

The full donut I've added and the original DeepNest and the Port are quite happy to place the smaller parts in the hole, but the C part you had never gets filled.

DeepNestPort/DeepNestLib/Background.cs

Line 1189 in 50dad9f

// returned data only contains outer nfp, we have to account for any holes separately in the synchronous portion

Noting the comment about holes in Background.cs (and the limitations of the C++ import) shouldn't the hole in the C shape be serviced by the OuterNfp?

[9swampy]

Reworked the example to:

First placement of the donut returns SheetNfp of:

First placement of donut done. First part 12 returns FinalNfp of

plus

The second is selected to place the first part 12.

Resetting the Nest to:

First placement SheetNfp can place part 7:

Second placement returns a single candidate Nfp of:

Which means it's the C++ Minkowski.Dll that's failing. That's a shame because C++ is way out of my comfort zone...

Any chance the ClipperNfp alternative would fare any better?

[fel88]

Sorry I didn't catch.. Does this problem also exist in the original DeepNest project?

[9swampy]

Yes, it's a limitation of the C++ import. Didn't know where the problem was coming from when first asked. Clipper isn't substitutable easily, trying to see why not or find an alternative implementation.

[fel88]

Actually, C++ portion of code is just a wrapper for the boost polygons convolve operation. I'll try to add a feature to choose C++/C# pure implementation.
https://www.boost.org/doc/libs/1_76_0/libs/polygon/doc/gtl_minkowski_tutorial.htm

[fel88]

Now it can work without Minkowski.dll

[9swampy]

Clipper fares no better unfortunately but losing the C++ dependency altogether is an interesting proposition; revealed a bug in the caching I put in though (just as well you didn't adapt that bit) but simply isn't performant without the caching. Can confirm it seems to generate more invalid nests - but I'm not yet convinced that that's purely down to Clipper.

@petrasvestartas re original DeepNest C# Port #34 and @dharani811
re C# port how did that conversation progress since 2018? Hi @dorkmo you've been using this port since then too?

[dorkmo]

Havent been nesting lately :( wish i was able to help