Verification vs construction

[boergens]

Wonderful tool, thank you for your hard work on this!
I could see a library like this used for two purposes:
(a) to supply snap-in information for a CAD program
(b) to verify existing LEGO builds for legality

If I understand correctly, the purpose of this library is (a). I am aware of some limitations for it to be used as (b) (my desired use case).
I have some ideas how to alleviate that, but maybe you could weigh in on that topic: how you would approach it if you had to?

Thanks in advance and best regards

[mxmws]

Hi!

You could test for legality by checking if all parts of a lego set are connected to at least one other part. But this would have some limitations:

1. Some parts of sets are not connected to any other parts and they are not supposed to be
2. Some parts may be connected but still be illegal (for example because they would bend - you may catch them by detecting overlap of the 3D parts but you will also have false positives from sloppy placements)
3. Some parts are not in the ShadowLibrary yet so they may appear to not be connected to any other parts

here is a graph I generated for set 10020 using the ShadowLibary

[boergens]

Thank you, these are great pointers!
One thing I encountered when applying the shadow library to preexisting sets is that 3641 and 4624 fail to connect because people mount the tire on the rim "flipped". So this would be a connection that should be allowed the other way round too. But for connecting 4624 to 4870 you have to be strict. Once you go out from there you hit a lot of special cases.

Cheers,
Kevin

[mxmws]

Yes, there are definitely a lot of edge cases. I guess the shadow library only contains the connection points, not the connection logic (?)
It seems like 3641 is not part of the ShadowLibrary so that's probably the reason it fails to connect. 4870 is also not in the ShadowLibrary, only the studs are inherited from a primitive. Or am I missing something?

I think we could easily add a lot more parts to the ShadowLibrary by using existing sets and probabilistically determining their position relative to connections of parts we already know the connection points of. I may do that soon

Best,
Maximilian

[boergens]

3641 was a poorly chosen example, for the reasons you are pointing out. I have an open PR for it, that's why it was on my mind: #12
But the point still stands.. For snap-in it's perfectly fine to enforce one orientation over the other, for verification the code needs to understand that for some parts the mirrored situation is also fine.

[mxmws]

There is also the issue that for verification you would need to choose a threshold distance. In some sets from the official model repository some parts are not snapped. I think this is because they were built before the parts were added to the shadow lib. If you choose the threshold too low you will miss real connections. If it is too high you will connect parts that are not actually connected. The image below shows that the green point and the green circle are not in the same place. One could write an algorithm to find these edge cases and approve them by hand.

[boergens]

Thanks for sharing that!
The tolerancing is indeed an issue, but hopefully not an insurmountable one:
Of the 1700 sets in my collection, these are the violations that my code found for 3024 stacked on top of 3024 (picked that just as a test)

Lateral (LDU shift)	Set	Note
0.010	10025-1 Mail Car
0.011	10264-1
0.020	10020-1 (×4 pairs)	Santa Fe Super Chief
0.050	10025-1 Mail Car
0.140	1682-1 (×2) + 1682 Space Shuttle (×2)	same set, two sources
0.142	75051 Jedi Scout Fighter (×4)
0.190	10022-1 Observation Car (×2)
0.193	10194-1

[mxmws]

Yes, the LDU shift in these cases seems negligible. But there are more extreme cases. For this one the two parts that are supposed to connect are almost 8 LDU apart from each other. (set=6054-1.mpd | part1=3623.dat | part2=2417.dat)

However this is rare and can probably be ignored depending on your use case. Do you mind sharing what kind of project you are working on? Feel free to reach out if you prefer to discuss this privately.

[RolandMelkert]

I never considered using the snapping engine to test legal connection etc.

It was initially targeted as a very simple placement tool, to be used together with normal grid editing.

Besides testing for used connections, it might also be needed to test on all unused ones to make sure they don't overlap in some impossible way.

Library wise I would also consider adding detailed bounding info.