Another shape cut off by autobounds

[omgitsraven]

(sequence
  (sphere 0.8)
  (twirl-x 3 1 #[0 0 -1])
  (rotate-z 0.3)
  (move #[0.3 1 1.2])
)

Is a shape being cut off a case where it's still worth reporting autobounds problems, or is autobounds basically completely at the limit of what it can be used for at this point?

[omgitsraven]

Maybe off-topic, but: it's my understanding that auto-bounds works by transforming a bounding box successively, right? Is there any possibility of trying alternate approaches, like doing a very low resolution pass on the full extent of the manual bounds, then bringing it in to roughly the size of the shape that emerges (plus a margin)? (And re-extending that margin every time a higher-resolution pass adds geometry too close to the existing margin? Is it difficult to just re-calculate whatever's in the new area without having to recalculate the rest of the shape along with it?)

[mkeeter]

Like issue #129, it looks like the bounds are technically correct, but are too big to be useful:

(define s (sequence
  (sphere 0.8)
  (twirl-x 3 1 #[0 0 -1])
  (rotate-z 0.3)
  (move #[0.3 1 1.2])
))
(shape-find-bounds s)
--> (#[-1289.328125 -1945.703125 -11.2574462890625] .
     #[120.2314453125 3729.46875 11.8204345703125])

I'm going to disable autobounding for the moment, to prevent people from running into this and being confused.

The bounds code emphasizes catching everything, at the cost of being very conservative. For example, if you've got a model that's mostly centered around the origin, but has a single sphere floating far away in space, it will detect that:

(shape-find-bounds
  (union (sphere 1 [1 1 1])
         (box [-2 -2 -2] [1 1 1])
         (sphere 0.2 [1032 500 3])))
--> (#[-2.00006103515625 -2.00006103515625 -2.00006103515625] .
     #[1032.203125 500.19921875 3.20001220703125])

I'm going to make a general-purpose tracking issue for "shapes that give too-large bounds", and accumulate things there.

[mkeeter]

FYI, I wrote a little about how the bounding box estimation works here.

[omgitsraven]

Oof, sorry; I forgot that it falls back to the manual bounds when auto-bounds fails. I just tried it again in a fresh document and you're right, it's too big, not too small.