General Algorithm for Primitive Calculation?

[Nils Schmidt]

I got a 0.5036499999999999.

LDPE says it's 0.5036.
PrimGen2 says it's 0.5037.

Arrgh!

I feel your pain... At least I learnt things about rounding ;)
Looks like we were lucky when we cloned the original primgen, once we figured out things, file comparison sturned out OK. Or maybe we didin't stumble on a tricky case? Now I am not sure that a 1/10000 error from time to time would be a real issue...
This last problem is related to this https://en.wikipedia.org/wiki/Rounding#T...7s_dilemma ?

Finally, I found a practical solution to this :)
I simply round twice.

Code:

private String formatDec(double d) { // d is 0.00004999999 (string representation)
 return bigDecimalToString(
        new BigDecimal(String.valueOf(d))
        .setScale(8, RoundingMode.HALF_UP)      // Round to 8 decimal places (to transform 0.00004999999 into 0.00005)
        .setScale(4, RoundingMode.HALF_UP));    // Round to 4 decimal places (0.00005 -> 0.0001)
}

Assume that N is < 9 and X > 4.
This will convert any 0.0000N999X number to the next 0.0000(N+1) number. It solves the problem for the trailing 9's.
Okay.. 0.000049990 will not work, but I guess this has no practical impact.

Just for demonstration, the following primitve is now identical in both tools (PrimGen2 and LDPE):

Code:

0 Torus Outside  1 x 0.5455 x 1
0 Name: 17\t01o5455.dat
0 Author: Primitive Generator 2
0 !LDRAW_ORG Unofficial_Primitive
0 !LICENSE Redistributable under CCAL version 2.0 : see CAreadme.txt

0 BFC CERTIFY CCW

0 // Major Radius: 11
0 // Tube(Minor) Radius: 6
0 // Segments(Sweep): 17/17 = 1.00
0 // 1  9  0 0 0  1 0 0  0 1 0  0 0 1 4-4edge.dat
0 // 1  12  1 0 0 0.5455 0 0  0 0 0.5455  0 1 0  4-4edge.dat

Nobody would even need that torus...
Nevertheless, I can sleep well again ;)