Subject:
|
Re: number notation in official parts
|
Newsgroups:
|
lugnet.cad.dev
|
Date:
|
Fri, 12 Mar 2010 00:39:24 GMT
|
Viewed:
|
23878 times
|
| |
|
|
In lugnet.cad.dev, Philippe Hurbain wrote:
> > My suggestion would be to read numbers in that notation but never write them.
> I mostly agree with this policy.
>
> > I'm not sure why LDDP would write them that way (it would require some strange
> > output routines) but if one thing writes them then it's best to read them.
> There is nothing strange in the routines, it is the standard format used to
> represent very large or very small numbers in a limited number of digits.
I definitely know this. I see them far too much in my job as a numerical
physicist ;)
My point is that to write in mixed format (some %f and others %e) requires some
strange coding unless there is a weird language which does it automatically. I'm
pretty sure that to do it in C or Fortran would be quite hard.
> Actually since values never get very large in LDraw parts (biggest is probably
> 50x50 baseplate, thus 1000 ldu), it's all a matter of proper rounding. As far as
> I can tell, 4 digits after comma is OK for position coordinates, and 6 digits is
> enough for matrices coefficients. The problem mentionned with values such as
> 1e-15 should not have occured if I had properly rounded the file.
Yes. Actually LDraw files should always use a fixed digit rounding and never a
percentage rounding.
> Note also that LDDP has an auto rounding feature (on by default) that would have
> prevented the problem from the beginning. But I disable it since this may
> accumulate errors during complex parts work.
>
> Also note that MLCad also cherfully generates scientific notation numbers (eg.
> 9.9e-005) for any number below 0.0001
>
> Philo
That is bad behaviour IMO. It implies an accuracy which is not there. You should
never display anything more accurate than the maximum error (give or take a
bit).
To expand on your comment about proper rounding: The nature of operations in
LDraw files is such that fixed digit rounding is better in almost all
circumstances (except where you have a scaling and rotation, a rotation of a
huge or poorly centered submodel or many multiples of rotations such as in
complex part design). In a model file you are only interested in results within
x LDU where x is something less than 1 and in a standard model x will be at
maximum around 1000 times the error in a rotation in the most extreme case.
Tim
|
|
Message has 1 Reply:
Message is in Reply To:
 | | Re: number notation in official parts
|
| (...) I mostly agree with this policy. (...) There is nothing strange in the routines, it is the standard format used to represent very large or very small numbers in a limited number of digits. Actually since values never get very large in LDraw (...) (15 years ago, 11-Mar-10, to lugnet.cad.dev)
|
13 Messages in This Thread:
 
 
- Entire Thread on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
This Message and its Replies on One Page:
- Nested:
All | Brief | Compact | Dots
Linear:
All | Brief | Compact
|
|
|
|
