Fix bug in cubic B spline and enhance class #22
Conversation
- instead of comparing output stream, compare floating point values with a
given threshold.
This but was effective only for non uniform splines.
|
👍 Looks good, thanks! However a test is missing for the new constructor. An example with non-uniform b-splines would probably be the best thing to do, since this was not really tested (and fixed) until now. Copying and adapting the |
- implement simpler dichotomy.
|
I have added a test that generates basis functions of non-uniform cubic B splines. |
| { | ||
| i = Double2SizeType::convert | ||
| (std::floor (static_cast<double> (imin) + (t - tmin) | ||
| / (tmax - tmin) * static_cast<double> (imax - imin))); | ||
| (std::floor (static_cast<double> (.5*(imin + imax)+.5))); |
There was a problem hiding this comment.
/home/travis/build/roboptim/roboptim-trajectory/src/cubic-b-spline.cc:196:53: error: conversion to ‘double’ from ‘roboptim::trajectory::Trajectory<3u>::size_type {aka long int}’ may alter its value [-Werror=conversion]
This should probably be:
(std::floor (.5 * static_cast<value_type> (imin + imax) + .5));
|
I fixed the warning. It compiles on my machine. |
|
Looks good. Just a question regarding d6c48e1 (first commit for the gradient test): was it related to a difference in the output for a different version of Eigen? Or a lack of precision in the output? |
|
The test failed because of numerical precision. At several places, the last decimal was different. I do not know precisely the reason, eigen version, different os... |
Fix bug in cubic B spline and enhance class
|
👍 |
Add a constructor with a vector of knots,
fix bug in computation of basis functions. The bug has no effect when knots are equally spaced.