Spline Toolbox 3.0 Release Notes (Spline Toolbox Release Notes)

Spline Toolbox Release Notes

Major Bug Fix

The Spline Toolbox 3.0 includes several bug fixes, including the following particularly important bug fix.

spmak and ppmak Use Size Argument to Correctly Construct Multivariate Splines

If the spline to be constructed by spmak (knots, coefs) is multivariate, but is meant to be a constant function without any interior knots in its last variable, then the last dimension of coefs is necessarily 1.

For example, a one-piece trivariate constant function would have knots stored in a length 3 cell array and a 3-D coefficient array.

knots = {[0 1],[0 1],[0 1]};
coefs = ones([1,1,1]);

However, this coefs array would be truncated to 2-D because MATLAB suppresses all trailing singleton dimensions (dimensions greater than 2 whose size is 1). Thus in earlier versions of the toolbox, this would fail.

spmak(knots, coefs)
??? Error using ==> spmak
coefs must be a ([1+]length(KNOTS))-dimensional array

Even if you managed to construct the B-form of this constant function, the various fn... functions could not work with it.

For the Spline Toolbox 3.0, all these fn... functions now handle splines with coefficient arrays of this kind correctly. You can now specify the intended size of the coefficient array by

```
spmak(knots,coefs,sizec) 
```

where sizec has the intended dimensionality of coefs. Now

spmak({[0 1],[0 1],[0 1]},ones([1,1,1]),[1 1 1])

will correctly construct a one-piece trivariate constant function, as will this.

```
spmak({[0 1],[0 1],[0 1]},1,[1 1 1]
```

Analogously, you can now use the optional third input argument, d, in

```
ppmak(breaks, coefs, d) 
```

to specify the intended dimensions of the coefficient array.

Spline Toolbox 3.0 Release Notes Upgrading from an Earlier Release