[yt-dev] field interpolater question

[Matthew Turk]

Hi Britton,

On Tue, May 7, 2013 at 5:48 PM, Britton Smith <brittonsmith at gmail.com> wrote:
> Hi all,
>
> I'm working on some analysis that requires interpolating from some data
> tables.  Normally, I would use the Unilinear, Bilinear, and
> TrilinearFieldInterpolators in yt, but they only accept evenly spaced bins,
> which I don't have.  It actually looks like the cython function which does
> the calculation accepts the bins as arrays, which the python function
> creates from the given bounds.

That's right, the Python code assumes evenly spaced bins.  Reading
over the Cython, my initial thought was that it wouldn't allow for
this, but it totally does:

dx_inv = 1.0 / (x_bins[x_i+1] - x_bins[x_i])

So yeah, I agree, it can handle it.  As long as the bins are correctly
identified (which it does using digitize) it should be fine.

>
> I'd like to change this behavior to allow me to give the bins to the python
> function.  Currently, the call to create the interpolator looks like this:
> BilinearInterpolator(table, boundaries, field_names)
>
> where boundaries is (x_0, x_n, y_0, y_n) or the min/max of the x and y bins.
> Would it be ok if I changed this so that boundaries can also be a pair of
> arrays?  I would just have the function check and still maintain the old way
> as well.  By doing it this way, it shouldn't break backward compatibility
> either.  Would this be an ok solution?

Yup, this would be nice -- thanks.  Can you update the tests to
reflect this, too?  The interpolators are tested somewhat extensively
with random values.

-Matt

>
> Britton
>
> _______________________________________________
> yt-dev mailing list
> yt-dev at lists.spacepope.org
> http://lists.spacepope.org/listinfo.cgi/yt-dev-spacepope.org
>