On 16 Oct 2009, at 14:06, Konrad Hinsen wrote:
Thanks! This should at least be efficient. I quickly checked the
Colt implementation, which does have a constructor for its array
class that takes the plain Java array plus the shape information, so
at least in theory this should be simple. If/when I get it to work,
I'll post my solution here.
Here is a working example (Clojure source code) for inverting a matrix
read from a netCDF file using Colt:
(ns netcdf-read
(:import (ucar.nc2 NetcdfFile)
(cern.colt.matrix.tdouble.impl DenseDoubleMatrix2D)
(cern.colt.matrix.tdouble.algo DenseDoubleAlgebra)))
(defn unidata-to-colt
[#^ucar.ma2.ArrayDouble$D2 array]
(let [java-array-1d (.get1DJavaArray array Double)
[rows, cols] (.getShape array)]
(new DenseDoubleMatrix2D rows cols java-array-1d 0 0 cols 1
false)))
(let [matrix (with-open [ncfile (NetcdfFile/open "/Users/hinsen/
Temp/matrix.nc")]
(.read (.findVariable ncfile "matrix")))
colt-matrix (unidata-to-colt matrix)
la (new DenseDoubleAlgebra)
inverse (.inverse la colt-matrix)]
(prn inverse))
For my current needs, conversion of 2D double arrays is sufficient,
but this could of course be generalized to ranks 1 and 3 (that's all
there is in Colt) and other element types.
Konrad.
