JLink NaN Support
- To: mathgroup at smc.vnet.net
- Subject: [mg110752] JLink NaN Support
- From: Raffy <adraffy at gmail.com>
- Date: Sun, 4 Jul 2010 06:10:40 -0400 (EDT)
The support for NaN -> Indeterminate over JLink from Java ->
Mathematica is completely broken.
double[] f1() { return new double[]{Double.NaN}; }
double[] f2() { return new double[]{1/0D}; }
Evaluating f1 from Mathematica will return {-1024.}
Evaluating f2 from Mathematica will kill the kernel.
The desired behavior should be {Indeterminate} in both cases.
Are there any fixes for this or am I stuck mapping NaN's to a
different value?