MultivarGaussian can't take literal mu and sigma

[cberzan]

This works:

fixed RealMatrix mu = [0.0; 0.0];
fixed RealMatrix sigma = [1.0, 0.0; 0.0, 2.0];
random RealMatrix x ~ MultivarGaussian(mu, sigma);
query x;

But this doesn't work:

random RealMatrix x ~ MultivarGaussian([0.0; 0.0], [1.0, 0.0; 0.0, 2.0]);
query x;

Error:

java.lang.reflect.InvocationTargetException
    at sun.reflect.NativeConstructorAccessorImpl.newInstance0(Native Method)
    at sun.reflect.NativeConstructorAccessorImpl.newInstance(NativeConstructorAccessorImpl.java:57)
    at sun.reflect.DelegatingConstructorAccessorImpl.newInstance(DelegatingConstructorAccessorImpl.java:45)
    at java.lang.reflect.Constructor.newInstance(Constructor.java:526)
    at blog.model.Clause.initCPD(Clause.java:358)
    at blog.model.Clause.compile(Clause.java:323)
    at blog.model.DependencyModel.compile(DependencyModel.java:262)
    at blog.model.RandomFunction.compile(RandomFunction.java:183)
    at blog.model.Model.compile(Model.java:566)
    at blog.Main.setup(Main.java:607)
    at blog.Main.main(Main.java:171)
Caused by: java.lang.ClassCastException: java.util.ArrayList cannot be cast to blog.common.numerical.MatrixLib
    at blog.distrib.MultivarGaussian.<init>(MultivarGaussian.java:88)
    ... 11 more
Error initializing CPD at 1: java.lang.ClassCastException (java.util.ArrayList cannot be cast to blog.common.numerical.MatrixLib)
Encountered 1 errors in compilation phase.

[lileicc]

try

fixed RealMatrix sigma = [1.0, 0.0; 0.0, 2.0];
random RealMatrix x ~ MultivarGaussian(__SCALAR_STACK(0.0, 0.0), sigma);

[cberzan]

Discussed this with @lileicc; recording here for posterity:

The reason fixed RealMatrix mu = [0.0; 0.0]; works is because we special-case it in Semant (transferToMatrix converts a ListSpec to a MatrixLib). But there is no special-casing for MultivarGaussian([0.0; 0.0], ...), so the ListSpecs get passed to the MultivarGaussian, and things break.