Add randn to create an array of standard gaussian numbers.

Cargo.toml

@@ -36,7 +36,8 @@ optional = true
 
 [dependencies]
 blas-sys = { version = "0.5", optional = true, default-features = false }
-
+asprim = "0.1"
+rand = "0.3"
 # deprecated! use ndarray-rblas instead
 rblas = { version = "0.0.13", optional = true }
 

src/impl_constructors.rs

@@ -8,6 +8,12 @@
 
 //! Constructor methods for ndarray
 //!
+extern crate rand;
+use self::rand::distributions::{Normal, IndependentSample};
+use self::rand::{Rng};
+
+extern crate asprim;
+use self::asprim::AsPrim;
 use libnum;
 
 use imp_prelude::*;
@@ -252,3 +258,28 @@ impl<S, A, D> ArrayBase<S, D>
     }
 
 }
+
+use super::NdFloat;
+
+impl<S, A, D> ArrayBase<S, D>
+    where S: DataOwned<Elem=A>,
+          A: NdFloat+AsPrim,
+          D: Dimension,
+{
+    pub fn randn_rng<R: Rng>(dim: D, rng: &mut R) -> ArrayBase<S, D>
+    {
+        let size = dim.size_checked().expect("Shape too large: overflow in size");
+        let normal = Normal::new(1.0, 0.0);
+        let iter = (0..size).map(|_| normal.ind_sample(rng).as_());
+        let arr = ArrayBase::from_iter(iter);
+        arr.into_shape(dim).unwrap()
+    }
+
+    pub fn randn(dim: D) -> ArrayBase<S, D>
+    {
+        // Use the thread_rng once to seed a fast, non-crypto grade rng.
+        // To use a crypto quality RNG, use randn_rng directly.
+        let mut rng = rand::weak_rng();
+        ArrayBase::<S, D>::randn_rng(dim, &mut rng)
+    }
+}