Merge pull request #34 from termoshtt/solve

LU decomposition and inverse matrix

Author: termoshtt

src/impl2/mod.rs

@@ -2,15 +2,17 @@
 pub mod opnorm;
 pub mod qr;
 pub mod svd;
+pub mod solve;
 
 pub use self::opnorm::*;
 pub use self::qr::*;
 pub use self::svd::*;
+pub use self::solve::*;
 
 use super::error::*;
 
-pub trait LapackScalar: OperatorNorm_ + QR_ + SVD_ {}
-impl<A> LapackScalar for A where A: OperatorNorm_ + QR_ + SVD_ {}
+pub trait LapackScalar: OperatorNorm_ + QR_ + SVD_ + Solve_ {}
+impl<A> LapackScalar for A where A: OperatorNorm_ + QR_ + SVD_ + Solve_ {}
 
 pub fn into_result<T>(info: i32, val: T) -> Result<T> {
     if info == 0 {

src/impl2/solve.rs

@@ -0,0 +1,58 @@
+
+use lapack::c;
+
+use types::*;
+use error::*;
+use layout::Layout;
+
+use super::into_result;
+
+pub type Pivot = Vec<i32>;
+
+#[derive(Debug, Clone, Copy)]
+#[repr(u8)]
+pub enum Transpose {
+    No = b'N',
+    Transpose = b'T',
+    Hermite = b'C',
+}
+
+pub trait Solve_: Sized {
+    fn lu(Layout, a: &mut [Self]) -> Result<Pivot>;
+    fn inv(Layout, a: &mut [Self], &Pivot) -> Result<()>;
+    fn solve(Layout, Transpose, a: &[Self], &Pivot, b: &mut [Self]) -> Result<()>;
+}
+
+macro_rules! impl_solve {
+    ($scalar:ty, $getrf:path, $getri:path, $getrs:path) => {
+
+impl Solve_ for $scalar {
+    fn lu(l: Layout, a: &mut [Self]) -> Result<Pivot> {
+        let (row, col) = l.size();
+        let k = ::std::cmp::min(row, col);
+        let mut ipiv = vec![0; k as usize];
+        let info = $getrf(l.lapacke_layout(), row, col, a, l.lda(), &mut ipiv);
+        into_result(info, ipiv)
+    }
+
+    fn inv(l: Layout, a: &mut [Self], ipiv: &Pivot) -> Result<()> {
+        let (n, _) = l.size();
+        let info = $getri(l.lapacke_layout(), n, a, l.lda(), ipiv);
+        into_result(info, ())
+    }
+
+    fn solve(l: Layout, t: Transpose, a: &[Self], ipiv: &Pivot, b: &mut [Self]) -> Result<()> {
+        let (n, _) = l.size();
+        let nrhs = 1;
+        let ldb = 1;
+        let info = $getrs(l.lapacke_layout(), t as u8, n, nrhs, a, l.lda(), ipiv, b, ldb);
+        into_result(info, ())
+    }
+}
+
+}} // impl_solve!
+
+impl_solve!(f64, c::dgetrf, c::dgetri, c::dgetrs);
+impl_solve!(f32, c::sgetrf, c::sgetri, c::sgetrs);
+impl_solve!(c64, c::zgetrf, c::zgetri, c::zgetrs);
+impl_solve!(c32, c::cgetrf, c::cgetri, c::cgetrs);

src/lib.rs

@@ -51,6 +51,7 @@ pub mod impl2;
 pub mod qr;
 pub mod svd;
 pub mod opnorm;
+pub mod solve;
 
 pub mod vector;
 pub mod matrix;

src/prelude.rs

@@ -9,3 +9,4 @@ pub use assert::*;
 pub use qr::*;
 pub use svd::*;
 pub use opnorm::*;
+pub use solve::*;

src/solve.rs

@@ -0,0 +1,92 @@
+
+use ndarray::*;
+use super::layout::*;
+use super::error::*;
+use super::impl2::*;
+
+pub use impl2::{Pivot, Transpose};
+
+pub struct Factorized<S: Data> {
+    pub a: ArrayBase<S, Ix2>,
+    pub ipiv: Pivot,
+}
+
+impl<A, S> Factorized<S>
+    where A: LapackScalar,
+          S: Data<Elem = A>
+{
+    pub fn solve<Sb>(&self, t: Transpose, mut rhs: ArrayBase<Sb, Ix1>) -> Result<ArrayBase<Sb, Ix1>>
+        where Sb: DataMut<Elem = A>
+    {
+        A::solve(self.a.square_layout()?,
+                 t,
+                 self.a.as_allocated()?,
+                 &self.ipiv,
+                 rhs.as_slice_mut().unwrap())?;
+        Ok(rhs)
+    }
+}
+
+impl<A, S> Factorized<S>
+    where A: LapackScalar,
+          S: DataMut<Elem = A>
+{
+    pub fn into_inverse(mut self) -> Result<ArrayBase<S, Ix2>> {
+        A::inv(self.a.square_layout()?,
+               self.a.as_allocated_mut()?,
+               &self.ipiv)?;
+        Ok(self.a)
+    }
+}
+
+pub trait Factorize<S: Data> {
+    fn factorize(self) -> Result<Factorized<S>>;
+}
+
+impl<A, S> Factorize<S> for ArrayBase<S, Ix2>
+    where A: LapackScalar,
+          S: DataMut<Elem = A>
+{
+    fn factorize(mut self) -> Result<Factorized<S>> {
+        let ipiv = A::lu(self.layout()?, self.as_allocated_mut()?)?;
+        Ok(Factorized {
+            a: self,
+            ipiv: ipiv,
+        })
+    }
+}
+
+impl<'a, A, S> Factorize<OwnedRepr<A>> for &'a ArrayBase<S, Ix2>
+    where A: LapackScalar + Clone,
+          S: Data<Elem = A>
+{
+    fn factorize(self) -> Result<Factorized<OwnedRepr<A>>> {
+        let mut a = self.to_owned();
+        let ipiv = A::lu(a.layout()?, a.as_allocated_mut()?)?;
+        Ok(Factorized { a: a, ipiv: ipiv })
+    }
+}
+
+pub trait Inverse<Inv> {
+    fn inv(self) -> Result<Inv>;
+}
+
+impl<A, S> Inverse<ArrayBase<S, Ix2>> for ArrayBase<S, Ix2>
+    where A: LapackScalar,
+          S: DataMut<Elem = A>
+{
+    fn inv(self) -> Result<ArrayBase<S, Ix2>> {
+        let f = self.factorize()?;
+        f.into_inverse()
+    }
+}
+
+impl<'a, A, S> Inverse<Array2<A>> for &'a ArrayBase<S, Ix2>
+    where A: LapackScalar + Clone,
+          S: Data<Elem = A>
+{
+    fn inv(self) -> Result<Array2<A>> {
+        let f = self.factorize()?;
+        f.into_inverse()
+    }
+}

src/square.rs

@@ -1,21 +1,16 @@
 //! Define trait for Hermite matrices
 
 use ndarray::{Ix2, Array, RcArray, ArrayBase, Data};
-use lapack::c::Layout;
 
 use super::matrix::{Matrix, MFloat};
 use super::error::{LinalgError, NotSquareError};
-use super::impls::solve::ImplSolve;
 
 /// Methods for square matrices
 ///
 /// This trait defines method for square matrices,
 /// but does not assure that the matrix is square.
 /// If not square, `NotSquareError` will be thrown.
 pub trait SquareMatrix: Matrix {
-    // fn eig(self) -> (Self::Vector, Self);
-    /// inverse matrix
-    fn inv(self) -> Result<Self, LinalgError>;
     /// trace of matrix
     fn trace(&self) -> Result<Self::Scalar, LinalgError>;
     #[doc(hidden)]
@@ -46,30 +41,13 @@ fn trace<A: MFloat, S>(a: &ArrayBase<S, Ix2>) -> A
 }
 
 impl<A: MFloat> SquareMatrix for Array<A, Ix2> {
-    fn inv(self) -> Result<Self, LinalgError> {
-        self.check_square()?;
-        let (n, _) = self.size();
-        let layout = self.layout()?;
-        let (ipiv, a) = ImplSolve::lu(layout, n, n, self.into_raw_vec())?;
-        let a = ImplSolve::inv(layout, n, a, &ipiv)?;
-        let m = Array::from_vec(a).into_shape((n, n)).unwrap();
-        match layout {
-            Layout::RowMajor => Ok(m),
-            Layout::ColumnMajor => Ok(m.reversed_axes()),
-        }
-    }
     fn trace(&self) -> Result<Self::Scalar, LinalgError> {
         self.check_square()?;
         Ok(trace(self))
     }
 }
 
 impl<A: MFloat> SquareMatrix for RcArray<A, Ix2> {
-    fn inv(self) -> Result<Self, LinalgError> {
-        // XXX unnecessary clone (should use into_owned())
-        let i = self.to_owned().inv()?;
-        Ok(i.into_shared())
-    }
     fn trace(&self) -> Result<Self::Scalar, LinalgError> {
         self.check_square()?;
         Ok(trace(self))