Here is a session in Octave demonstrating how to compute the QR decomposition of a matrix:

octave-3.2.4.exe:1> m = [1, 2; 3, 4]
m =

   1   2
   3   4

octave-3.2.4.exe:2> [Q R] = qr(m)
Q =

  -0.31623  -0.94868
  -0.94868   0.31623

R =

  -3.16228  -4.42719
   0.00000  -0.63246

To emulate this functional approach (versus object-oriented approach) to decomposing a matrix in PHP, we need to wrap calls to the JAMA Matrix.php object inside a qr( ) function as follows:

<?php

// Set path to JAMA Matrix object.
require_once "../Matrix.php";

// Wrap JAMA matrix functionality inside a qr function.
function qr($m) {
  $A  = new Matrix($m);
  $QR = $A->qr();
  $ans['R']  = $QR->getR();
  $ans['Q']  = $QR->getQ();
  return $ans;
}

// Create matrix.
$m = array(array(1, 2),array(3, 4));

// Qet QR decompostion of a matrix.
$ans = qr($m);

// Output Q and R components of answer object.
?>

<pre>
Q = 
<?php print_r($ans['Q']); ?>
</pre>

<pre>
R = 
<?php print_r($ans['R']); ?>
</pre>



The output of this script looks like this:

Q = 
Matrix Object
(
    [A] => Array
        (
            [0] => Array
                (
                    [1] => 0.94868329805051
                    [0] => -0.31622776601684
                )

            [1] => Array
                (
                    [1] => -0.31622776601684
                    [0] => -0.94868329805051
                )

        )

    [m] => 2
    [n] => 2
)

R = 
Matrix Object
(
    [A] => Array
        (
            [0] => Array
                (
                    [0] => -3.1622776601684
                    [1] => -4.4271887242357
                )

            [1] => Array
                (
                    [0] => 0
                    [1] => 0.63245553203368
                )

        )

    [m] => 2
    [n] => 2
)

Interestingly, the answers produced by the Octave and PHP scripts differ in the signs associated with a few of the answer values. Does this matter? Well, if we take the values of Q and R output by PHP, enter them into Octave, and then multiply Q and R together, we get the original matrix back. This reconstruction of the original matrix demonstrates the correctness of the decomposition and that the decomposition of a matrix into Q and R components is not unique.

octave-3.2.4.exe:1> Q = [-0.31622, 0.94868; -0.94868, -0.31622]
Q =

  -0.31622   0.94868
  -0.94868  -0.31622

octave-3.2.4.exe:2> R = [-3.1622, -4.42718; 0, 0.63245]
R =

  -3.16220  -4.42718
   0.00000   0.63245

octave-3.2.4.exe:3> Q * R
ans =

   0.99995   1.99996
   2.99992   3.99998

Note that if I didn't reduce the precision of the Q and R matrix values, the output would be [1 2; 3 4] or the exact version of the original matrix.

This proof-of-concept demonstrates how we can develop a matlab/octave-like functional API for linear algrebra decompositions in PHP by incorporating calls to the JAMA library methods inside qr(), svd(), lu(), eig(), and chol() functions. A functional API for linear algebra would make doing linear algrebra in PHP much more direct, interactive and enjoyable. Object-oriented linear algebra libraries like JAMA play a supporting role, hidden behind a set of linear algebra functions like qr(), svd(), lu(), eig() and chol() for operating on matrices.