view scripts/general/quadv.m @ 31058:12f8fb75fc30

quadgk.m: Overhaul function to add new "ArrayValued" option (bug #62468) * quadgk.m: Add "ArrayValued" option to docstring. Re-phrase bits of existing documentation for clarity. Rename input validation variable "str" to "prop" for clarity. Add decode for "arrayvalued" property. If limits of integration are high to low then reverse them in function rather than recursively calling quadgk(). Add FIXME note about missing input validation for several properties. Delete variable 'h0' which was never used. New code branch if "ArrayValued" is true which uses same code strategy as for normal quadgk, but subfunction __quadgk_eval_array__ for evaluating the function in a vectorized manner. Add BIST tests for "ArrayValued" input. * quadgk.m (__quadgk_eval__): Eliminate "too_close" output and code to calculate "too_close" which was never used. * quadgk.m (__quadgk_eval_array__): New subfunction.
author Michael Leitner <michael.leitner@frm2.tum.de>, Rik <rik@octave.org>
date Thu, 02 Jun 2022 08:45:10 -0700
parents e1788b1a315f
children 597f3ee61a48
line wrap: on
line source

########################################################################
##
## Copyright (C) 2008-2022 The Octave Project Developers
##
## See the file COPYRIGHT.md in the top-level directory of this
## distribution or <https://octave.org/copyright/>.
##
## This file is part of Octave.
##
## Octave is free software: you can redistribute it and/or modify it
## under the terms of the GNU General Public License as published by
## the Free Software Foundation, either version 3 of the License, or
## (at your option) any later version.
##
## Octave is distributed in the hope that it will be useful, but
## WITHOUT ANY WARRANTY; without even the implied warranty of
## MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
## GNU General Public License for more details.
##
## You should have received a copy of the GNU General Public License
## along with Octave; see the file COPYING.  If not, see
## <https://www.gnu.org/licenses/>.
##
########################################################################

## -*- texinfo -*-
## @deftypefn  {} {@var{q} =} quadv (@var{f}, @var{a}, @var{b})
## @deftypefnx {} {@var{q} =} quadv (@var{f}, @var{a}, @var{b}, @var{tol})
## @deftypefnx {} {@var{q} =} quadv (@var{f}, @var{a}, @var{b}, @var{tol}, @var{trace})
## @deftypefnx {} {@var{q} =} quadv (@var{f}, @var{a}, @var{b}, @var{tol}, @var{trace}, @var{p1}, @var{p2}, @dots{})
## @deftypefnx {} {[@var{q}, @var{nfev}] =} quadv (@dots{})
##
## Numerically evaluate the integral of @var{f} from @var{a} to @var{b}
## using an adaptive Simpson's rule.
##
## @var{f} is a function handle, inline function, or string containing the name
## of the function to evaluate.  @code{quadv} is a vectorized version of
## @code{quad} and the function defined by @var{f} must accept a scalar or
## vector as input and return a scalar, vector, or array as output.
##
## @var{a} and @var{b} are the lower and upper limits of integration.  Both
## limits must be finite.
##
## The optional argument @var{tol} defines the absolute tolerance used to stop
## the adaptation procedure.  The default value is 1e-6.
##
## The algorithm used by @code{quadv} involves recursively subdividing the
## integration interval and applying Simpson's rule on each subinterval.
## If @var{trace} is true then after computing each of these partial
## integrals display: (1) the total number of function evaluations,
## (2) the left end of the subinterval, (3) the length of the subinterval,
## (4) the approximation of the integral over the subinterval.
##
## Additional arguments @var{p1}, etc., are passed directly to the function
## @var{f}.  To use default values for @var{tol} and @var{trace}, one may pass
## empty matrices ([]).
##
## The result of the integration is returned in @var{q}.
##
## The optional output @var{nfev} indicates the total number of function
## evaluations performed.
##
## Note: @code{quadv} is written in Octave's scripting language and can be
## used recursively in @code{dblquad} and @code{triplequad}, unlike the
## @code{quad} function.
## @seealso{quad, quadl, quadgk, quadcc, trapz, dblquad, triplequad, integral,
##          integral2, integral3}
## @end deftypefn

## Algorithm: See https://en.wikipedia.org/wiki/Adaptive_Simpson%27s_method
## for basic explanation.  See NOTEs and FIXME for Octave modifications.

function [q, nfev] = quadv (f, a, b, tol = [], trace = false, varargin)

  if (nargin < 3)
    print_usage ();
  endif

  if (isa (a, "single") || isa (b, "single"))
    eps = eps ("single");
  else
    eps = eps ("double");
  endif
  if (isempty (tol))
    tol = 1e-6;
  elseif (! isscalar (tol) || tol < 0)
    error ("quadv: TOL must be a scalar >=0");
  endif

  if (trace)
    ## Print column headers once above trace display.
    printf ("  nfev          a            (b - a)         q_interval\n");
  endif

  ## NOTE: Split the interval into 3 parts which avoids problems with periodic
  ## functions when a, b, and (a + b)/2 fall on boundaries such as 0, 2*pi.
  ## For compatibility with Matlab, split in to two equal size regions on the
  ## left and right, and one larger central region.
  alpha = 0.27158;   # factor for region 1 & region 3 size (~27%)
  b1 = a + alpha * (b - a);
  b2 = b - alpha * (b - a);
  c1 = (a + b1) / 2;
  c2 = (a + b)  / 2;
  c3 = (b2 + b) / 2;

  fa  = feval (f, a,  varargin{:});
  fc1 = feval (f, c1, varargin{:});
  fb1 = feval (f, b1, varargin{:});
  fc2 = feval (f, c2, varargin{:});
  fb2 = feval (f, b2, varargin{:});
  fc3 = feval (f, c3, varargin{:});
  fb  = feval (f, b,  varargin{:});
  nfev = 7;

  ## NOTE: If there are edge singularities, move edge point by eps*(b-a) as
  ## discussed in Shampine paper used to implement quadgk.
  if (any (isinf (fa(:))))
    fa = feval (f, a + eps * (b-a), varargin{:});
    nfev++;
  endif
  if (any (isinf (fb(:))))
    fb = feval (f, b - eps * (b-a), varargin{:});
    nfev++;
  endif

  ## Region 1
  h = (b1 - a);
  q1 = h / 6 * (fa + 4*fc1 + fb1);

  [q1, nfev, hmin1] = simpsonstp (f, a, b1, c1, fa, fb1, fc1, q1, tol,
                                  nfev, abs (h), trace, varargin{:});

  ## Region 2
  h = (b2 - b1);
  q2 = h / 6 * (fb1 + 4*fc2 + fb2);

  [q2, nfev, hmin2] = simpsonstp (f, b1, b2, c2, fb1, fb2, fc2, q2, tol,
                                  nfev, abs (h), trace, varargin{:});

  ## Region 3
  h = (b - b2);
  q3 = h / 6 * (fb2 + 4*fc3 + fb);

  [q3, nfev, hmin3] = simpsonstp (f, b2, b, c3, fb2, fb, fc3, q3, tol,
                                  nfev, abs (h), trace, varargin{:});

  ## Total integral over all 3 regions and verify results
  q = q1 + q2 + q3;

  hmin = min ([hmin1, hmin2, hmin3]);

  if (nfev > 10_000)
    warning ("quadv: maximum iteration count reached -- possible singular integral");
  elseif (any (! isfinite (q(:))))
    warning ("quadv: infinite or NaN function evaluations were returned");
  elseif (hmin < (b - a) * eps)
    warning ("quadv: minimum step size reached -- possible singular integral");
  endif

endfunction

function [q, nfev, hmin] = simpsonstp (f, a, b, c, fa, fb, fc, q0, tol,
                                       nfev, hmin, trace, varargin)

  if (nfev > 10_000)   # stop endless recursion
    q = q0;
    return;
  endif

  d = (a + c) / 2;
  e = (c + b) / 2;
  fd = feval (f, d, varargin{:});
  fe = feval (f, e, varargin{:});
  nfev += 2;
  q1 = (c - a) / 6 * (fa + 4*fd + fc);
  q2 = (b - c) / 6 * (fc + 4*fe + fb);
  q = q1 + q2;

  if (abs (a - c) < hmin)
    hmin = abs (a - c);
  endif

  delta = q - q0;   # error term between new estimate and old estimate

  if (trace)
    printf ("%5d   %#14.11g   %16.10e   %-16.11g\n",
            nfev,  a,         b-a,      q + delta/15);
  endif

  ## NOTE: Not vectorizing q-q0 in the norm provides a more rigid criterion
  ##       for matrix-valued functions.
  if (norm (delta, Inf) > 15*tol)
    ## FIXME: To keep sum of sub-interval integrands within overall tolerance
    ## each bisection interval should use tol/2.  However, Matlab does not
    ## do this, and it would also profoundly increase the number of function
    ## evaluations required.
    [q1, nfev, hmin] = simpsonstp (f, a, c, d, fa, fc, fd, q1, tol,
                                   nfev, hmin, trace, varargin{:});
    [q2, nfev, hmin] = simpsonstp (f, c, b, e, fc, fb, fe, q2, tol,
                                   nfev, hmin, trace, varargin{:});
    q = q1 + q2;
  else
    q += delta / 15;   # NOTE: Richardson extrapolation correction
  endif

endfunction


%!assert (quadv (@sin, 0, 2*pi), 0, 1e-6)
%!assert (quadv (@sin, 0, pi), 2, 1e-6)

## Test weak singularities at the edge
%!assert (quadv (@(x) 1 ./ sqrt (x), 0, 1), 2, 15*1e-6)

## Test vector-valued functions
%!assert (quadv (@(x) [(sin (x)), (sin (2 * x))], 0, pi), [2, 0], 1e-6)

## Test matrix-valued functions
%!assert (quadv (@(x) [ x,x,x; x,1./sqrt(x),x; x,x,x ], 0, 1),
%!        [0.5,0.5,0.5; 0.5,2,0.5; 0.5,0.5,0.5], 15*1e-6);

## Test periodic function
%!assert <*57603> (quadv (@(t) sin (t) .^ 2, 0, 8*pi), 4*pi, 1e-6)

## Test input validation
%!error <Invalid call> quadv ()
%!error <Invalid call> quadv (@sin)
%!error <Invalid call> quadv (@sin,1)
%!error <TOL must be a scalar> quadv (@sin,0,1, ones (2,2))
%!error <TOL must be .* .=0> quadv (@sin,0,1, -1)