view scripts/set/intersect.m @ 31120:4581402b1c5b

uniquetol.m: Simplify code for "ByRows" option (bug #59850). Use standard form for error() messages and update input validation BIST tests to pass. * uniquetol.m: Use "real" instead of "non-complex" in documentation and in error() messages. Update input validation BIST tests to pass with new messages. Update documentation example so that text matches actual output of Octave. Use isreal() rather than iscomplex() in input validation. New variable "calc_indices" which indicates whether outputs ia, ic should be calculated. Use "calc_indices" to reduce running unnecessary code. In ByRows code, eliminate Iall variable and use J for the same purpose. Eliminate linear search ("any (Iall == i)") with direct lookup ("if (J(i))"). Eliminate variables sumeq, ii. Introduce intermediate variable "Arow_i" for clarity. Rename "equ" to "eq_rows" for clarity. Use '!' instead of '~' for logical negation.
author Rik <rik@octave.org>
date Tue, 05 Jul 2022 17:14:44 -0700
parents 796f54d4ddbf
children 597f3ee61a48
line wrap: on
line source

########################################################################
##
## Copyright (C) 2000-2022 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{c} =} intersect (@var{a}, @var{b})
## @deftypefnx {} {@var{c} =} intersect (@var{a}, @var{b}, "rows")
## @deftypefnx {} {@var{c} =} intersect (@dots{}, "sorted")
## @deftypefnx {} {@var{c} =} intersect (@dots{}, "stable")
## @deftypefnx {} {@var{c} =} intersect (@dots{}, "legacy")
## @deftypefnx {} {[@var{c}, @var{ia}, @var{ib}] =} intersect (@dots{})
##
## Return the unique elements common to both @var{a} and @var{b}.
##
## If @var{a} and @var{b} are both row vectors then return a row vector;
## Otherwise, return a column vector.  The inputs may also be cell arrays of
## strings.
##
## If the optional input @qcode{"rows"} is given then return the common rows of
## @var{a} and @var{b}.  The inputs must be 2-D numeric matrices to use this
## option.
##
## The optional argument @qcode{"sorted"}/@qcode{"stable"} controls the order
## in which unique values appear in the output.  The default is
## @qcode{"sorted"} and values in the output are placed in ascending order.
## The alternative @qcode{"stable"} preserves the order found in the input.
##
## If requested, return column index vectors @var{ia} and @var{ib} such that
## @code{@var{c} = @var{a}(@var{ia})} and @code{@var{c} = @var{b}(@var{ib})}.
##
## Programming Note: The input flag @qcode{"legacy"} changes the algorithm
## to be compatible with @sc{matlab} releases prior to R2012b.
##
## @seealso{unique, union, setdiff, setxor, ismember}
## @end deftypefn

function [c, ia, ib] = intersect (a, b, varargin)

  if (nargin < 2 || nargin > 4)
    print_usage ();
  endif

  [a, b] = validsetargs ("intersect", a, b, varargin{:});

  ## Special case of empty matrices
  if (isempty (a) || isempty (b))
    ## Lots of type checking required for Matlab compatibility.
    if (isnumeric (a) && isnumeric (b))
      c = [];
    elseif (iscell (b))
      c = {};
    else
      c = "";
    endif
    ia = ib = [];
    return;
  endif

  by_rows = any (strcmp ("rows", varargin));
  optsorted = ! any (strcmp ("stable", varargin));
  optlegacy = any (strcmp ("legacy", varargin));

  if (optlegacy)
    isrowvec = ! iscolumn (a) || ! iscolumn (b);
  else
    isrowvec = isrow (a) && isrow (b);
  endif

  ## Form A and B into sets
  if (nargout > 1 || ! optsorted)
    [a, ia] = unique (a, varargin{:});
    ia = ia(:);
    [b, ib] = unique (b, varargin{:});
    ib = ib(:);
  else
    a = unique (a, varargin{:});
    b = unique (b, varargin{:});
  endif

  if (by_rows)
    c = [a; b];
    if (nargout > 1 || ! optsorted)
      [c, ic] = sortrows (c);
    else
      c = sortrows (c);
    endif
    match = find (all (c(1:end-1,:) == c(2:end,:), 2));
    if (optsorted)
      c = c(match, :);
    else
      c = [a; b];
      ## FIXME: Is there a way to avoid a call to sort?
      c = c(sort (ic(match)), :);
    endif
    len_a = rows (a);
  else
    c = [a(:); b(:)];
    if (nargout > 1 || ! optsorted)
      [c, ic] = sort (c);
    else
      c = sort (c);
    endif
    if (iscellstr (c))
      match = find (strcmp (c(1:end-1), c(2:end)));
    else
      match = find (c(1:end-1) == c(2:end));
    endif
    len_a = length (a);
    if (optsorted)
      c = c(match);
    else
      c = [a(:); b(:)];
      ## FIXME: Is there a way to avoid a call to sort?
      c = c(sort (ic(match)));
    endif

    ## Adjust output orientation for Matlab compatibility
    if (isrowvec)
      c = c.';
    endif
  endif

  if (nargout > 1)
    ia = ia(ic(match));            # a(ia) == c
    ib = ib(ic(match+1) - len_a);  # b(ib) == c
    if (! optsorted)
      ## FIXME: Is there a way to avoid a call to sort?
      ia = sort (ia);
      [~, idx] = min (ib);
      ib = [ib(idx:end); ib(1:idx-1)];
    endif
    if (optlegacy && isrowvec && ! by_rows)
      ia = ia.';
      ib = ib.';
    endif
  endif

endfunction


%!assert (intersect ([1 2 3 4], [9 8 4 2]), [2, 4])
%!assert (intersect ([1 2; 2 3; 4 5], [2 3; 3 4; 5 6], "rows"), [2 3])
%!assert (intersect ([1 NaN], [NaN NaN 5]), zeros (1,0))

%!test
%! a = [1 1 1 2 2 2];
%! b = [1 2 3 4 5 6];
%! c = intersect (a, b);
%! assert (c, [1,2]);

## Test multi-dimensional arrays
%!test
%! a = rand (3,3,3);
%! b = a;
%! b(1,1,1) = 2;
%! assert (intersect (a, b), sort (a(2:end)'));

## Test the routine for index vectors ia and ib
%!test
%! a = [3 2 4 5 7 6 5 1 0 13 13];
%! b = [3 5 12 1 1 7];
%! [c, ia, ib] = intersect (a, b);
%! assert (c, [1, 3, 5, 7]);
%! assert (ia, [8; 1; 4; 5]);
%! assert (ib, [4; 1; 2; 6]);
%! assert (a(ia), c);
%! assert (b(ib), c);

## Test "rows" argument
%!test
%! a = [1,1,2;1,4,5;2,1,7];
%! b = [1,4,5;2,3,4;1,1,2;9,8,7];
%! [c,ia,ib] = intersect (a, b, "rows");
%! assert (c, [1,1,2;1,4,5]);
%! assert (ia, [1;2]);
%! assert (ib, [3;1]);
%! assert (a(ia,:), c);
%! assert (b(ib,:), c);

%!test
%! a = [1 2 3 4; 5 6 7 8; 9 10 11 12];
%! [b, ia, ib] = intersect (a, a, "rows");
%! assert (b, a);
%! assert (ia, [1:3]');
%! assert (ib, [1:3]');

## Test "stable" argument
%!test
%! a = [2 2 2 1 1 1];
%! b = [1 2 3 4 5 6];
%! c = intersect (a, b, "stable");
%! assert (c, [2,1]);

%!test
%! a = [3 2 4 5 7 6 5 1 0 13 13];
%! b = [3 5 12 1 1 7];
%! [c, ia, ib] = intersect (a, b, "stable");
%! assert (c, [3, 5, 7, 1]);
%! assert (ia, [1; 4; 5; 8]);
%! assert (ib, [1; 2; 6; 4]);
%! assert (a(ia), c);
%! assert (b(ib), c);

%!test
%! a = [1,4,5;1,1,2;2,1,7];
%! b = [1,4,5;2,3,4;1,1,2;9,8,7];
%! [c, ia, ib] = intersect (a, b, "rows", "stable");
%! assert (c, [1,4,5; 1,1,2]);
%! assert (ia, [1;2]);
%! assert (ib, [1;3]);
%! assert (a(ia,:), c);
%! assert (b(ib,:), c);

%!test
%! a = [1 2 3 4; 5 6 7 8; 9 10 11 12];
%! [b, ia, ib] = intersect (a, a, "rows", "stable");
%! assert (b, a);
%! assert (ia, [1:3]');
%! assert (ib, [1:3]');

## Test "legacy" argument
%!test
%! a = [7 1 7 7 4];
%! b = [7 0 4 4 0];
%! [c, ia, ib] = intersect (a, b);
%! assert (c, [4, 7]);
%! assert (ia, [5; 1]);
%! assert (ib, [3; 1]);
%! [c, ia, ib] = intersect (a, b, "legacy");
%! assert (c, [4, 7]);
%! assert (ia, [5, 4]);
%! assert (ib, [4, 1]);

%!test  # "legacy" + "rows"
%! A = [ 1 2; 3 4; 5 6; 3 4; 7 8 ];
%! B = [ 3 4; 7 8; 9 10 ];
%! [c, ia, ib] = intersect (A, B, "rows");
%! assert (c, [3, 4; 7, 8]);
%! assert (ia, [2; 5]);
%! assert (ib, [1; 2]);
%! [c, ia, ib] = intersect (A, B, "rows", "legacy");
%! assert (c, [3, 4; 7, 8]);
%! assert (ia, [4; 5]);
%! assert (ib, [1; 2]);

## Test orientation of output
%!shared a,b
%! a = 1:4;
%! b = 2:5;

%!assert (size (intersect (a, b)), [1, 3])
%!assert (size (intersect (a', b)), [3, 1])
%!assert (size (intersect (a, b')), [3, 1])
%!assert (size (intersect (a', b')), [3, 1])
%!assert (size (intersect (a, b, "legacy")), [1, 3])
%!assert (size (intersect (a', b, "legacy")), [1, 3])
%!assert (size (intersect (a, b', "legacy")), [1, 3])
%!assert (size (intersect (a', b', "legacy")), [3, 1])

## Test return type of empty intersections
%!assert (intersect (['a', 'b'], {}), {})
%!assert (intersect ([], {'a', 'b'}), {})
%!assert (intersect ([], {}), {})
%!assert (intersect ({'a', 'b'}, []), {})
%!assert (intersect ([], ['a', 'b']), "")
%!assert (intersect ({}, []), {})
%!assert (intersect (['a', 'b'], []), "")