什么是计算笛卡尔积的好的非递归算法?

时间:2008-10-19 02:31:07

标签: algorithm language-agnostic rebol cartesian-product

注意

这不是特定于REBOL的问题。你可以用任何语言回答它。

背景

REBOL语言支持在REBOL parlance 中创建称为“方言”的特定于域的语言。我已经为列表推导创建了这样的方言,这在REBOL中本身不受支持。

列表推导需要一个好的笛卡尔积算法。

问题

我已经使用元编程来解决这个问题,通过动态创建然后执行一系列嵌套的foreach语句。它工作得很漂亮。但是,因为它是动态的,所以代码不是很易读。 REBOL不能很好地进行递归。它迅速耗尽堆栈空间和崩溃。因此,递归解决方案是不可能的。

总之,如果可能的话,我想用可读的,非递归的“内联”算法替换我的元编程。解决方案可以使用任何语言,只要我可以在REBOL中重现它。 (我可以阅读任何编程语言:C#,C,C ++,Perl,Oz,Haskell,Erlang,等等。)

我应该强调,这个算法需要支持任意数量的集合才能“加入”,因为列表理解可能涉及任意数量的集合。

6 个答案:

答案 0 :(得分:6)

使用速度提高3倍,使用的内存减少(回收次数减少)。

cartesian: func [
 d [block! ] 
 /local len set i res

][
 d: copy d
 len: 1
 res: make block! foreach d d [len: len * length? d]
 len: length? d
 until [
  set: clear []
  loop i: len [insert set d/:i/1   i: i - 1]
  res: change/only res copy set
  loop i: len [
   unless tail? d/:i: next d/:i [break]
   if i = 1 [break]
   d/:i: head d/:i
   i: i - 1
  ]
  tail? d/1
 ]
 head res
]

答案 1 :(得分:5)

这样的事情怎么样:

#!/usr/bin/perl

use strict;
use warnings;

my @list1 = qw(1 2);
my @list2 = qw(3 4);
my @list3 = qw(5 6);

# Calculate the Cartesian Product
my @cp = cart_prod(\@list1, \@list2, \@list3);

# Print the result
foreach my $elem (@cp) {
  print join(' ', @$elem), "\n";
}

sub cart_prod {
  my @sets = @_;
  my @result;
  my $result_elems = 1;

  # Calculate the number of elements needed in the result
  map { $result_elems *= scalar @$_ } @sets;
  return undef if $result_elems == 0;

  # Go through each set and add the appropriate element
  # to each element of the result
  my $scale_factor = $result_elems;
  foreach my $set (@sets)
  {
    my $set_elems = scalar @$set;  # Elements in this set
    $scale_factor /= $set_elems;
    foreach my $i (0 .. $result_elems - 1) {
      # Calculate the set element to place in this position
      # of the result set.
      my $pos = $i / $scale_factor % $set_elems;
      push @{$result[$i]}, $$set[ $pos ];
    }
  }

  return @result;
}

产生以下输出:

1 3 5
1 3 6
1 4 5
1 4 6
2 3 5
2 3 6
2 4 5
2 4 6

答案 2 :(得分:3)

为了完整起见,这里罗伯特·甘博的答案被翻译成REBOL:

REBOL []

cartesian: func [
    {Given a block of sets, returns the Cartesian product of said sets.}
    sets [block!] {A block containing one or more series! values}
    /local
        elems
        result
        row
][
    result: copy []

    elems: 1
    foreach set sets [
        elems: elems * (length? set)
    ]

    for n 0 (elems - 1) 1 [
        row: copy []
        skip: elems
        foreach set sets [
            skip: skip / length? set
            index: (mod to-integer (n / skip) length? set) + 1 ; REBOL is 1-based, not 0-based
            append row set/(index)
        ]
        append/only result row
    ]

    result
]

foreach set cartesian [[1 2] [3 4] [5 6]] [
    print set
]

; This returns the same thing Robert Gamble's solution did:

1 3 5
1 3 6
1 4 5
1 4 6
2 3 5
2 3 6
2 4 5
2 4 6

答案 3 :(得分:1)

use strict;

print "@$_\n" for getCartesian(
  [qw(1 2)],
  [qw(3 4)],
  [qw(5 6)],
);

sub getCartesian {
#
  my @input = @_;
  my @ret = map [$_], @{ shift @input };

  for my $a2 (@input) {
    @ret = map {
      my $v = $_;
      map [@$v, $_], @$a2;
    }
    @ret;
  }
  return @ret;
}

输出

1 3 5
1 3 6
1 4 5
1 4 6
2 3 5
2 3 6
2 4 5
2 4 6

答案 4 :(得分:1)

这是一个Java代码,用于为具有任意数量元素的任意数量的集合生成笛卡尔积。

在此示例中,列表“ls”包含4个集合(ls1,ls2,ls3和ls4),因为您可以看到“ls”可以包含任意数量的具有任意数量元素的集合。

import java.util.*;

public class CartesianProduct {

    private List <List <String>> ls = new ArrayList <List <String>> ();
    private List <String> ls1 = new ArrayList <String> ();
    private List <String> ls2 = new ArrayList <String> ();
    private List <String> ls3 = new ArrayList <String> ();
    private List <String> ls4 = new ArrayList <String> ();

    public List <String> generateCartesianProduct () {
        List <String> set1 = null;
        List <String> set2 = null;

        ls1.add ("a");
        ls1.add ("b");
        ls1.add ("c");

        ls2.add ("a2");
        ls2.add ("b2");
        ls2.add ("c2");

        ls3.add ("a3");
        ls3.add ("b3");
        ls3.add ("c3");
        ls3.add ("d3");

        ls4.add ("a4");
        ls4.add ("b4");

        ls.add (ls1);
        ls.add (ls2);
        ls.add (ls3);
        ls.add (ls4);

        boolean subsetAvailabe = true;
        int setCount = 0;

        try{    
            set1 = augmentSet (ls.get (setCount++), ls.get (setCount));
        } catch (IndexOutOfBoundsException ex) {
            if (set1 == null) {
                set1 = ls.get(0);
            }
            return set1;
        }

        do {
            try {
                setCount++;      
                set1 = augmentSet(set1,ls.get(setCount));
            } catch (IndexOutOfBoundsException ex) {
                subsetAvailabe = false;
            }
        } while (subsetAvailabe);
        return set1;
    }

    public List <String> augmentSet (List <String> set1, List <String> set2) {

        List <String> augmentedSet = new ArrayList <String> (set1.size () * set2.size ());
        for (String elem1 : set1) {
            for(String elem2 : set2) {
                augmentedSet.add (elem1 + "," + elem2);
            }
        }
        set1 = null; set2 = null;
        return augmentedSet;
    }

    public static void main (String [] arg) {
        CartesianProduct cp = new CartesianProduct ();
        List<String> cartesionProduct = cp.generateCartesianProduct ();
        for (String val : cartesionProduct) {
            System.out.println (val);
        }
    }
}

答案 5 :(得分:0)

编辑:此解决方案不起作用。 Robert Gamble是正确的解决方案。

我集思广益,想出了这个解决方案:

(我知道大多数人都不会知道REBOL,但它是一种相当可读的语言。)

REBOL []

sets: [[1 2 3] [4 5] [6]] ; Here's a set of sets
elems: 1
result: copy []
foreach set sets [elems: elems * (length? set)]
for n 1 elems 1 [
    row: copy []
    foreach set sets [
        index: 1 + (mod (n - 1) length? set)
        append row set/(index)
    ]
    append/only result row
]

foreach row result [
    print result
]

此代码生成:

1 4 6
2 5 6
3 4 6
1 5 6
2 4 6
3 5 6

(首先阅读上面的数字,你可能认为有重复数据。我做过。但是没有。)

有趣的是,这段代码几乎使用了相同的算法(1 +((n - 1)%9)来破坏我的Digital Root问题。