Goto Chapter: Top 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 Bib Ind

16 Combinatorics

16.2 Combinations, Arrangements and Tuples

16.2-1 Combinations

16.2-2 Iterator and enumerator of combinations

16.2-3 NrCombinations

16.2-4 Arrangements

16.2-5 NrArrangements

16.2-6 UnorderedTuples

16.2-7 NrUnorderedTuples

16.2-8 Tuples

16.2-9 EnumeratorOfTuples

16.2-10 IteratorOfTuples

16.2-11 NrTuples

16.2-12 PermutationsList

16.2-13 NrPermutationsList

16.2-14 Derangements

16.2-15 NrDerangements

16.2-16 PartitionsSet

16.2-17 NrPartitionsSet

16.2-18 Partitions

16.2-19 IteratorOfPartitions

16.2-20 IteratorOfPartitionsSet

16.2-21 NrPartitions

16.2-22 OrderedPartitions

16.2-23 NrOrderedPartitions

16.2-24 PartitionsGreatestLE

16.2-25 PartitionsGreatestEQ

16.2-26 RestrictedPartitions

16.2-27 NrRestrictedPartitions

16.2-28 SignPartition

16.2-29 AssociatedPartition

16.2-30 PowerPartition

16.2-31 PartitionTuples

16.2-32 NrPartitionTuples

16.2-33 BetaSet

16.2-1 Combinations

16.2-2 Iterator and enumerator of combinations

16.2-3 NrCombinations

16.2-4 Arrangements

16.2-5 NrArrangements

16.2-6 UnorderedTuples

16.2-7 NrUnorderedTuples

16.2-8 Tuples

16.2-9 EnumeratorOfTuples

16.2-10 IteratorOfTuples

16.2-11 NrTuples

16.2-12 PermutationsList

16.2-13 NrPermutationsList

16.2-14 Derangements

16.2-15 NrDerangements

16.2-16 PartitionsSet

16.2-17 NrPartitionsSet

16.2-18 Partitions

16.2-19 IteratorOfPartitions

16.2-20 IteratorOfPartitionsSet

16.2-21 NrPartitions

16.2-22 OrderedPartitions

16.2-23 NrOrderedPartitions

16.2-24 PartitionsGreatestLE

16.2-25 PartitionsGreatestEQ

16.2-26 RestrictedPartitions

16.2-27 NrRestrictedPartitions

16.2-28 SignPartition

16.2-29 AssociatedPartition

16.2-30 PowerPartition

16.2-31 PartitionTuples

16.2-32 NrPartitionTuples

16.2-33 BetaSet

This chapter describes functions that deal with combinatorics. We mainly concentrate on two areas. One is about *selections*, that is the ways one can select elements from a set. The other is about *partitions*, that is the ways one can partition a set into the union of pairwise disjoint subsets.

`‣ Factorial` ( n ) | ( function ) |

returns the *factorial* \(n!\) of the positive integer `n`, which is defined as the product \(1 \cdot 2 \cdot 3 \cdots n\).

\(n!\) is the number of permutations of a set of \(n\) elements. \(1 / n!\) is the coefficient of \(x^n\) in the formal series \(\exp(x)\), which is the generating function for factorial.

gap> List( [0..10], Factorial ); [ 1, 1, 2, 6, 24, 120, 720, 5040, 40320, 362880, 3628800 ] gap> Factorial( 30 ); 265252859812191058636308480000000

`PermutationsList`

(16.2-12) computes the set of all permutations of a list.

`‣ Binomial` ( n, k ) | ( function ) |

returns the *binomial coefficient* \({{n \choose k}}\) of integers `n` and `k`. This is defined by the conditions \({{n \choose k}} = 0\) for \(k < 0\), \({{0 \choose k}} = 0\) for \(k \neq 0\), \({{0 \choose 0}} = 1\) and the relation \({{n \choose k}} = {{n-1 \choose k}} + {{n-1 \choose k-1}}\) for all \(n\) and \(k\).

There are many ways of describing this function. For example, if \(n \geq 0\) and \(0 \leq k \leq n\), then \({{n \choose k}} = n! / (k! (n-k)!)\) and for \(n < 0\) and \(k \geq 0\) we have \({{n \choose k}} = (-1)^k {{-n+k-1 \choose k}}\).

If \(n \geq 0\) then \({{n \choose k}}\) is the number of subsets with \(k\) elements of a set with \(n\) elements. Also, \({{n \choose k}}\) is the coefficient of \(x^k\) in the polynomial \((x + 1)^n\), which is the generating function for \({{n \choose .}}\), hence the name.

gap> # Knuth calls this the trademark of Binomial: gap> List( [0..4], k->Binomial( 4, k ) ); [ 1, 4, 6, 4, 1 ] gap> List( [0..6], n->List( [0..6], k->Binomial( n, k ) ) );; gap> # the lower triangle is called Pascal's triangle: gap> PrintArray( last ); [ [ 1, 0, 0, 0, 0, 0, 0 ], [ 1, 1, 0, 0, 0, 0, 0 ], [ 1, 2, 1, 0, 0, 0, 0 ], [ 1, 3, 3, 1, 0, 0, 0 ], [ 1, 4, 6, 4, 1, 0, 0 ], [ 1, 5, 10, 10, 5, 1, 0 ], [ 1, 6, 15, 20, 15, 6, 1 ] ] gap> Binomial( 50, 10 ); 10272278170

`NrCombinations`

(16.2-3) is the generalization of `Binomial`

for multisets. `Combinations`

(16.2-1) computes the set of all combinations of a multiset.

`‣ Bell` ( n ) | ( function ) |

returns the *Bell number* \(B(n)\). The Bell numbers are defined by \(B(0) = 1\) and the recurrence \(B(n+1) = \sum_{{k = 0}}^n {{n \choose k}} B(k)\).

\(B(n)\) is the number of ways to partition a set of `n` elements into pairwise disjoint nonempty subsets (see `PartitionsSet`

(16.2-16)). This implies of course that \(B(n) = \sum_{{k = 0}}^n S_2(n,k)\) (see `Stirling2`

(16.1-6)). \(B(n)/n!\) is the coefficient of \(x^n\) in the formal series \(\exp( \exp(x)-1 )\), which is the generating function for \(B(n)\).

gap> List( [0..6], n -> Bell( n ) ); [ 1, 1, 2, 5, 15, 52, 203 ] gap> Bell( 14 ); 190899322

`‣ Bernoulli` ( n ) | ( function ) |

returns the `n`-th *Bernoulli number* \(B_n\), which is defined by \(B_0 = 1\) and \(B_n = -\sum_{{k = 0}}^{{n-1}} {{n+1 \choose k}} B_k/(n+1)\).

\(B_n / n!\) is the coefficient of \(x^n\) in the power series of \(x / (\exp(x)-1)\). Except for \(B_1 = -1/2\) the Bernoulli numbers for odd indices are zero.

gap> Bernoulli( 4 ); -1/30 gap> Bernoulli( 10 ); 5/66 gap> Bernoulli( 12 ); # there is no simple pattern in Bernoulli numbers -691/2730 gap> Bernoulli( 50 ); # and they grow fairly fast 495057205241079648212477525/66

`‣ Stirling1` ( n, k ) | ( function ) |

returns the *Stirling number of the first kind* \(S_1(n,k)\) of the integers `n` and `k`. Stirling numbers of the first kind are defined by \(S_1(0,0) = 1\), \(S_1(n,0) = S_1(0,k) = 0\) if \(n, k \ne 0\) and the recurrence \(S_1(n,k) = (n-1) S_1(n-1,k) + S_1(n-1,k-1)\).

\(S_1(n,k)\) is the number of permutations of `n` points with `k` cycles. Stirling numbers of the first kind appear as coefficients in the series \(n! {{x \choose n}} = \sum_{{k = 0}}^n S_1(n,k) x^k\) which is the generating function for Stirling numbers of the first kind. Note the similarity to \(x^n = \sum_{{k = 0}}^n S_2(n,k) k! {{x \choose k}}\) (see `Stirling2`

(16.1-6)). Also the definition of \(S_1\) implies \(S_1(n,k) = S_2(-k,-n)\) if \(n, k < 0\). There are many formulae relating Stirling numbers of the first kind to Stirling numbers of the second kind, Bell numbers, and Binomial coefficients.

gap> # Knuth calls this the trademark of S_1: gap> List( [0..4], k -> Stirling1( 4, k ) ); [ 0, 6, 11, 6, 1 ] gap> List( [0..6], n->List( [0..6], k->Stirling1( n, k ) ) );; gap> # note the similarity with Pascal's triangle for Binomial numbers gap> PrintArray( last ); [ [ 1, 0, 0, 0, 0, 0, 0 ], [ 0, 1, 0, 0, 0, 0, 0 ], [ 0, 1, 1, 0, 0, 0, 0 ], [ 0, 2, 3, 1, 0, 0, 0 ], [ 0, 6, 11, 6, 1, 0, 0 ], [ 0, 24, 50, 35, 10, 1, 0 ], [ 0, 120, 274, 225, 85, 15, 1 ] ] gap> Stirling1(50,10); 101623020926367490059043797119309944043405505380503665627365376

`‣ Stirling2` ( n, k ) | ( function ) |

returns the *Stirling number of the second kind* \(S_2(n,k)\) of the integers `n` and `k`. Stirling numbers of the second kind are defined by \(S_2(0,0) = 1\), \(S_2(n,0) = S_2(0,k) = 0\) if \(n, k \ne 0\) and the recurrence \(S_2(n,k) = k S_2(n-1,k) + S_2(n-1,k-1)\).

\(S_2(n,k)\) is the number of ways to partition a set of `n` elements into `k` pairwise disjoint nonempty subsets (see `PartitionsSet`

(16.2-16)). Stirling numbers of the second kind appear as coefficients in the expansion of \(x^n = \sum_{{k = 0}}^n S_2(n,k) k! {{x \choose k}}\). Note the similarity to \(n! {{x \choose n}} = \sum_{{k = 0}}^n S_1(n,k) x^k\) (see `Stirling1`

(16.1-5)). Also the definition of \(S_2\) implies \(S_2(n,k) = S_1(-k,-n)\) if \(n, k < 0\). There are many formulae relating Stirling numbers of the second kind to Stirling numbers of the first kind, Bell numbers, and Binomial coefficients.

gap> # Knuth calls this the trademark of S_2: gap> List( [0..4], k->Stirling2( 4, k ) ); [ 0, 1, 7, 6, 1 ] gap> List( [0..6], n->List( [0..6], k->Stirling2( n, k ) ) );; gap> # note the similarity with Pascal's triangle for Binomial numbers gap> PrintArray( last ); [ [ 1, 0, 0, 0, 0, 0, 0 ], [ 0, 1, 0, 0, 0, 0, 0 ], [ 0, 1, 1, 0, 0, 0, 0 ], [ 0, 1, 3, 1, 0, 0, 0 ], [ 0, 1, 7, 6, 1, 0, 0 ], [ 0, 1, 15, 25, 10, 1, 0 ], [ 0, 1, 31, 90, 65, 15, 1 ] ] gap> Stirling2( 50, 10 ); 26154716515862881292012777396577993781727011

`‣ Combinations` ( mset[, k] ) | ( function ) |

returns the set of all combinations of the multiset `mset` (a list of objects which may contain the same object several times) with `k` elements; if `k` is not given it returns all combinations of `mset`.

A *combination* of `mset` is an unordered selection without repetitions and is represented by a sorted sublist of `mset`. If `mset` is a proper set, there are \({{|\textit{mset}| \choose \textit{k}}}\) (see `Binomial`

(16.1-2)) combinations with `k` elements, and the set of all combinations is just the *power set* of `mset`, which contains all *subsets* of `mset` and has cardinality \(2^{{|\textit{mset}|}}\).

To loop over combinations of a larger multiset use `IteratorOfCombinations`

(16.2-2) which produces combinations one by one and may save a lot of memory. Another memory efficient representation of the list of all combinations is provided by `EnumeratorOfCombinations`

(16.2-2).

`‣ IteratorOfCombinations` ( mset[, k] ) | ( function ) |

`‣ EnumeratorOfCombinations` ( mset ) | ( function ) |

`IteratorOfCombinations`

returns an `Iterator`

(30.8-1) for combinations (see `Combinations`

(16.2-1)) of the given multiset `mset`. If a non-negative integer `k` is given as second argument then only the combinations with `k` entries are produced, otherwise all combinations.

`EnumeratorOfCombinations`

returns an `Enumerator`

(30.3-2) of the given multiset `mset`. Currently only a variant without second argument `k` is implemented.

The ordering of combinations from these functions can be different and also different from the list returned by `Combinations`

(16.2-1).

gap> m:=[1..15];; Add(m, 15); gap> NrCombinations(m); 49152 gap> i := 0;; for c in Combinations(m) do i := i+1; od; gap> i; 49152 gap> cm := EnumeratorOfCombinations(m);; gap> cm[1000]; [ 1, 2, 3, 6, 7, 8, 9, 10 ] gap> Position(cm, [1,13,15,15]); 36866

`‣ NrCombinations` ( mset[, k] ) | ( function ) |

returns the number of `Combinations(`

.`mset`,`k`)

gap> Combinations( [1,2,2,3] ); [ [ ], [ 1 ], [ 1, 2 ], [ 1, 2, 2 ], [ 1, 2, 2, 3 ], [ 1, 2, 3 ], [ 1, 3 ], [ 2 ], [ 2, 2 ], [ 2, 2, 3 ], [ 2, 3 ], [ 3 ] ] gap> # number of different hands in a game of poker: gap> NrCombinations( [1..52], 5 ); 2598960

The function `Arrangements`

(16.2-4) computes ordered selections without repetitions, `UnorderedTuples`

(16.2-6) computes unordered selections with repetitions, and `Tuples`

(16.2-8) computes ordered selections with repetitions.

`‣ Arrangements` ( mset[, k] ) | ( function ) |

returns the set of arrangements of the multiset `mset` that contain `k` elements. If `k` is not given it returns all arrangements of `mset`.

An *arrangement* of `mset` is an ordered selection without repetitions and is represented by a list that contains only elements from `mset`, but maybe in a different order. If `mset` is a proper set there are \(|mset|! / (|mset|-k)!\) (see `Factorial`

(16.1-1)) arrangements with `k` elements.

`‣ NrArrangements` ( mset[, k] ) | ( function ) |

returns the number of `Arrangements(`

.`mset`,`k`)

As an example of arrangements of a multiset, think of the game Scrabble. Suppose you have the six characters of the word `"settle"`

and you have to make a four letter word. Then the possibilities are given by

gap> Arrangements( ["s","e","t","t","l","e"], 4 ); [ [ "e", "e", "l", "s" ], [ "e", "e", "l", "t" ], [ "e", "e", "s", "l" ], [ "e", "e", "s", "t" ], [ "e", "e", "t", "l" ], [ "e", "e", "t", "s" ], ... 93 more possibilities ... [ "t", "t", "l", "s" ], [ "t", "t", "s", "e" ], [ "t", "t", "s", "l" ] ]

Can you find the five proper English words, where `"lets"`

does not count? Note that the fact that the list returned by `Arrangements`

(16.2-4) is a proper set means in this example that the possibilities are listed in the same order as they appear in the dictionary.

gap> NrArrangements( ["s","e","t","t","l","e"] ); 523

The function `Combinations`

(16.2-1) computes unordered selections without repetitions, `UnorderedTuples`

(16.2-6) computes unordered selections with repetitions, and `Tuples`

(16.2-8) computes ordered selections with repetitions.

`‣ UnorderedTuples` ( set, k ) | ( function ) |

returns the set of all unordered tuples of length `k` of the set `set`.

An *unordered tuple* of length `k` of `set` is an unordered selection with repetitions of `set` and is represented by a sorted list of length `k` containing elements from `set`. There are \({{|set| + k - 1 \choose k}}\) (see `Binomial`

(16.1-2)) such unordered tuples.

Note that the fact that `UnorderedTuples`

returns a set implies that the last index runs fastest. That means the first tuple contains the smallest element from `set` `k` times, the second tuple contains the smallest element of `set` at all positions except at the last positions, where it contains the second smallest element from `set` and so on.

`‣ NrUnorderedTuples` ( set, k ) | ( function ) |

returns the number of `UnorderedTuples(`

.`set`,`k`)

As an example for unordered tuples think of a poker-like game played with 5 dice. Then each possible hand corresponds to an unordered five-tuple from the set \(\{ 1, 2, \ldots, 6 \}\).

gap> NrUnorderedTuples( [1..6], 5 ); 252 gap> UnorderedTuples( [1..6], 5 ); [ [ 1, 1, 1, 1, 1 ], [ 1, 1, 1, 1, 2 ], [ 1, 1, 1, 1, 3 ], [ 1, 1, 1, 1, 4 ], [ 1, 1, 1, 1, 5 ], [ 1, 1, 1, 1, 6 ], [ 1, 1, 1, 2, 2 ], [ 1, 1, 1, 2, 3 ], ... 100 more tuples ... [ 1, 3, 5, 5, 6 ], [ 1, 3, 5, 6, 6 ], [ 1, 3, 6, 6, 6 ], [ 1, 4, 4, 4, 4 ], ... 100 more tuples ... [ 3, 3, 5, 5, 5 ], [ 3, 3, 5, 5, 6 ], [ 3, 3, 5, 6, 6 ], [ 3, 3, 6, 6, 6 ], ... 32 more tuples ... [ 5, 5, 5, 6, 6 ], [ 5, 5, 6, 6, 6 ], [ 5, 6, 6, 6, 6 ], [ 6, 6, 6, 6, 6 ] ]

The function `Combinations`

(16.2-1) computes unordered selections without repetitions, `Arrangements`

(16.2-4) computes ordered selections without repetitions, and `Tuples`

(16.2-8) computes ordered selections with repetitions.

`‣ Tuples` ( set, k ) | ( function ) |

returns the set of all ordered tuples of length `k` of the set `set`.

An *ordered tuple* of length `k` of `set` is an ordered selection with repetition and is represented by a list of length `k` containing elements of `set`. There are \(|\textit{set}|^{\textit{k}}\) such ordered tuples.

Note that the fact that `Tuples`

returns a set implies that the last index runs fastest. That means the first tuple contains the smallest element from `set` `k` times, the second tuple contains the smallest element of `set` at all positions except at the last positions, where it contains the second smallest element from `set` and so on.

`‣ EnumeratorOfTuples` ( set, k ) | ( function ) |

This function is referred to as an example of enumerators that are defined by functions but are not constructed from a domain. The result is equal to that of `Tuples( `

. However, the entries are not stored physically in the list but are created/identified on demand.`set`, `k` )

`‣ IteratorOfTuples` ( set, k ) | ( function ) |

For a set `set` and a positive integer `k`, `IteratorOfTuples`

returns an iterator (see 30.8) of the set of all ordered tuples (see `Tuples`

(16.2-8)) of length `k` of the set `set`. The tuples are returned in lexicographic order.

`‣ NrTuples` ( set, k ) | ( function ) |

returns the number of `Tuples(`

.`set`,`k`)

gap> Tuples( [1,2,3], 2 ); [ [ 1, 1 ], [ 1, 2 ], [ 1, 3 ], [ 2, 1 ], [ 2, 2 ], [ 2, 3 ], [ 3, 1 ], [ 3, 2 ], [ 3, 3 ] ] gap> NrTuples( [1..10], 5 ); 100000

`Tuples(`

can also be viewed as the `set`,`k`)`k`-fold cartesian product of `set` (see `Cartesian`

(21.20-15)).

The function `Combinations`

(16.2-1) computes unordered selections without repetitions, `Arrangements`

(16.2-4) computes ordered selections without repetitions, and finally the function `UnorderedTuples`

(16.2-6) computes unordered selections with repetitions.

`‣ PermutationsList` ( mset ) | ( function ) |

`PermutationsList`

returns the set of permutations of the multiset `mset`.

A *permutation* is represented by a list that contains exactly the same elements as `mset`, but possibly in different order. If `mset` is a proper set there are \(|\textit{mset}| !\) (see `Factorial`

(16.1-1)) such permutations. Otherwise if the first elements appears \(k_1\) times, the second element appears \(k_2\) times and so on, the number of permutations is \(|\textit{mset}| ! / (k_1! k_2! \ldots)\), which is sometimes called multinomial coefficient.

`‣ NrPermutationsList` ( mset ) | ( function ) |

returns the number of `PermutationsList(`

.`mset`)

gap> PermutationsList( [1,2,3] ); [ [ 1, 2, 3 ], [ 1, 3, 2 ], [ 2, 1, 3 ], [ 2, 3, 1 ], [ 3, 1, 2 ], [ 3, 2, 1 ] ] gap> PermutationsList( [1,1,2,2] ); [ [ 1, 1, 2, 2 ], [ 1, 2, 1, 2 ], [ 1, 2, 2, 1 ], [ 2, 1, 1, 2 ], [ 2, 1, 2, 1 ], [ 2, 2, 1, 1 ] ] gap> NrPermutationsList( [1,2,2,3,3,3,4,4,4,4] ); 12600

The function `Arrangements`

(16.2-4) is the generalization of `PermutationsList`

(16.2-12) that allows you to specify the size of the permutations. `Derangements`

(16.2-14) computes permutations that have no fixed points.

`‣ Derangements` ( list ) | ( function ) |

returns the set of all derangements of the list `list`.

A *derangement* is a fixpointfree permutation of `list` and is represented by a list that contains exactly the same elements as `list`, but in such an order that the derangement has at no position the same element as `list`. If the list `list` contains no element twice there are exactly \(|\textit{list}|! (1/2! - 1/3! + 1/4! - \cdots + (-1)^n / n!)\) derangements.

Note that the ratio `NrPermutationsList( [ 1 .. n ] ) / NrDerangements( [ 1 .. n ] )`

, which is \(n! / (n! (1/2! - 1/3! + 1/4! - \cdots + (-1)^n / n!))\) is an approximation for the base of the natural logarithm \(e = 2.7182818285\ldots\), which is correct to about \(n\) digits.

`‣ NrDerangements` ( list ) | ( function ) |

returns the number of `Derangements(`

.`list`)

As an example of derangements suppose that you have to send four different letters to four different people. Then a derangement corresponds to a way to send those letters such that no letter reaches the intended person.

gap> Derangements( [1,2,3,4] ); [ [ 2, 1, 4, 3 ], [ 2, 3, 4, 1 ], [ 2, 4, 1, 3 ], [ 3, 1, 4, 2 ], [ 3, 4, 1, 2 ], [ 3, 4, 2, 1 ], [ 4, 1, 2, 3 ], [ 4, 3, 1, 2 ], [ 4, 3, 2, 1 ] ] gap> NrDerangements( [1..10] ); 1334961 gap> Int( 10^7*NrPermutationsList([1..10])/last ); 27182816 gap> Derangements( [1,1,2,2,3,3] ); [ [ 2, 2, 3, 3, 1, 1 ], [ 2, 3, 1, 3, 1, 2 ], [ 2, 3, 1, 3, 2, 1 ], [ 2, 3, 3, 1, 1, 2 ], [ 2, 3, 3, 1, 2, 1 ], [ 3, 2, 1, 3, 1, 2 ], [ 3, 2, 1, 3, 2, 1 ], [ 3, 2, 3, 1, 1, 2 ], [ 3, 2, 3, 1, 2, 1 ], [ 3, 3, 1, 1, 2, 2 ] ] gap> NrDerangements( [1,2,2,3,3,3,4,4,4,4] ); 338

The function `PermutationsList`

(16.2-12) computes all permutations of a list.

`‣ PartitionsSet` ( set[, k] ) | ( function ) |

returns the set of all unordered partitions of the set `set` into `k` pairwise disjoint nonempty sets. If `k` is not given it returns all unordered partitions of `set` for all `k`.

An *unordered partition* of `set` is a set of pairwise disjoint nonempty sets with union `set` and is represented by a sorted list of such sets. There are \(B( |set| )\) (see `Bell`

(16.1-3)) partitions of the set `set` and \(S_2( |set|, k )\) (see `Stirling2`

(16.1-6)) partitions with `k` elements.

`‣ NrPartitionsSet` ( set[, k] ) | ( function ) |

returns the number of `PartitionsSet(`

.`set`,`k`)

gap> PartitionsSet( [1,2,3] ); [ [ [ 1 ], [ 2 ], [ 3 ] ], [ [ 1 ], [ 2, 3 ] ], [ [ 1, 2 ], [ 3 ] ], [ [ 1, 2, 3 ] ], [ [ 1, 3 ], [ 2 ] ] ] gap> PartitionsSet( [1,2,3,4], 2 ); [ [ [ 1 ], [ 2, 3, 4 ] ], [ [ 1, 2 ], [ 3, 4 ] ], [ [ 1, 2, 3 ], [ 4 ] ], [ [ 1, 2, 4 ], [ 3 ] ], [ [ 1, 3 ], [ 2, 4 ] ], [ [ 1, 3, 4 ], [ 2 ] ], [ [ 1, 4 ], [ 2, 3 ] ] ] gap> NrPartitionsSet( [1..6] ); 203 gap> NrPartitionsSet( [1..10], 3 ); 9330

Note that `PartitionsSet`

(16.2-16) does currently not support multisets and that there is currently no ordered counterpart.

`‣ Partitions` ( n[, k] ) | ( function ) |

returns the set of all (unordered) partitions of the positive integer `n` into sums with `k` summands. If `k` is not given it returns all unordered partitions of `n` for all `k`.

An *unordered partition* is an unordered sum \(n = p_1 + p_2 + \cdots + p_k\) of positive integers and is represented by the list \(p = [ p_1, p_2, \ldots, p_k ]\), in nonincreasing order, i.e., \(p_1 \geq p_2 \geq \ldots \geq p_k\). We write \(p \vdash n\). There are approximately \(\exp(\pi \sqrt{{2/3 n}}) / (4 \sqrt{{3}} n)\) such partitions, use `NrPartitions`

(16.2-21) to compute the precise number.

If you want to loop over all partitions of some larger `n` use the more memory efficient `IteratorOfPartitions`

(16.2-19).

It is possible to associate with every partition of the integer `n` a conjugacy class of permutations in the symmetric group on `n` points and vice versa. Therefore \(p(n) := \)`NrPartitions`

\((n)\) is the number of conjugacy classes of the symmetric group on `n` points.

Ramanujan found the identities \(p(5i+4) = 0\) mod 5, \(p(7i+5) = 0\) mod 7 and \(p(11i+6) = 0\) mod 11 and many other fascinating things about the number of partitions.

`‣ IteratorOfPartitions` ( n ) | ( function ) |

For a positive integer `n`, `IteratorOfPartitions`

returns an iterator (see 30.8) of the set of partitions of `n` (see `Partitions`

(16.2-18)). The partitions of `n` are returned in lexicographic order.

`‣ IteratorOfPartitionsSet` ( set[, k[, flag]] ) | ( function ) |

`IteratorOfPartitionsSet`

returns an iterator (see 30.8) for all unordered partitions of the set `set` into pairwise disjoint nonempty sets (see `PartitionsSet`

(16.2-16)). If `k` given and `flag` is omitted or equal to `false`

, then only partitions of size `k` are computed. If `k` is given and `flag` is equal to `true`

, then only partitions of size at most `k` are computed.

`‣ NrPartitions` ( n[, k] ) | ( function ) |

returns the number of `Partitions(`

.`set`,`k`)

gap> Partitions( 7 ); [ [ 1, 1, 1, 1, 1, 1, 1 ], [ 2, 1, 1, 1, 1, 1 ], [ 2, 2, 1, 1, 1 ], [ 2, 2, 2, 1 ], [ 3, 1, 1, 1, 1 ], [ 3, 2, 1, 1 ], [ 3, 2, 2 ], [ 3, 3, 1 ], [ 4, 1, 1, 1 ], [ 4, 2, 1 ], [ 4, 3 ], [ 5, 1, 1 ], [ 5, 2 ], [ 6, 1 ], [ 7 ] ] gap> Partitions( 8, 3 ); [ [ 3, 3, 2 ], [ 4, 2, 2 ], [ 4, 3, 1 ], [ 5, 2, 1 ], [ 6, 1, 1 ] ] gap> NrPartitions( 7 ); 15 gap> NrPartitions( 100 ); 190569292

The function `OrderedPartitions`

(16.2-22) is the ordered counterpart of `Partitions`

(16.2-18).

`‣ OrderedPartitions` ( n[, k] ) | ( function ) |

returns the set of all ordered partitions of the positive integer `n` into sums with `k` summands. If `k` is not given it returns all ordered partitions of `set` for all `k`.

An *ordered partition* is an ordered sum \(n = p_1 + p_2 + \ldots + p_k\) of positive integers and is represented by the list \([ p_1, p_2, \ldots, p_k ]\). There are totally \(2^{{n-1}}\) ordered partitions and \({{n-1 \choose k-1}}\) (see `Binomial`

(16.1-2)) ordered partitions with `k` summands.

Do not call `OrderedPartitions`

with an `n` much larger than \(15\), the list will simply become too large.

`‣ NrOrderedPartitions` ( n[, k] ) | ( function ) |

returns the number of `OrderedPartitions(`

.`set`,`k`)

gap> OrderedPartitions( 5 ); [ [ 1, 1, 1, 1, 1 ], [ 1, 1, 1, 2 ], [ 1, 1, 2, 1 ], [ 1, 1, 3 ], [ 1, 2, 1, 1 ], [ 1, 2, 2 ], [ 1, 3, 1 ], [ 1, 4 ], [ 2, 1, 1, 1 ], [ 2, 1, 2 ], [ 2, 2, 1 ], [ 2, 3 ], [ 3, 1, 1 ], [ 3, 2 ], [ 4, 1 ], [ 5 ] ] gap> OrderedPartitions( 6, 3 ); [ [ 1, 1, 4 ], [ 1, 2, 3 ], [ 1, 3, 2 ], [ 1, 4, 1 ], [ 2, 1, 3 ], [ 2, 2, 2 ], [ 2, 3, 1 ], [ 3, 1, 2 ], [ 3, 2, 1 ], [ 4, 1, 1 ] ] gap> NrOrderedPartitions(20); 524288

The function `Partitions`

(16.2-18) is the unordered counterpart of `OrderedPartitions`

(16.2-22).

`‣ PartitionsGreatestLE` ( n, m ) | ( function ) |

returns the set of all (unordered) partitions of the integer `n` having parts less or equal to the integer `m`.

`‣ PartitionsGreatestEQ` ( n, m ) | ( function ) |

returns the set of all (unordered) partitions of the integer `n` having greatest part equal to the integer `m`.

`‣ RestrictedPartitions` ( n, set[, k] ) | ( function ) |

In the first form `RestrictedPartitions`

returns the set of all restricted partitions of the positive integer `n` into sums with `k` summands with the summands of the partition coming from the set `set`. If `k` is not given all restricted partitions for all `k` are returned.

A *restricted partition* is like an ordinary partition (see `Partitions`

(16.2-18)) an unordered sum \(n = p_1 + p_2 + \ldots + p_k\) of positive integers and is represented by the list \(p = [ p_1, p_2, \ldots, p_k ]\), in nonincreasing order. The difference is that here the \(p_i\) must be elements from the set `set`, while for ordinary partitions they may be elements from `[ 1 .. n ]`

.

`‣ NrRestrictedPartitions` ( n, set[, k] ) | ( function ) |

returns the number of `RestrictedPartitions(`

.`n`,`set`,`k`)

gap> RestrictedPartitions( 8, [1,3,5,7] ); [ [ 1, 1, 1, 1, 1, 1, 1, 1 ], [ 3, 1, 1, 1, 1, 1 ], [ 3, 3, 1, 1 ], [ 5, 1, 1, 1 ], [ 5, 3 ], [ 7, 1 ] ] gap> NrRestrictedPartitions(50,[1,2,5,10,20,50]); 451

The last example tells us that there are 451 ways to return 50 pence change using 1, 2, 5, 10, 20 and 50 pence coins.

`‣ SignPartition` ( pi ) | ( function ) |

returns the sign of a permutation with cycle structure `pi`.

This function actually describes a homomorphism from the symmetric group \(S_n\) into the cyclic group of order 2, whose kernel is exactly the alternating group \(A_n\) (see `SignPerm`

(42.4-1)). Partitions of sign 1 are called *even* partitions while partitions of sign \(-1\) are called *odd*.

gap> SignPartition([6,5,4,3,2,1]); -1

`‣ AssociatedPartition` ( pi ) | ( function ) |

`AssociatedPartition`

returns the associated partition of the partition `pi` which is obtained by transposing the corresponding Young diagram.

gap> AssociatedPartition([4,2,1]); [ 3, 2, 1, 1 ] gap> AssociatedPartition([6]); [ 1, 1, 1, 1, 1, 1 ]

`‣ PowerPartition` ( pi, k ) | ( function ) |

`PowerPartition`

returns the partition corresponding to the `k`-th power of a permutation with cycle structure `pi`.

Each part \(l\) of `pi` is replaced by \(d = \gcd(l, k)\) parts \(l/d\). So if `pi` is a partition of \(n\) then \(\textit{pi}^{\textit{k}}\) also is a partition of \(n\). `PowerPartition`

describes the power map of symmetric groups.

gap> PowerPartition([6,5,4,3,2,1], 3); [ 5, 4, 2, 2, 2, 2, 1, 1, 1, 1 ]

`‣ PartitionTuples` ( n, r ) | ( function ) |

`PartitionTuples`

returns the list of all `r`-tuples of partitions which together form a partition of `n`.

`r`-tuples of partitions describe the classes and the characters of wreath products of groups with `r` conjugacy classes with the symmetric group on `n` points, see `CharacterTableWreathSymmetric`

(71.20-6) and `CharacterValueWreathSymmetric`

(71.20-7).

`‣ NrPartitionTuples` ( n, r ) | ( function ) |

returns the number of `PartitionTuples( `

.`n`, `r` )

gap> PartitionTuples(3, 2); [ [ [ 1, 1, 1 ], [ ] ], [ [ 1, 1 ], [ 1 ] ], [ [ 1 ], [ 1, 1 ] ], [ [ ], [ 1, 1, 1 ] ], [ [ 2, 1 ], [ ] ], [ [ 1 ], [ 2 ] ], [ [ 2 ], [ 1 ] ], [ [ ], [ 2, 1 ] ], [ [ 3 ], [ ] ], [ [ ], [ 3 ] ] ]

`‣ BetaSet` ( alpha ) | ( function ) |

For a list `alpha` that describes a partition of a nonnegative integer (see `Partitions`

(16.2-18)), `BetaSet`

returns the list of integers obtained by reversing the order of `alpha` and then adding the sequence `[ 0, 1, 2, ... ]`

of the same length, cf. [JK81, Section 2.7].

gap> BetaSet( [ 4, 2, 1 ] ); [ 1, 3, 6 ] gap> BetaSet( [] ); [ ]

`‣ Fibonacci` ( n ) | ( function ) |

returns the `n`th number of the *Fibonacci sequence*. The Fibonacci sequence \(F_n\) is defined by the initial conditions \(F_1 = F_2 = 1\) and the recurrence relation \(F_{{n+2}} = F_{{n+1}} + F_n\). For negative \(n\) we define \(F_n = (-1)^{{n+1}} F_{{-n}}\), which is consistent with the recurrence relation.

Using generating functions one can prove that \(F_n = \phi^n - 1/\phi^n\), where \(\phi\) is \((\sqrt{{5}} + 1)/2\), i.e., one root of \(x^2 - x - 1 = 0\). Fibonacci numbers have the property \(\gcd( F_m, F_n ) = F_{{\gcd(m,n)}}\). But a pair of Fibonacci numbers requires more division steps in Euclid's algorithm (see `Gcd`

(56.7-1)) than any other pair of integers of the same size. `Fibonacci(`

is the special case `k`)`Lucas(1,-1,`

(see `k`)[1]`Lucas`

(16.3-2)).

gap> Fibonacci( 10 ); 55 gap> Fibonacci( 35 ); 9227465 gap> Fibonacci( -10 ); -55

`‣ Lucas` ( P, Q, k ) | ( function ) |

returns the `k`-th values of the *Lucas sequence* with parameters `P` and `Q`, which must be integers, as a list of three integers. If `k` is a negative integer, then the values of the Lucas sequence may be nonintegral rational numbers, with denominator roughly `Q`^`k`.

Let \(\alpha, \beta\) be the two roots of \(x^2 - P x + Q\) then we define `Lucas( `

\(= U_k = (\alpha^k - \beta^k) / (\alpha - \beta)\) and `P`, `Q`, `k` )[1]`Lucas( `

\(= V_k = (\alpha^k + \beta^k)\) and as a convenience `P`, `Q`, `k` )[2]`Lucas( `

\(= Q^k\).`P`, `Q`, `k` )[3]

The following recurrence relations are easily derived from the definition \(U_0 = 0, U_1 = 1, U_k = P U_{{k-1}} - Q U_{{k-2}}\) and \(V_0 = 2, V_1 = P, V_k = P V_{{k-1}} - Q V_{{k-2}}\). Those relations are actually used to define `Lucas`

if \(\alpha = \beta\).

Also the more complex relations used in `Lucas`

can be easily derived \(U_{2k} = U_k V_k\), \(U_{{2k+1}} = (P U_{2k} + V_{2k}) / 2\) and \(V_{2k} = V_k^2 - 2 Q^k\), \(V_{{2k+1}} = ((P^2-4Q) U_{2k} + P V_{2k}) / 2\).

`Fibonacci(`

(see `k`)`Fibonacci`

(16.3-1)) is simply `Lucas(1,-1,`

. In an abuse of notation, the sequence `k`)[1]`Lucas(1,-1,`

is sometimes called the Lucas sequence.`k`)[2]

gap> List( [0..10], i -> Lucas(1,-2,i)[1] ); # 2^k - (-1)^k)/3 [ 0, 1, 1, 3, 5, 11, 21, 43, 85, 171, 341 ] gap> List( [0..10], i -> Lucas(1,-2,i)[2] ); # 2^k + (-1)^k [ 2, 1, 5, 7, 17, 31, 65, 127, 257, 511, 1025 ] gap> List( [0..10], i -> Lucas(1,-1,i)[1] ); # Fibonacci sequence [ 0, 1, 1, 2, 3, 5, 8, 13, 21, 34, 55 ] gap> List( [0..10], i -> Lucas(2,1,i)[1] ); # the roots are equal [ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10 ]

`‣ Permanent` ( mat ) | ( attribute ) |

returns the *permanent* of the matrix `mat`. The permanent is defined by \(\sum_{{p \in Sym(n)}} \prod_{{i = 1}}^n mat[i][i^p]\).

Note the similarity of the definition of the permanent to the definition of the determinant (see `DeterminantMat`

(24.4-4)). In fact the only difference is the missing sign of the permutation. However the permanent is quite unlike the determinant, for example it is not multilinear or alternating. It has however important combinatorial properties.

gap> Permanent( [[0,1,1,1], > [1,0,1,1], > [1,1,0,1], > [1,1,1,0]] ); # inefficient way to compute NrDerangements([1..4]) 9 gap> # 24 permutations fit the projective plane of order 2: gap> Permanent( [[1,1,0,1,0,0,0], > [0,1,1,0,1,0,0], > [0,0,1,1,0,1,0], > [0,0,0,1,1,0,1], > [1,0,0,0,1,1,0], > [0,1,0,0,0,1,1], > [1,0,1,0,0,0,1]] ); 24

Goto Chapter: Top 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 Bib Ind

generated by GAPDoc2HTML