TL;DR

We go on with finding all partitions of a set, following the track started with PWC108 - Bell Numbers.

In previous post All positive integer sums we laid out a possible strategy for finding all (distinct) partitions of a set.

The first half was to find out all possible ways to express a positive integer as the sum of other (lower or equal) positive integers; this has been addressed and led us to the iterator-based solution described in All positive integer sums, as iterator.

Now… we’re only left with generating the sets starting from these partial sums. Let’s take a first look at the case for $N = 3$:

$$(3) \Rightarrow \{\{a, b, c\}\} \\ (2 + 1) \Rightarrow \{\{a, b\}, \{c\}\} \\ (2 + 1) \Rightarrow \{\{a, c\}, \{b\}\} \\ (2 + 1) \Rightarrow \{\{b, c\}, \{a\}\} \\ (1 + 1 + 1) \Rightarrow \{\{a\}, \{b\}, \{c\}\}$$

There is an obvious factor that has to be taken into considerations: we have three distinct expansions for $2 + 1$, but only one for $1 + 1 + 1$.

In general, any subset of equal addends in the sum have to be taken with care in order to avoid duplicates; this does not happen, of course, across different values. For this reason, the $2 + 2$ decomposition for $4$ has to be taken with care too, or we would have duplicates. In other words, the following are the only distinct partitions of the type $2 + 2$:

$$\{\{a, b\}, \{c, d\}\} \\ \{\{a, c\}, \{b, d\}\} \\ \{\{a, d\}, \{b, c\}\} \\$$

Any partition with the $a$ in the second subset would lead to a partition that is equivalent to one of the above, i.e. a duplicate.

Summing up, when generating all partitions starting from our decomposition of the integer input into possible sums, we will have to address the subsets of equal addends as a kind of unit with a specific algorithm.

For this reason, it’s useful to express the generic sum decomposition like this:

$$N = \sum_{j = 1}^{J}{k_j \cdot n_j}$$

where $n_j$ represents the addendum value and $k_j$ represents how many that addendum appears in the decomposition. This would mean the following:

$$3 = 3 = 1 \cdot 3 \\ 3 = 2 + 1 = 1 \cdot 2 + 1 \cdot 1 \\ 3 = 1 + 1 + 1 = 3 \cdot 1$$

Enough for the preliminary considerations… stay safe!