In basic combinatorics, when dividing a total of N objects into k classes each with $r_k$ objects we have Partition Formula: $$ \binom{N}{r_1}\binom{N-r_1}{r_2}...\binom{r_k}{r_k}=\frac{N!}{r_1! r_2!...r_k!} $$
But if you try to divide $2$ into $2$ classes, or $3$ into $3$ classes, the formula gives you $2$ and $6$, but apparently there is only one way to divide it. But it functions normally when you divide $2$ into $2$ and zero and $3$ into $1$ and $2$.
Since we deduce the Partition Formula from n choose k method, we can see that in the case of dividing $2$ into $1$ and $1$, it first chooses $1$ out of $2$, which has $2$ ways. And then choose the left one. So this process actually arranges the separation orderly. $(a,b)$ and $(b,a)$ are different. But in the case of dividing $3$ into $1$ and $2$, we end up with $(a, bc)$, $(b, ac)$ and $(c, ab)$. Not ordered, which is good.
Why? I'm confused. Help me.
This is the case when the classes are identical, i.e., whether your elements $a$ and $b$ in the first class or in the second class is not important. This formula is given for distinct classes, therefore for "2 into 2" case, we have $2! = 2$ such ways and for "3 into 3" case, we have $3! = 6$ such cases.
There is a problem here since formula does not have to count $(bc,a)$, $(ac,b)$, $(ab,c)$ when we divide them with $1+2$ and not $2+1$. So, there are $3$ ways to divide them, which is $\binom{3}{1}\binom{2}{2}$, consistent with the formula.
I think the confusion comes from $1+1+1$ and $1+2$ cases. Notice that when we have $3$ classes for $3$ objects, we can permute the objects into classes since the size of the classes are the same for all three. However, in $1+2$ case, we cannot permute them since it restricts us to have $1$ object in the first class and $2$ objects in the second class.