Consider $N=2^k$ and the set $[N] = \{0,1,\dots,N-1\}$. For each $n\in[N]$ there are $k$ unique numbers $n_i \in \{0,1\}$, $i = 0,\dots,k-1$ such that $n = \sum_i\ n_i\cdot 2^i$. The tupel $\langle n_{k-1}\dots n_0\rangle$ is the binary representation of $n$.
The tuple can be interpreted in another way, namely as a tuple of truth values, indicating which of $k$ independent arithmetic properties $p_0,\dots p_{k-1}$ a number $n$ possesses or not.
These properties can be defined via the parametrized arithmetic function $r_i(n) = n / 2^i \mod 2$:
$$p_i(n) :\equiv r_i(n) = 1$$
This yields a "truth table", e.g. for the numbers $n\in [8] = \{0,\dots,7\}$ (with $0$ for FALSE and $1$ for TRUE):
\begin{array}{|c|c|c|c|} \hline n& p_2 & p_1 & p_0 \\ \hline 0 & 0 & 0 & 0\\ \hline 1 & 0 & 0 & 1\\ \hline 2 & 0 & 1 & 0 \\ \hline 3 & 0 & 1 & 1 \\ \hline 4 & 1 & 0 & 0 \\ \hline 5 & 1 & 0 & 1 \\ \hline 6 & 1 & 1 & 0 \\ \hline 7 & 1 & 1 & 1 \\ \hline \end{array}
Note, that the columns indicate the extensions of the predicates $p_i$.
By permuation of $\{0,\dots,7\}$ it is easy to see that there are exactly $8!=40320$ triples of subsets $\langle q_2,q_1,q_0\rangle$ such that each $n \in [8]$ is uniquely defined by its membership in these subsets.
Some of these triples can be defined not just by the extensions of the $q_i$ but by some arithmetic properties.
We know already one such triple of arithmetic properties, and there are eight similar triples, because for each $p_i$ we could have independently chosen $r_i(n) = 1$ or $r_i(n) = 0$.
My question is:
Which "naturally" definable triples $\langle q_2,q_1,q_0\rangle$ of arithmetic properties come to mind, such that each $n < 8$ can be uniquely defined by them?
"Naturally" means: not by just giving their extensions.
Are there other parametrized properties like the $p_i$?
Are there properties not making use of the functions $r_i$?
The language in which the properties may be defined is intentionally unspecified.
(1) I found one triple of arithmetic properties, that is not parametrized like the $p_i$, but still makes use of the functions $r_i$.
They have their origin in somehow "physical" properties of the original binary representation:
Let $\mathsf{symm}(n)$ iff the binary representation of $n$ is symmetric, i.e.
$\mathsf{symm}(n) :\equiv r_0(n) = r_2(n)$.
Let $\mathsf{right}(n)$ iff (the binary representation is NOT symmetric AND the right most digit is $1$) OR (the binary representation is symmetric AND the middle digit is $1$), i.e.
$\mathsf{right}(n) :\equiv (r_0(n) \neq r_2(n) \wedge r_0(n)=1) \vee (r_0(n) = r_2(n) \wedge r_1(n)=1)$
Let $\mathsf{heavy}(n)$ iff the binary representation of $n$ contains more $1$s than $0$s, i.e.
$\mathsf{heavy}(n) :\equiv (r_0(n) = r_1(n) = 1) \vee (r_0(n) = r_2(n) = 1) \vee (r_1(n) = r_2(n) = 1)$.
This is the "truth" or "property table":
\begin{array}{|c|c|c|c|c|c|c|c|} \hline n& p_2 & p_1 & p_0 & \mathsf{symm} & \mathsf{right} & \mathsf{heavy} & \pi(n)\\ \hline 0 & 0 & 0 & 0 & 1 & 0 & 0 & 4\\ \hline 1 & 0 & 0 & 1 & 0 & 1 & 0 & 2\\ \hline 2 & 0 & 1 & 0 & 1 & 1 & 0 & 6\\ \hline 3 & 0 & 1 & 1 & 0 & 1 & 1 & 3\\ \hline 4 & 1 & 0 & 0 & 0 & 0 & 0 & 0\\ \hline 5 & 1 & 0 & 1 & 1 & 0 & 1 & 5\\ \hline 6 & 1 & 1 & 0 & 0 & 0 & 1 & 1\\ \hline 7 & 1 & 1 & 1 & 1 & 1 & 1 & 7\\ \hline \end{array}
Could this be some kind of a blue print for constructing further such triples of properties?