Struct ContentKey 

pub struct ContentKey<T> { /* private fields */ }

A content address derived by construction from a value of type T.

The type-level seal (M3 — the stale-key/decoupling axis)

The general ContentMemo::get_or_compute lets the key and the compute closure disagree: a caller can pass a key that is not a function of what compute actually reads (the stale-key footgun — the Sharing::PerSite/libxcrypt divergence class). ContentKey<T> removes that footgun structurally: its sole constructor is ContentKey::of, which hashes T’s structural read-set. So “the key IS the content of the input” is not a caller obligation — it holds by construction, and a key decoupled from its input has no way to be built.

Paired with ContentMemo::get_or_compute_keyed, which derives the key from the same &T it hands to compute, the key↔content decoupling axis is parse-time-rejected (there is no expressible program that memoizes under a key not derived from the computed input).

Honest ceiling — what this does NOT seal

This seals the KEY↔CONTENT structural axis only. It does not seal the purity of T → V: compute could still read wall-clock, getEnv, or a mutable filesystem — those are opaque to the type system. There is no PureFn in safe Rust and there cannot be, so the purity axis stays only-mitigated (C1) forever (the module invariant + the CI byte gate are the correct terminal enforcement). Do not read ContentKey<T> as a purity proof — it is a decoupling proof.

The digest is a 32-byte BLAKE3 over T’s Hash serialization, so the key is a stable, collision-resistant content address of the value’s structural fields.

The trait impls below are hand-written (not #[derive]d) so they hold for any T — the standard derive would demand T: Clone + Eq + Hash + … even though the only real field is the 32-byte digest (T lives only in a zero-size PhantomData).

Implementations

impl<T> ContentKey<T>

where T: Hash,

pub fn of(input: &T) -> ContentKey<T>

Derive the content key from input — the sole constructor.

Hashes input’s structural read-set (everything its Hash impl writes) through BLAKE3, so the returned key is a pure, deterministic function of input’s content. Two structurally-equal Ts produce the identical key; there is no way to construct a ContentKey<T> that is not the content address of some &T.

pub fn digest(&self) -> &[u8; 32]

The raw 32-byte BLAKE3 digest (for debugging / cross-checking).

Trait Implementations

impl<T> Clone for ContentKey<T>

fn clone(&self) -> ContentKey<T>

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl<T> Copy for ContentKey<T>

impl<T> Debug for ContentKey<T>

fn fmt(&self, f: &mut Formatter<'_>) -> Result<(), Error>

Formats the value using the given formatter.

impl<T> Eq for ContentKey<T>

impl<T> Hash for ContentKey<T>

fn hash<H>(&self, state: &mut H)

where H: Hasher,

Feeds this value into the given Hasher.

fn hash_slice<H>(data: &[Self], state: &mut H)

where H: Hasher, Self: Sized,

Feeds a slice of this type into the given Hasher.

impl<T> PartialEq for ContentKey<T>

fn eq(&self, other: &ContentKey<T>) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.