feat: blake2s

[AvivYossef-starkware]

Pull Request type:

Feature

Current behavior:
No Blake2s‐based StarkHash implementation.

New behavior:
Adds Blake2s struct implementing StarkHash (hash, hash_array, hash_single) plus encoding helpers and tests.

---

This change is 

[AvivYossef-starkware]

Reviewable status: 0 of 4 files reviewed, 1 unresolved discussion (waiting on @dorimedini-starkware)

---

ensure_no_std/Cargo.toml line 13 at r1 (raw file):

    "parity-scale-codec",
    "num-traits",
    "hash",

blake2 uses std ...

[noaov1]

Reviewed 3 of 4 files at r1, all commit messages.
Reviewable status: 3 of 4 files reviewed, 9 unresolved discussions (waiting on @dan-starkware and @dorimedini-starkware)

---

crates/starknet-types-core/src/hash/blake2s.rs line 14 at r1 (raw file):

impl StarkHash for Blake2s {
    fn hash(felt_0: &Felt, felt_1: &Felt) -> Felt {
        Self::encode_felt252_data_and_calc_224_bit_blake_hash(&[*felt_0, *felt_1])

Consider using hash_array.
Same below.

Code quote:

 Self::encode_felt252_data_and_calc_224_bit_blake_hash(&[*felt_0, *felt_1])

---

crates/starknet-types-core/src/hash/blake2s.rs line 34 at r1 (raw file):

    ///
    /// # Returns
    /// A flat `Vec<u32>` containing all the unpacked words, in the same order.

Suggestion:

    // Encode each `Felt` into 32-bit words:
    /// - **Small** values `< 2^63` gets **2** words: `[ high_32_bits, low_32_bits ]` from the last 8
    ///   bytes of the 256-bit BE representation.
    /// - **Large** values `>= 2^63` get **8** words: the full 32-byte big-endian split, **with** the
    ///   MSB of the first word set as a marker (`+2^255`).
    ///
    /// # Returns
    /// A flat `Vec<u32>` containing all the unpacked words, in the same order.

---

crates/starknet-types-core/src/hash/blake2s.rs line 35 at r1 (raw file):

    /// # Returns
    /// A flat `Vec<u32>` containing all the unpacked words, in the same order.
    pub fn encode_felts_to_u32s(felts: Vec<Felt>) -> Vec<u32> {

Suggestion:

pub fn encode_felts_to_u32s(felts: &[Felt]) -> Vec<u32> {

---

crates/starknet-types-core/src/hash/blake2s.rs line 41 at r1 (raw file):

        const BIG_MARKER: u32 = 1 << 31;

        let mut unpacked_u32s = Vec::new();

What does unpacked stand for?

Code quote:

let mut unpacked_u32s = Vec::new();

---

crates/starknet-types-core/src/hash/blake2s.rs line 45 at r1 (raw file):

            let felt_as_be_bytes = felt.to_bytes_be();
            if felt < SMALL_THRESHOLD {
                // small: 2 limbs only, high‐32 then low‐32 of the last 8 bytes

Suggestion:

                // small: 2 limbs only, extracts the LSB.

---

crates/starknet-types-core/src/hash/blake2s.rs line 47 at r1 (raw file):

                // small: 2 limbs only, high‐32 then low‐32 of the last 8 bytes
                let hi = u32::from_be_bytes(felt_as_be_bytes[24..28].try_into().unwrap());
                let lo = u32::from_be_bytes(felt_as_be_bytes[28..32].try_into().unwrap());

Suggestion:

                let high = u32::from_be_bytes(felt_as_be_bytes[24..28].try_into().unwrap());
                let low = u32::from_be_bytes(felt_as_be_bytes[28..32].try_into().unwrap());

---

crates/starknet-types-core/src/hash/blake2s.rs line 53 at r1 (raw file):

                // big: 8 limbs, big‐endian order
                let start = unpacked_u32s.len();
                for chunk in felt_as_be_bytes.chunks_exact(4) {

Consider defining it as a const.

Code quote:

4

---

crates/starknet-types-core/src/hash/blake2s.rs line 87 at r1 (raw file):

    /// with Blake2s-256 and returns the 224-bit truncated digest as a `Felt`.
    pub fn encode_felt252_data_and_calc_224_bit_blake_hash(data: &[Felt]) -> Felt {
        // 1) Unpack each Felt into 2 or 8 u32 limbs

?

Code quote:

limbs

[AvivYossef-starkware]

Reviewable status: 3 of 4 files reviewed, 5 unresolved discussions (waiting on @dan-starkware, @dorimedini-starkware, and @noaov1)

---

crates/starknet-types-core/src/hash/blake2s.rs line 14 at r1 (raw file):

Previously, noaov1 (Noa Oved) wrote…

Consider using hash_array.
Same below.

Done

---

crates/starknet-types-core/src/hash/blake2s.rs line 41 at r1 (raw file):

Previously, noaov1 (Noa Oved) wrote…

What does unpacked stand for?

they use this word in the cairo function
I can change it to encoded_u32s

---

crates/starknet-types-core/src/hash/blake2s.rs line 34 at r1 (raw file):

    ///
    /// # Returns
    /// A flat `Vec<u32>` containing all the unpacked words, in the same order.

Why? "values" is plural.

---

crates/starknet-types-core/src/hash/blake2s.rs line 35 at r1 (raw file):

    /// # Returns
    /// A flat `Vec<u32>` containing all the unpacked words, in the same order.
    pub fn encode_felts_to_u32s(felts: Vec<Felt>) -> Vec<u32> {

Why not take ownership?
it's like into

---

crates/starknet-types-core/src/hash/blake2s.rs line 45 at r1 (raw file):

            let felt_as_be_bytes = felt.to_bytes_be();
            if felt < SMALL_THRESHOLD {
                // small: 2 limbs only, high‐32 then low‐32 of the last 8 bytes

WTYT?

[Yoni-Starkware]

Reviewable status: 3 of 4 files reviewed, 7 unresolved discussions (waiting on @dan-starkware, @dorimedini-starkware, and @noaov1)

---

crates/starknet-types-core/src/hash/blake2s.rs line 10 at r2 (raw file):

use super::traits::StarkHash;

pub struct Blake2s;

Rename to reflect the fact that it is not a standard impl of blake.

Code quote:

pub struct Blake2s;

---

crates/starknet-types-core/src/hash/blake2s.rs line 27 at r2 (raw file):

impl Blake2s {
    // Encode each `Felt` into 32-bit words:

Suggestion:

/ Encodes eac

[AvivYossef-starkware]

Reviewable status: 3 of 4 files reviewed, 7 unresolved discussions (waiting on @dan-starkware, @dorimedini-starkware, @noaov1, and @Yoni-Starkware)

---

crates/starknet-types-core/src/hash/blake2s.rs line 10 at r2 (raw file):

Previously, Yoni-Starkware (Yoni) wrote…

Rename to reflect the fact that it is not a standard impl of blake.

WTYT?

---

crates/starknet-types-core/src/hash/blake2s.rs line 27 at r2 (raw file):

impl Blake2s {
    // Encode each `Felt` into 32-bit words:

Done.

[Yoni-Starkware]

Reviewable status: 3 of 4 files reviewed, 6 unresolved discussions (waiting on @AvivYossef-starkware, @dan-starkware, @dorimedini-starkware, and @noaov1)

---

crates/starknet-types-core/src/hash/blake2s.rs line 10 at r2 (raw file):

Previously, AvivYossef-starkware wrote…

WTYT?

Blake2Felt252 for consistency with Blake2s256 :)

[AvivYossef-starkware]

Reviewable status: 3 of 4 files reviewed, 6 unresolved discussions (waiting on @dan-starkware, @dorimedini-starkware, @noaov1, and @Yoni-Starkware)

---

crates/starknet-types-core/src/hash/blake2s.rs line 10 at r2 (raw file):

Previously, Yoni-Starkware (Yoni) wrote…

Blake2Felt252 for consistency with Blake2s256 :)

Done.

[tdelabro]

@AvivYossef-starkware @Yoni-Starkware is it ready to be reviewed and merged?

[AvivYossef-starkware]

Hi, I still have some tests to complete before the merge, but it’s ready for review.

Reviewable status: 3 of 4 files reviewed, 8 unresolved discussions (waiting on @dan-starkware, @dorimedini-starkware, @noaov1, @tdelabro, and @Yoni-Starkware)

[noaov1]

Reviewed 1 of 1 files at r5, all commit messages.
Reviewable status: all files reviewed, 7 unresolved discussions (waiting on @AvivYossef-starkware, @dan-starkware, @dorimedini-starkware, @tdelabro, and @Yoni-Starkware)

---

crates/starknet-types-core/src/hash/blake2s.rs line 35 at r1 (raw file):

Previously, AvivYossef-starkware wrote…

Why not take ownership?
it's like into

The input vector is read-only and not reused, mutated, or moved elsewhere — so there's no reason to take ownership.
(non-blocking)

---

crates/starknet-types-core/src/hash/blake2s.rs line 41 at r1 (raw file):

Previously, AvivYossef-starkware wrote…

they use this word in the cairo function
I can change it to encoded_u32s

I think that encoded_u32s is better (non-blocking)

---

crates/starknet-types-core/src/hash/blake2s.rs line 103 at r5 (raw file):

        let mut hasher = Blake2s256::new();
        hasher.update(data);
        let hash32 = hasher.finalize();

How do you know that hash32 contains at least 7 words?

Code quote:

        let hash32 = hasher.finalize();

[AvivYossef-starkware]

Reviewable status: all files reviewed, 7 unresolved discussions (waiting on @dan-starkware, @dorimedini-starkware, @noaov1, @tdelabro, and @Yoni-Starkware)

---

crates/starknet-types-core/src/hash/blake2s.rs line 35 at r1 (raw file):

Previously, noaov1 (Noa Oved) wrote…

The input vector is read-only and not reused, mutated, or moved elsewhere — so there's no reason to take ownership.
(non-blocking)

sababa

---

crates/starknet-types-core/src/hash/blake2s.rs line 41 at r1 (raw file):

Previously, noaov1 (Noa Oved) wrote…

I think that encoded_u32s is better (non-blocking)

I prefer to stick with the Cairo impl

---

crates/starknet-types-core/src/hash/blake2s.rs line 103 at r5 (raw file):

Previously, noaov1 (Noa Oved) wrote…

How do you know that hash32 contains at least 7 words?

Blake2s256 returns 256 bits

[Yoni-Starkware]

Reviewed all commit messages.
Reviewable status: 3 of 4 files reviewed, 3 unresolved discussions (waiting on @dan-starkware, @dorimedini-starkware, @noaov1, and @tdelabro)

[dorimedini-starkware]

Reviewed 3 of 4 files at r1, 1 of 1 files at r6, all commit messages.
Reviewable status: all files reviewed, 2 unresolved discussions (waiting on @dan-starkware and @tdelabro)