refactor: define hasher traits in nomt-core hasher module and feature…

…-gate

benchtop/src/nomt.rs

@@ -1,8 +1,8 @@
 use crate::{backend::Transaction, timer::Timer, workload::Workload};
 use fxhash::FxHashMap;
 use nomt::{
-    trie::KeyPath, Blake3Hasher, KeyReadWrite, Nomt, Options, Overlay, Session, SessionParams,
-    WitnessMode,
+    hasher::Blake3Hasher, trie::KeyPath, KeyReadWrite, Nomt, Options, Overlay, Session,
+    SessionParams, WitnessMode,
 };
 use sha2::Digest;
 use std::{

core/Cargo.toml

@@ -16,11 +16,15 @@ hex = { version = "0.4.3", default-features = false, features = ["alloc"] }
 ruint = { version = "1.12.1", default-features = false }
 arrayvec = { version = "0.7", default-features = false }
 borsh = { version = ">=1.4, <1.5.0", default-features = false, features = ["derive"], optional = true }
+blake3 = { version = "1.5.1", default-features = false, optional = true }
+sha2 = { version = "0.10.6" , default-features = false, optional = true }
 
 [dev-dependencies]
 blake3 = "1.5.1"
 
 [features]
-default = ["std"]
+default = ["std", "blake3-hasher", "sha2-hasher"]
 std = ["bitvec/std", "borsh/std"]
 borsh = ["dep:borsh"]
+blake3-hasher = ["dep:blake3"]
+sha2-hasher = ["dep:sha2"]

core/src/hasher.rs

@@ -0,0 +1,162 @@
+//! Hashers (feature-gated) and utilities for implementing them.
+
+use crate::trie::{InternalData, LeafData, Node, NodeKind, TERMINATOR};
+
+/// A trie node hash function specialized for 64 bytes of data.
+///
+/// Note that it is illegal for the produced hash to equal [0; 32], as this value is reserved
+/// for the terminator node.
+///
+/// A node hasher should domain-separate internal and leaf nodes in some specific way. The
+/// recommended approach for binary hashes is to set the MSB to 0 or 1 depending on the node kind.
+/// However, for other kinds of hashes (e.g. Poseidon2 or other algebraic hashes), other labeling
+/// schemes may be required.
+pub trait NodeHasher {
+    /// Hash a leaf. This should domain-separate the hash
+    /// according to the node kind.
+    fn hash_leaf(data: &LeafData) -> [u8; 32];
+
+    /// Hash an internal node. This should domain-separate
+    /// the hash according to the node kind.
+    fn hash_internal(data: &InternalData) -> [u8; 32];
+
+    /// Get the kind of the given node.
+    fn node_kind(node: &Node) -> NodeKind;
+}
+
+/// A hasher for arbitrary-length values.
+pub trait ValueHasher {
+    /// Hash an arbitrary-length value.
+    fn hash_value(value: &[u8]) -> [u8; 32];
+}
+
+/// Get the node kind, according to a most-significant bit labeling scheme.
+///
+/// If the MSB is true, it's a leaf. If the node is empty, it's a [`TERMINATOR`]. Otherwise, it's
+/// an internal node.
+pub fn node_kind_by_msb(node: &Node) -> NodeKind {
+    if node[0] >> 7 == 1 {
+        NodeKind::Leaf
+    } else if node == &TERMINATOR {
+        NodeKind::Terminator
+    } else {
+        NodeKind::Internal
+    }
+}
+
+/// Set the most-significant bit of the node.
+pub fn set_msb(node: &mut Node) {
+    node[0] |= 0b10000000;
+}
+
+pub fn unset_msb(node: &mut Node) {
+    node[0] &= 0b01111111;
+}
+
+/// A simple trait for representing binary hash functions.
+pub trait BinaryHash {
+    /// Given a bit-string, produce a 32-bit hash.
+    fn hash(input: &[u8]) -> [u8; 32];
+
+    /// An optional specialization of `hash` where there are two 32-byte inputs, left and right.
+    fn hash2_32_concat(left: &[u8; 32], right: &[u8; 32]) -> [u8; 32] {
+        let mut buf = [0u8; 64];
+        buf[0..32].copy_from_slice(left);
+        buf[32..64].copy_from_slice(right);
+        Self::hash(&buf)
+    }
+}
+
+/// A node and value hasher constructed from a simple binary hasher.
+///
+/// This implements a [`ValueHasher`] and [`NodeHasher`] where the node kind is tagged by setting
+/// or unsetting the MSB of the hash value.
+///
+/// The binary hash wrapped by this structure must behave approximately like a random oracle over
+/// the space 2^256, i.e. all 256 bit outputs are valid and inputs are uniformly distributed.
+///
+/// Functions like Sha2/Blake3/Keccak/Groestl all meet these criteria.
+pub struct BinaryHasher<H>(std::marker::PhantomData<H>);
+
+impl<H: BinaryHash> ValueHasher for BinaryHasher<H> {
+    fn hash_value(value: &[u8]) -> [u8; 32] {
+        H::hash(value)
+    }
+}
+
+impl<H: BinaryHash> NodeHasher for BinaryHasher<H> {
+    fn hash_leaf(data: &LeafData) -> [u8; 32] {
+        let mut h = H::hash2_32_concat(&data.key_path, &data.value_hash);
+        set_msb(&mut h);
+        h
+    }
+
+    fn hash_internal(data: &InternalData) -> [u8; 32] {
+        let mut h = H::hash2_32_concat(&data.left, &data.right);
+        unset_msb(&mut h);
+        h
+    }
+
+    fn node_kind(node: &Node) -> NodeKind {
+        node_kind_by_msb(node)
+    }
+}
+
+#[cfg(any(feature = "blake3-hasher", test))]
+pub use blake3::Blake3Hasher;
+
+/// A node hasher making use of blake3.
+#[cfg(any(feature = "blake3-hasher", test))]
+pub mod blake3 {
+    use super::{BinaryHash, BinaryHasher};
+
+    /// A [`BinaryHash`] implementation for Blake3.
+    pub struct Blake3BinaryHasher;
+
+    /// A wrapper around Blake3 for use in NOMT.
+    pub type Blake3Hasher = BinaryHasher<Blake3BinaryHasher>;
+
+    impl BinaryHash for Blake3BinaryHasher {
+        fn hash(value: &[u8]) -> [u8; 32] {
+            blake3::hash(value).into()
+        }
+
+        fn hash2_32_concat(left: &[u8; 32], right: &[u8; 32]) -> [u8; 32] {
+            let mut hasher = blake3::Hasher::new();
+            hasher.update(left);
+            hasher.update(right);
+            hasher.finalize().into()
+        }
+    }
+}
+
+#[cfg(feature = "sha2-hasher")]
+pub use sha2::Sha2Hasher;
+
+/// A node and value hasher making use of sha2-256.
+#[cfg(feature = "sha2-hasher")]
+pub mod sha2 {
+    use super::{BinaryHash, BinaryHasher};
+    use sha2::{Digest, Sha256};
+
+    /// A [`BinaryHash`] implementation for Sha2.
+    pub struct Sha2BinaryHasher;
+
+    /// A wrapper around sha2-256 for use in NOMT.
+    pub type Sha2Hasher = BinaryHasher<Sha2BinaryHasher>;
+
+    impl BinaryHash for Sha2BinaryHasher {
+        fn hash(value: &[u8]) -> [u8; 32] {
+            let mut hasher = Sha256::new();
+            hasher.update(value);
+            hasher.finalize().into()
+        }
+
+        fn hash2_32_concat(left: &[u8; 32], right: &[u8; 32]) -> [u8; 32] {
+            let mut hasher = Sha256::new();
+            hasher.update(left);
+            hasher.update(right);
+            hasher.finalize().into()
+        }
+    }
+}

core/src/lib.rs

@@ -10,6 +10,7 @@
 
 extern crate alloc;
 
+pub mod hasher;
 pub mod page;
 pub mod page_id;
 pub mod proof;

core/src/proof/multi_proof.rs

@@ -3,11 +3,12 @@
 //! the inclusion of all provided path proofs.
 
 use crate::{
+    hasher::NodeHasher,
     proof::{
         path_proof::{hash_path, shared_bits},
         KeyOutOfScope, PathProof, PathProofTerminal,
     },
-    trie::{InternalData, KeyPath, LeafData, Node, NodeHasher, TERMINATOR},
+    trie::{InternalData, KeyPath, LeafData, Node, TERMINATOR},
 };
 
 #[cfg(not(feature = "std"))]
@@ -497,8 +498,9 @@ mod tests {
     use super::{verify, MultiProof};
 
     use crate::{
+        hasher::{Blake3Hasher, NodeHasher},
         proof::{PathProof, PathProofTerminal},
-        trie::{self, InternalData, LeafData, Node, NodeHasher, NodeKind, TERMINATOR},
+        trie::{InternalData, LeafData, TERMINATOR},
         trie_pos::TriePosition,
     };
 
@@ -884,43 +886,6 @@ mod tests {
         );
     }
 
-    /// Hash nodes with blake3.
-    pub struct Blake3Hasher;
-
-    impl NodeHasher for Blake3Hasher {
-        fn hash_leaf(data: &trie::LeafData) -> [u8; 32] {
-            let mut hasher = blake3::Hasher::new();
-            hasher.update(&data.key_path);
-            hasher.update(&data.value_hash);
-            let mut hash: [u8; 32] = hasher.finalize().into();
-
-            // Label with MSB
-            hash[0] |= 0b10000000;
-            hash
-        }
-
-        fn hash_internal(data: &trie::InternalData) -> [u8; 32] {
-            let mut hasher = blake3::Hasher::new();
-            hasher.update(&data.left);
-            hasher.update(&data.right);
-            let mut hash: [u8; 32] = hasher.finalize().into();
-
-            // Label with MSB
-            hash[0] &= 0b01111111;
-            hash
-        }
-
-        fn node_kind(node: &Node) -> NodeKind {
-            if node[0] >> 7 == 1 {
-                NodeKind::Leaf
-            } else if node == &TERMINATOR {
-                NodeKind::Terminator
-            } else {
-                NodeKind::Internal
-            }
-        }
-    }
-
     #[test]
     pub fn test_verify_multiproof_two_leafs() {
         //     root

core/src/proof/path_proof.rs

@@ -1,6 +1,7 @@
 //! Proving and verifying inclusion, non-inclusion, and updates to the trie.
 
-use crate::trie::{self, InternalData, KeyPath, LeafData, Node, NodeHasher, NodeKind, TERMINATOR};
+use crate::hasher::NodeHasher;
+use crate::trie::{self, InternalData, KeyPath, LeafData, Node, NodeKind, TERMINATOR};
 use crate::trie_pos::TriePosition;
 
 use bitvec::prelude::*;

core/src/trie.rs

@@ -13,6 +13,8 @@
 //!
 //! All node preimages are 512 bits.
 
+use crate::hasher::NodeHasher;
+
 /// A node in the binary trie. In this schema, it is always 256 bits and is the hash of either
 /// an [`LeafData`] or [`InternalData`], or zeroed if it's a [`TERMINATOR`].
 ///
@@ -90,25 +92,3 @@ pub struct LeafData {
     /// The hash of the value carried in this leaf.
     pub value_hash: ValueHash,
 }
-
-/// A trie node hash function specialized for 64 bytes of data.
-///
-/// Note that it is illegal for the produced hash to equal [0; 32], as this value is reserved
-/// for the terminator node.
-///
-/// A node hasher should domain-separate internal and leaf nodes in some specific way. The
-/// recommended approach for binary hashes is to set the MSB to 0 or 1 depending on the node kind.
-/// However, for other kinds of hashes (e.g. Poseidon2 or other algebraic hashes), other labeling
-/// schemes may be required.
-pub trait NodeHasher {
-    /// Hash a leaf. This should domain-separate the hash
-    /// according to the node kind.
-    fn hash_leaf(data: &LeafData) -> [u8; 32];
-
-    /// Hash an internal node. This should domain-separate
-    /// the hash according to the node kind.
-    fn hash_internal(data: &InternalData) -> [u8; 32];
-
-    /// Get the kind of the given node.
-    fn node_kind(node: &Node) -> NodeKind;
-}

core/src/update.rs

@@ -1,6 +1,7 @@
 //! Trie update logic helpers.
 
-use crate::trie::{self, KeyPath, LeafData, Node, NodeHasher, ValueHash};
+use crate::hasher::NodeHasher;
+use crate::trie::{self, KeyPath, LeafData, Node, ValueHash};
 
 use bitvec::prelude::*;
 

examples/commit_batch/src/lib.rs

@@ -1,5 +1,7 @@
 use anyhow::Result;
-use nomt::{Blake3Hasher, KeyReadWrite, Nomt, Options, Root, SessionParams, Witness, WitnessMode};
+use nomt::{
+    hasher::Blake3Hasher, KeyReadWrite, Nomt, Options, Root, SessionParams, Witness, WitnessMode,
+};
 use sha2::Digest;
 
 const NOMT_DB_FOLDER: &str = "nomt_db";

examples/read_value/src/main.rs

@@ -1,5 +1,5 @@
 use anyhow::Result;
-use nomt::{Blake3Hasher, KeyReadWrite, Nomt, Options, SessionParams, WitnessMode};
+use nomt::{hasher::Blake3Hasher, KeyReadWrite, Nomt, Options, SessionParams, WitnessMode};
 use sha2::Digest;
 
 const NOMT_DB_FOLDER: &str = "nomt_db";