Make HashDB generic (#8739)

The `patricia_trie` crate is generic over the hasher (by way of HashDB) and node encoding scheme. Adds a new `patricia_trie_ethereum` crate with concrete impls for Keccak/RLP.

Cargo.toml

@@ -1,10 +1,10 @@
 [package]
 name = "hashdb"
-version = "0.1.1"
+version = "0.2.0"
 authors = ["Parity Technologies <admin@parity.io>"]
 description = "trait for hash-keyed databases."
 license = "GPL-3.0"
 
 [dependencies]
 elastic-array = "0.10"
-ethereum-types = "0.3"
+heapsize = "0.4"

src/lib.rs

@@ -14,65 +14,70 @@
 // You should have received a copy of the GNU General Public License
 // along with Parity.  If not, see <http://www.gnu.org/licenses/>.
 
-//! Database of byte-slices keyed to their Keccak hash.
+//! Database of byte-slices keyed to their hash.
 extern crate elastic_array;
-extern crate ethereum_types;
+extern crate heapsize;
 
-use std::collections::HashMap;
 use elastic_array::ElasticArray128;
-use ethereum_types::H256;
+use heapsize::HeapSizeOf;
+use std::collections::HashMap;
+use std::{fmt::Debug, hash::Hash};
+
+/// Trait describing an object that can hash a slice of bytes. Used to abstract 
+/// other types over the hashing algorithm. Defines a single `hash` method and an
+/// `Out` associated type with the necessary bounds.
+pub trait Hasher: Sync + Send {
+	/// The output type of the `Hasher`
+	type Out: AsRef<[u8]> + AsMut<[u8]> + Default + HeapSizeOf + Debug + PartialEq + Eq + Hash + Send + Sync + Clone + Copy;
+	/// What to use to build `HashMap`s with this `Hasher`
+	type StdHasher: Sync + Send + Default + std::hash::Hasher;
+	/// The length in bytes of the `Hasher` output
+	const LENGTH: usize;
+
+	/// Compute the hash of the provided slice of bytes returning the `Out` type of the `Hasher`
+	fn hash(x: &[u8]) -> Self::Out;
+}
 
 /// `HashDB` value type.
 pub type DBValue = ElasticArray128<u8>;
 
-/// Trait modelling datastore keyed by a 32-byte Keccak hash.
+/// Trait modelling datastore keyed by a hash defined by the `Hasher`.
-pub trait HashDB: AsHashDB + Send + Sync {
+pub trait HashDB<H: Hasher>: Send + Sync + AsHashDB<H> {
 	/// Get the keys in the database together with number of underlying references.
-	fn keys(&self) -> HashMap<H256, i32>;
+	fn keys(&self) -> HashMap<H::Out, i32>;
 
 	/// Look up a given hash into the bytes that hash to it, returning None if the
 	/// hash is not known.
-	fn get(&self, key: &H256) -> Option<DBValue>;
+	fn get(&self, key: &H::Out) -> Option<DBValue>;
 
 	/// Check for the existance of a hash-key.
-	fn contains(&self, key: &H256) -> bool;
+	fn contains(&self, key: &H::Out) -> bool;
 
 	/// Insert a datum item into the DB and return the datum's hash for a later lookup. Insertions
 	/// are counted and the equivalent number of `remove()`s must be performed before the data
 	/// is considered dead.
-	fn insert(&mut self, value: &[u8]) -> H256;
+	fn insert(&mut self, value: &[u8]) -> H::Out;
 
-	/// Like `insert()` , except you provide the key and the data is all moved.
+	/// Like `insert()`, except you provide the key and the data is all moved.
-	fn emplace(&mut self, key: H256, value: DBValue);
+	fn emplace(&mut self, key: H::Out, value: DBValue);
 
 	/// Remove a datum previously inserted. Insertions can be "owed" such that the same number of `insert()`s may
 	/// happen without the data being eventually being inserted into the DB. It can be "owed" more than once.
-	fn remove(&mut self, key: &H256);
+	fn remove(&mut self, key: &H::Out);
 }
 
 /// Upcast trait.
-pub trait AsHashDB {
+pub trait AsHashDB<H: Hasher> {
 	/// Perform upcast to HashDB for anything that derives from HashDB.
-	fn as_hashdb(&self) -> &HashDB;
+	fn as_hashdb(&self) -> &HashDB<H>;
 	/// Perform mutable upcast to HashDB for anything that derives from HashDB.
-	fn as_hashdb_mut(&mut self) -> &mut HashDB;
+	fn as_hashdb_mut(&mut self) -> &mut HashDB<H>;
 }
 
-impl<T: HashDB> AsHashDB for T {
+// NOTE: There used to be a `impl<T> AsHashDB for T` but that does not work with generics. See https://stackoverflow.com/questions/48432842/implementing-a-trait-for-reference-and-non-reference-types-causes-conflicting-im
-	fn as_hashdb(&self) -> &HashDB {
+// This means we need concrete impls of AsHashDB in several places, which somewhat defeats the point of the trait.
-		self
+impl<'a, H: Hasher> AsHashDB<H> for &'a mut HashDB<H> {
-	}
+	fn as_hashdb(&self) -> &HashDB<H> { &**self }
-	fn as_hashdb_mut(&mut self) -> &mut HashDB {
+	fn as_hashdb_mut(&mut self) -> &mut HashDB<H> { &mut **self }
-		self
-	}
 }
 
-impl<'a> AsHashDB for &'a mut HashDB {
-	fn as_hashdb(&self) -> &HashDB {
-		&**self
-	}
-
-	fn as_hashdb_mut(&mut self) -> &mut HashDB {
-		&mut **self
-	}
-}