[][src]Crate exonum_merkledb

Interfaces to work with persisted blockchain data. The data can be Merkelized, i.e., stored into authenticated data structures, which allow to prove presence or absence of data with logarithmic overhead.

See also the documentation page on storage.


A Database is a container for data persistence. Internally, a Database is a collection of named key-value stores (aka column families) with reading isolation and atomic writes. The database is assumed to be embedded, that is, the Exonum process has exclusive access to the DB during blockchain operation. You can interact with the Database from multiple threads by cloning its instance.

This crate provides two database types: RocksDB and TemporaryDB.

Snapshot and Fork

Snapshots and forks facilitate access to the database.

If you need to read the data, you can create a Snapshot using the snapshot method of the Database instance. Snapshots provide read isolation, so you are guaranteed to work with consistent values even if the data in the database changes between reads. Snapshot provides all the necessary methods for reading data from the database, so &Snapshot is used as a storage view for creating a read-only representation of the indexes.

If you need to make changes to the database, you need to create a Fork using the fork method of the Database. Like Snapshot, Fork provides read isolation, but also allows creating a sequence of changes to the database that are specified as a Patch. A patch can be atomically merged into a database. Different threads may call merge concurrently.

BinaryKey and BinaryValue traits

If you need to use your own data types as keys or values in the storage, you need to implement the BinaryKey or BinaryValue traits respectively. These traits have already been implemented for most standard types.


Indexes are structures representing data collections stored in the database. This concept is similar to tables in relational databases. The interfaces of the indexes are similar to ordinary collections (like arrays, maps and sets).

Each index occupies a certain set of keys in a single column family of the Database. On the other hand, multiple indexes can be stored in the same column family, provided that their key spaces do not intersect. Isolation is commonly achieved with the help of Groups or keyed IndexAddresses.

Merkelized indexes can generate cryptographic proofs about inclusion of entries. Having such a proof, an external client may verify locally that the received data was authorized by the blockchain validators, without having to replicate the entire blockchain contents.

This crate provides the following index types:

  • Entry is a specific index that stores only one value. Useful for global values, such as configuration. Similar to a combination of Box and Option.
  • ListIndex is a list of items stored in a sequential order. Similar to Vec.
  • SparseListIndex is a list of items stored in a sequential order. Similar to ListIndex, but may contain indexes without elements.
  • MapIndex is a map of keys and values. Similar to BTreeMap.
  • ProofEntry is a Merkelized version of Entry.
  • ProofListIndex is a Merkelized version of ListIndex that supports cryptographic proofs of existence and is implemented as a Merkle tree.
  • ProofMapIndex is a Merkelized version of MapIndex that supports cryptographic proofs of existence and is implemented as a binary Merkle Patricia tree.
  • KeySetIndex and ValueSetIndex are sets of items, similar to BTreeSet and HashSet accordingly.

State aggregation

The database automatically aggregates its contents into a single state_hash, which commits to the entire Merkelized database contents. For example, this is used in Exonum to achieve consensus as to the database state.

The state_hash of the database is the hash of state_aggregator, a system ProofMapIndex with keys being UTF-8 names of aggregated indexes, and values their hashes as per ObjectHash implementation. An index is aggregated if and only if it satisfies the following constraints:

  • Index has a matching type (ProofListIndex, ProofMapIndex, or ProofEntry)
  • Index is not a part of a group, i.e., its address does not contain the bytes part

The aggregation is automatically updated when a Fork is converted into a Patch. Thus, Snapshots (including Patches!) are always consistent with respect to the aggregated state; the index hashes in the state_aggregator match their actual values. This is not the case for Forks, in which state_aggregator may be stale.


The database provides tooling for data migrations. With the help of migration, it is possible to gradually accumulate changes to a set of indexes (including across process restarts) and then atomically apply or discard these changes.


pub use self::indexes::proof_list;
pub use self::indexes::proof_list::ListProof;
pub use self::indexes::proof_list::ProofListIndex;
pub use self::indexes::proof_map;
pub use self::indexes::proof_map::MapProof;
pub use self::indexes::proof_map::ProofMapIndex;
pub use self::indexes::proof_map::RawProofMapIndex;
pub use self::indexes::Entry;
pub use self::indexes::Group;
pub use self::indexes::KeySetIndex;
pub use self::indexes::ListIndex;
pub use self::indexes::MapIndex;
pub use self::indexes::ProofEntry;
pub use self::indexes::SparseListIndex;
pub use self::indexes::ValueSetIndex;



High-level access to database.


Access generalizations, mainly useful for bindings.


All available MerkleDB indexes.


Migration utilities.


Validation helpers for index names.



Implements BinaryKey trait for any type that implements BinaryValue.


Implement ObjectHash trait for any type that implements BinaryValue.


Hex conversions for the given BinaryValue.



Options for the database.


The error type for I/O operations with the Database.


A combination of a database snapshot and changes on top of it.


Represents the address of an index in the database.


Lazily initialized object in the database.


Version of ReadonlyFork with a static lifetime. Can be produced from an Rc<Fork> using the AsReadonly trait.


A set of changes that can be atomically applied to a Database.


Readonly wrapper for a Fork.


Resolved address of a view.


Database implementation on top of RocksDB backend.


System-wide information about the database.


A wrapper over the RocksDB backend which stores data in the temporary directory using the tempfile crate.



Prefixes for different types of objects stored in the database. These prefixes are necessary to provide domain separation among hashed objects of different types.


Type of an index supported by Exonum.


Errors that can occur while validating a ListProof or MapProof against a trusted collection hash.



Converts index access to a readonly presentation. The conversion operation is cheap.


A type that can be (de)serialized as a key in the blockchain storage.


A type that can be (de)serialized as a value in the blockchain storage.


Low-level storage backend implementing a collection of named key-value stores (aka column families).


Extension trait for Database.


A trait that defines a streaming iterator over storage view entries. Unlike the standard Iterator trait, Iterator in Exonum is low-level and, therefore, operates with bytes.


A common trait for the ability to compute a unique hash.


A read-only snapshot of a storage backend.



Computes a Merkle root hash for a the given list of hashes.

Type Definitions


A generalized iterator over the storage views.


A specialized Result type for I/O operations with storage.