This packages implements Merkle Tree Logs similar to those described in RFC 6962 in Haskell.
Merkle Logs are a append-only data structure. The tree layout in this implementation of Merkle trees is based on the description of Merkle trees in RFC 6962. With this tree layout extending a Merkle tree requires chaining a logarithmic number of nodes at the end of the tree. Unlike RFC 6962 the Merkle trees in this module support the creation of unbalanced MerkleTrees by nesting sub-trees as leafs of Merkle trees. Also, unlike RFC 6962 this module generates fully self-contained inclusion proofs that don't rely on the client being aware of the balancing of the Merkle Tree that was used to generate the proof.
The implementation stores Merkle trees in a packed format in memory. This allows for efficient construction, serialization, and querying. Trees are opaque objects that are allocated and deallocated as well as serialized and deserialized as a whole, which matches many use cases. Also, trees can be nested, by building larger Merkle trees that have smaller trees as inputs to their leafs.
The overhead per indexed item is 64 bytes when 256 bit hashes are used. Thus, about 16,000 items can be index in 1MB of memory.
We plan to make the trees extensible and support loading and storing trees in chunks that represent immutable full subtrees. Please file an issue on GitHub if you need this feature.
Proofs are self contained and don't rely on a particular implementation of Merkle tree. In particular, proofs don't depend on how the tree is balanced.
A proof contains the proof subject (the input for which inclusion is proven) as
a plain ByteString. The result of validating a proof is a Merkle tree root
that must match the root of the Merkle tree that includes the subject. A proof
doesn't include the root hash of the Merkle tree, because the root must be
obtained from a trusted / authenticated source. Including it in the proof would
thus be redundant and may even be misleading.
At the moment only inclusion / audit proofs are supported. We plan to also implement consistency proofs. Please file an issue on GitHub if you need consistency proofs.
The package can be build with cabal via
cabal new-update
cabal new-build merkle-logThe test suite can be run with
cabal new-test merkle-logBenchmarks are available via
cabal new-bench merkle-logThe following benchmark results compare the performance of this package with the Merkle tree implementations in the packages merkle-tree and hash-tree for different hash functions on a 64bit Intel CPU.
Create tree:
Create inclusion proof:
Verify inclusion proof:
