Proof Calc Rewrite: Leaf-only Implementation

[mediocregopher]

trafficstars

Describe the feature

Implementation of base scaffolding and types which will be iterated upon in subsequent tasks. To start with we will only use leaf data (HashedAccounts/Storages) to generate proofs, leaving room to later introduce cached branch nodes from Accounts/StoragesTrie tables.

Requirements:

• Given one or more B256 proof targets, sorted lexicographically, the new proof calculator should return the proof nodes for all proof targets as a Vec, sorted lexicographically by proof node paths.

• Given zero proof targets, the proof calculator should return just the root.

• Once a proof is calculated, the calculator should automatically "reset" itself, such that it is ready for another calculation with new proof targets.

• The calculator should re-use cursors from one calculation to the next.

• The value type should be a generic. The generic should have a method which returns an FnOnce. The FnOnce should synchronously return the RlpNode for the value.

  • For storage proofs this will simply return the RlpNode of the slot value.
  • For account proofs the default implementation will synchronously call a storage proof for the account with empty targets.
  • This will leave room for injecting parallelization and caching of storage roots in later iterations.
  • Calling of the FnOnce should be done as lazily as possible in the algorithm, to maximize the time that storage proof workers have to calculate the root.

• Implement a proptest which generates random sets of accounts, storage slots, and proof targets. The proptest should compare the results of a proof calculated against a mock cursor factory using the legacy proof implementation vs a proof calculated against that same factory with the new implementation.

Additional context

No response