store API proto AccountProof: optimize merkle node compression
#1178
Conversation
drahnr
changed the title
AccountProof: optimize merkle node compression
There was a problem hiding this comment.
Looking at the complexity of the code, I wonder if this should actually live in miden-base. The main reason is that the client would need to deserialize this data but the client doesn't get anything from the node except for protobuf files.
In miden-base, we could attach the logic to the PartialStorageMap struct. Basically, we need two things there:
- Given a
PartialStorageMapwe need to getSmtLeafs andInnerNodes from it. Not sure what the name of the function would be - but getting this data shouldn't be too difficult. - Given a set of
SmtLeafs andInnerNodes, we need a constructor that would build the underlyingPartialSmtfrom this.
Then, here in miden-node we'll just need to convert these to/from protobuf structs - so, the logic will be pretty straight-forward.
| message AccountStateHeader { | ||
| // Represents a single storage slot with the requested keys and their respective values. | ||
| message StorageSlotMapProof { | ||
| message StorageSlotMapPartialSmt { |
There was a problem hiding this comment.
I think this message could be just PartialStorageMap.
| impl From<NodeIndex> for proto::primitives::NodeIndex { | ||
| fn from(value: NodeIndex) -> Self { | ||
| proto::primitives::NodeIndex { | ||
| depth: value.depth() as u32, | ||
| value: value.value(), | ||
| } | ||
| } | ||
| } | ||
| impl TryFrom<proto::primitives::NodeIndex> for NodeIndex { | ||
| type Error = ConversionError; | ||
| fn try_from(index: proto::primitives::NodeIndex) -> Result<Self, Self::Error> { | ||
| let depth = u8::try_from(index.depth)?; | ||
| let value = index.value; | ||
| Ok(NodeIndex::new(depth, value)?) | ||
| } | ||
| } |
There was a problem hiding this comment.
Probably nothing, but we encode the depth as u32 and then cast it to u8. Should a comment be added about why is it always valid?
Follow up to #1158
Changes
Previously we used the
Serializable/Deserializableimplementations forPartialSmtwhich uses more leaves than the theoretical minimum required to reconstruct all intermediate nodes.The PR implements a DFS per sibling required, calculating all inner nodes dynamically. Recomputation is avoided by using lookup cache per node index.