Database
How Blockchain Development Kit's (BDK) internal database is structured and how data is stored in it.
BDK validators use an in-disk database for storing data about themselves and other nodes in the network, such as block/transaction data, contract data, node metadata, etc. Depending on the component (e.g. Storage), they might have their own database folder reserved just for them.
The database itself is an abstraction of a Speedb database - a simple key/value database, but handled in a different way: keys use prefixes, which makes it possible to batch read and write, so we can get around the "simple key/value" limitation and divide data into sectors.
The database requires a filesystem path to open it (if it already exists) or create it on the spot (if it doesn't exist) during construction. It closes itself automatically on destruction.
Content in the database is stored as raw bytes. This is due to space optimizations, as one raw byte equals two UTF-8 characters (e.g. an address like 0x1234567890123456789012345678901234567890, ignoring the "0x" prefix, occupies 20 raw bytes - "12 34 56 ..." - , but 40 bytes if converted to a string, since each byte becomes two separate characters - "1 2 3 4 5 6 ...").
For the main CRUD operations, refer to the has(), get(), put() and del() functions. Due to how the database works internally, updating an entry is the same as inserting a different value in a key that already exists, effectively replacing the value that existed before (e.g. put(oldKey, newValue)). There's also getBatch() and putBatch() for batched operations, as well as getKeys() for fetching only the keys.
Structs and Prefixes
We have three helper structs to ease database manipulation:
DBServer- struct that contains the host and version of the database that will be connected toDBEntry- struct that contains an entry to be inserted or read by the database, and has only two members: key and value, both stringsDBBatch- struct that contains multipleDBEntrys to be inserted and/or deleted all at once
We also have a DBPrefix namespace to reference the database's prefixes in a simpler way:
Those prefixes are concatenated to the start of the key, so an entry that would have, for example, a key named "abc" and a value of "123", if inserted to the "0003" prefix, would be like this inside the database (in raw bytes format, strings here are just for the sake of the explanation): {"0003abc": "123"}
Prefixes Overview
blocks
Used to store serialized blocks based on their hashes.
blockHeightMaps
Used to store block hashes based on their heights.
Padded means that the uint64_t is padded with 0's to the left to ensure a fixed length of 8 bytes.
nativeAccounts
Used to store serialized native accounts ("balance + nonce") based on their addresses.
Serialization for a native account goes like this: requiredBytes(balance) + bytes(balance) + requiredBytes(nonce) + bytes(nonce).
For example, an account with balance 1000000 and nonce 2 would be serialized as 03 + 0f4240 + 01 + 02.
An account with balance 0 and nonce 0 would be serialized as 0000.
txToBlocks
Used to store block hashes, the tx indexes within that block and the block heights, based on their transaction hashes.
Padded means that the uint32_t and uint64_t are padded with 0's to the left to ensure a fixed length of 4 and 8 bytes respectively.
rdPoS
Used to store Validator addresses based on their index within the rdPoS list.
contracts
Used in a multitude of ways, where an "additional" prefix is used per contract (the contract address).
Contracts are free to use the contract prefix as they wish, but the following is the default structure for the contract prefix:
contractManager
Used to store a contract class name based on their address.
events
Used to store events emitted from contracts.
vmStorage
Used to store EVM-related stuff like storage keys and values (essentially the EVM's "permanent memory").
txToAddr
Used to store EVM transactions that created contracts, storing the transaction hash and the address where the contract was deployed.
Last updated