[Ethereum] How to read block data from chaindata

Data structures are stored in Merkle Patricia tries (read this and this), usually inside a LevelDB store. The chaindata is exactly that.

I think the structure might slightly depend on the actual node implementation?

Here are two of the trie implementations, could be a good starting point:

• in Go (geth): https://github.com/ethereum/go-ethereum/tree/master/trie
• in Javascript: https://github.com/ethereumjs/merkle-patricia-tree

Once that is understood, you will need to look up the data structures sent over the network, especially blocks, transactions, accounts and storage. Accounts refer to both externally owned and contract accounts, the latter with contract code. Storage refers to contract storage.

Back to the point of data analysis

You might be better off building a traditional NoSQL or SQL database if you want to have quick lookups about any transactions done by an account or contract. What the chaindata is optimised for is validating transactions and maintaining a valid state, and not so for historical lookups.

Blockchains are databases from a high-level view, but the underlying technology is different enough to make some assumptions inaccurate. An Excel file is also a database, but with characteristics unneeded/harmful for say, a NoSQL database.

The data "is" on every full node. That is, every client replicates the entire blockchain. This is indeed inefficient, and future development is aimed towards removing this requirement.

1. The official specification for the state (that is, the set of all contracts and accounts and their storage and balance) is a Merkle Patricia Tree. However, an individual node may implement it internally in whatever way it finds best. (geth uses levelDB for example.)
2. Every full node has its own copy of the blockchain. Reading consists of querying one's own node for the data. There's no guarantee that you have the latest block--an inherent limitation of blockchains. There are several ways for a good dapp to come with this environment--the primary is simply not to assume that the local latest block is the actual latest.