Block Matrix
Menu

What an EVM Subgraph Can Index—and Where Its Limits Begin

The first question for an EVM subgraph is what kind of record it needs to capture. A November 4, 2026 overview of The Graph’s indexing capabilities distinguishes emitted events, contract state, function calls and execution traces. They are not interchangeable sources of data.

Events are the starting point. Event handlers are the primary and most reliable way to index activity exposed through contract events. They can also access transaction and block metadata associated with those events.

State reads answer a different question. A mapping can use eth_call to read contract state, but the result reflects state at the block level rather than a complete record of every change within that block. Historical reads require an archive node that supports EIP-1898. A block handler can instead sample state on every block or at a fixed interval; sampling does not capture every intervening change.

Calls are not the same as events or full traces. A call handler can follow contract function calls that emit no event, but it depends on trace_filter, which many networks and RPC providers do not offer. Full execution traces, internal transfers, pending transactions and a chain’s complete transaction history fall outside standard subgraph capabilities. The overview points to debug_traceTransaction, Substreams or a custom indexing pipeline for those needs.

The overview also separates querying historical subgraph entities from reading historical on-chain state: stored indexing results do not, by themselves, provide arbitrary past contract state. It notes isolation limits for IPFS and Arweave file data sources without presenting them as substitutes for chain indexing.

The practical selection order is to check emitted events first, then consider state reads and call or block handlers where needed. Trace tooling or full-chain indexing comes last, when the required records cannot be obtained through those approaches.