Duneanalytics SQL queries

Ethereum’s ethereum.transactions contains one row per transaction. Logs and traces describe different parts of execution, so counting their rows doesn’t produce the same metric as counting transactions.

Decoded event tables expose readable contract fields. Dune’s Event Logs documentation identifies uniswap_v3_ethereum.Factory_evt_PoolCreated as the Ethereum table for Uniswap V3 pool-creation events. The related uniswap_v3_ethereum.Factory_call_createPool table describes calls to the function. Use the event table for emitted events and the call table for recorded invocations; their counts aren’t interchangeable.

Decoded tables can collect multiple instances of a contract type. The contract_address field distinguishes the instance emitting an event or receiving a call. A protocol namespace alone doesn’t necessarily restrict the result to one pool or contract.

Call tables describe onchain execution. A local read-only function call that isn’t broadcast doesn’t appear as an onchain call record. The same function can be recorded when another contract invokes it within a transaction.

Dune’s curated datasets normalize activity into consistent schemas. For DEX analysis, dex.trades covers supported EVM activity, while dex_solana.trades provides Solana DEX data. A table’s cross-chain scope doesn’t imply identical protocol coverage on every network.

The dex.trades fields blockchain , project and version describe the network and DEX implementation. Each recorded pool segment is a separate row, so a transaction routed through several pools can contribute several rows.

Amount columns also carry different meanings. In dex.trades , token_bought_amount_raw is the quantity before decimal adjustment, token_bought_amount is in token display units and amount_usd is the trade’s USD value at execution. Those columns aren’t interchangeable measures of volume.

Pool Trade Rows Versus Transactions

A reader wants the number of Ethereum transactions containing a recorded DEX trade. Assume a hypothetical 13-day sample with 83 pool trade rows and 53 distinct, non-null transaction hashes. The period and counts are illustrative inputs, not measured activity.

View full-size image

SELECT
  COUNT(*) AS pool_trade_rows,
  COUNT(DISTINCT tx_hash) AS transactions_with_trades,
  COUNT(*) - COUNT(DISTINCT tx_hash) AS additional_rows
FROM dex.trades
WHERE blockchain = 'ethereum'
  AND block_time >= NOW() - INTERVAL '13' DAY
  AND block_time < NOW();

For that hypothetical sample, the results are 83 pool trade rows, 53 transactions with recorded trades and 30 additional rows: 83 - 53 = 30 . Reporting the row count as the transaction count would overstate transactions by 30. Another recorded pool segment within an existing transaction increases the row count without increasing the distinct transaction count. Both counts remain limited to the selected table, network and period.

Dune’s data explorer provides table schemas and sample data. Confirm the available columns, contract coverage and historical range before relying on an aggregation. A missing decoded event or unsupported protocol version can’t establish that no onchain activity occurred.

Time filters and a blockchain filter on applicable cross-chain tables help reduce the data scanned. Use the table’s actual partition fields and select the columns the analysis needs. When joining event or trade rows to transaction records, account for multiple matching rows before summing transaction-level values; otherwise the join can repeat those values in the total.