# evm.execution_attempts One row per contextual owner association to one physical attempt. Program-step owners have step_index and no observation_key; observation-call owners have observation_key and no step_index. Reuse may expose the same attempt_id under multiple owners. Canonical HTML: - Kind: View - Layer: execution_read_model - Category: Simulation - Tags: Program, Execution, Public, View ## Contract and lifecycle **Grain:** One contextual attempt owner per (execution_id, attempt_kind, step_index or observation_index). The same attempt_id may expand under multiple owners. **Lifecycle:** run publishes this immutable row set atomically with its owning run. Statement failure or caller rollback publishes no durable prefix; direct DML is unsupported because it can corrupt the documented invariants. **Status and NULLs:** Interpret nullable payloads only through their owning status fields. Defensive poisoned and engine_error outcomes mean the engine refused to claim a trustworthy normal result; broad view nullability also covers owner alternatives and left-joined evidence. ### Join guidance - Join evm.executions on execution_id for candidate and run context. - Use attempt_id to deduplicate physical facts expanded across contextual owners. - For program owners use step_index; for observation-call owners use observation_index and observation_key. ### Effect semantics - Access flags describe attempted EVM access, not committed state change. - Effect facts retain attempted evidence and final committed, reverted, discarded, or poisoned disposition. - execution_state_deltas contains committed checkpoint transitions only; observation-call effects are excluded from execution_effect_summary. ## Columns ### Identity and ownership Contextual keys and reusable identities that establish row ownership. - `execution_id` (UUID; nullable) — Contextual execution association identity; do not use it as reusable artifact identity. - `attempt_kind` (VARCHAR; nullable) — Owner kind: program_step or observation_call. - `observation_key` (VARCHAR; nullable) — Byte-exact caller observation identity for an observation_call owner; otherwise NULL. - `attempt_id` (UUID; nullable) — Reusable physical attempt identity; contextual owners may share this value. ### Chain context Pinned chain, block, profile, and revision context. - `chain_id` (UBIGINT; nullable) — EVM chain identifier frozen for the owning execution. - `chain_profile` (VARCHAR; nullable) — Named chain-rules profile frozen for the owning execution. ### Ordering and cardinality Stable indexes and counts used to reconstruct deterministic order. - `step_index` (UINTEGER; nullable) — Zero-based program order; NULL for observation_call attempt owners. - `observation_index` (UINTEGER; nullable) — Zero-based expanded point order; identifies observation-call ownership. ### Payload and evidence Operation-specific inputs, outputs, and authenticated evidence. - `evm_execution_gas` (UBIGINT; nullable) — Gas consumed by EVM execution when a complete measurement is available. - `return_data` (BYTES; nullable) — Successful return bytes when the attempt produced them; otherwise NULL. - `revert_data` (BYTES; nullable) — EVM revert bytes when the attempt reverted with data; otherwise NULL. - `engine_semantics` (VARCHAR; nullable) — Versioned evaluator-semantics identity used for this result. - `evidence_semantics` (VARCHAR; nullable) — Versioned evidence-publication semantics identity used for this result. ### Status and disposition Outcome vocabulary and NULL-controlling disposition fields. - `evm_status` (VARCHAR; nullable) — Physical attempt outcome: success, revert, or failure. - `evm_error_code` (VARCHAR; nullable) — Stable EVM failure code when evm_status is not success; otherwise NULL. - `fact_completeness` (VARCHAR; nullable) — Whether the selected fact relations are complete for the represented result. ### Commitments Hashes and roots authenticating specifications, facts, state, or results. - `input_state_identity_hash` (BYTES32; nullable) — Identity of the immutable checkpoint supplied to the physical attempt. - `call_message_root` (BYTES32; nullable) — Commitment to the EVM call message evaluated by the attempt. - `attempt_spec_root` (BYTES32; nullable) — Commitment to the input state, call message, chain rules, and attempt semantics. - `trace_root` (BYTES32; nullable) — Commitment to the semantic call-frame trace for the physical attempt. - `attempt_facts_root` (BYTES32; nullable) — Commitment to the exact retained fact relations for the physical attempt. - `attempt_result_root` (BYTES32; nullable) — Commitment to the complete physical attempt result. ### Timestamps Contextual publication times that are excluded from reusable semantic identity. - `created_at` (TIMESTAMP; nullable) — Context association or publication timestamp; not a reusable semantic identity. ## Linked joins - `execution_id` → [evm.executions](/docs/tables/evm-executions.md) (`execution_id`) — Contextual execution that exposes this public row. ## Related public objects - [evm.executions](/docs/tables/evm-executions.md) — upstream ## Verified queries ### Confirm fixture coverage Confirm that this object has rows owned by the stable seven-candidate public fixture. ```sql SELECT 'evm.execution_attempts'::VARCHAR AS object_name, CASE WHEN count(*) > 0 THEN 'PASS' ELSE 'FAIL' END::VARCHAR AS status FROM evm."execution_attempts" o JOIN evm.executions e USING (execution_id) WHERE e.candidate_key IN ('complete-effects', 'later-revert', 'ordered-logs', 'create', 'selfdestruct', 'blockhash', 'transient-access'); ``` ### Check candidate ownership Diagnose whether this object remains linked to at least one stable fixture candidate through public keys. ```sql SELECT 'evm.execution_attempts'::VARCHAR AS object_name, CASE WHEN count(DISTINCT e.candidate_key) > 0 THEN 'PASS' ELSE 'FAIL' END::VARCHAR AS status FROM evm."execution_attempts" o JOIN evm.executions e USING (execution_id) WHERE e.candidate_key IN ('complete-effects', 'later-revert', 'ordered-logs', 'create', 'selfdestruct', 'blockhash', 'transient-access'); ```