$ show pawurb/mevlog-rs
Mevlog-rs is a Rust-based, open-source alternative to commercial blockchain explorers like Dune Analytics, providing a comprehensive toolchain for indexing, storing, and querying EVM-compatible blockchain data. The core system downloads raw chain data from RPC endpoints, stores it in per-chain SQLite databases, and allows read-only SQL queries enriched with custom U256 arithmetic, formatting, and ENS resolution. It includes a CLI with 20+ subcommands, a web backend (Axum + React) for hosted exploration, a TUI (ratatui) for interactive browsing, and an MCP server for AI agent integration. The system is heavily oriented toward MEV (maximal extractable value) analysis, with coinbase transfer detection, state diffs, and call tracing via local revm replay or RPC. Key architectural patterns include a shared dependency initializer, a secure read-only SQL sandbox with progress limits, a config-driven custom table engine, and a reorg-aware indexing pipeline that handles chain reorganizations. While the codebase demonstrates strong craftsmanship and modularity, its large dependency surface, reliance on external services (ChainList, Pinata, cryo), and some commented-out tests indicate ongoing evolution and moderate maintenance risk.
The LLM identified 5 files worth highlighting for reuse:
Implements a secure, read-only SQL query execution layer using rusqlite with an authorizer callback that blocks mutations, ATTACH, and dangerous PRAGMAs. It enforces a maximum row count, a progress timeout, duplicate column detection, and registers custom U256 SQL functions. This pattern is directly reusable in any application that needs to expose raw SQL to users without risking the database.
Manages a parquet-based cache layer for blockchain data fetched via the cryo tool. It computes coverage gaps by analyzing parquet file ranges, batches missing ranges for efficient fetching, and purges stale files after reorgs. The gap-detection and batch-fetch pattern is reusable for any project needing incremental data synchronization with a local cache.
Provides an abstraction over IPFS upload methods (Pinata API or local Kubo daemon) with environment-variable-based configuration and gateway URL resolution. It handles both simple file upload and result content hash retrieval. This pattern can be extracted for any project that needs decentralized storage with multiple backends.
Implements a config-driven custom table engine where users define derived tables from log events (topic0 + column mappings) in a TOML config file. The engine syncs table definitions with the database using fingerprints, performs atomic create-and-backfill, incremental range population, and rebuilds on config changes. This pattern is reusable for any application that needs user-defined materialized views on top of event data.
Detects blockchain reorganizations by comparing stored block hashes with live chain data, finds the fork point via parent hash chain traversal, and rolls back stale blocks/transactions/logs from SQLite. It also purges overlapping cryo parquet cache. The rollback logic is database-agnostic and could be adapted for any system that needs to revert to a consistent state after detecting a fork.