vectorized-execution

Learn advanced Rust by building a type-safe, auto-vectorized database expression engine

[SIGMOD'25] We show the data chunk compaction problem in vectorized execution, and propose practical compaction solutions.

Production-grade genpark-columnar-storage-vectorized-execution-skill skill for AI agents

Production-grade genpark-columnar-storage-vectorized-execution-skill skill for AI agents

A high-performance, distributed SQL query engine built in Rust. Powered by Apache Arrow for vectorized execution, it supports complex SQL (JOINs, CTEs, Window Functions), distributed processing, and Arrow Flight.

Columnar query engine built from scratch in Python , Parquet row group pruning, rule-based optimizer, vectorized operators, and a hash join that spills to disk. Within 1.2x of DuckDB on statistics-pruned filters.

C++20 relational executor comparing iterator, vectorized, and code-generated pipelines on identical plans.

In-process columnar analytics engine: Float64Array storage, vectorized GROUP BY, 1M rows in <10ms

Distributed SQL analytics engine for Go: embed it in a Go application, or deploy it as a fault-tolerant query cluster with durable S3 exchange. Vectorized columnar OLAP over Parquet and Apache Iceberg; PostgreSQL wire protocol; no JVM, no CGo. Built for network telemetry: first-class IPv4/IPv6/CIDR/MAC/Port/Protocol types, 100+ network functions.

A columnar SQL query engine written from the wire up in Rust — hand-written parser, cost-based optimiser, vectorised execution, and its own on-disk format with zone-map pruning.

A SQL engine that moves typed column batches through vector operators and chooses physical plans by estimated cost

Lamina SQL: vectorized analytical SQL engine from scratch in Rust -- parsing, query planning, predicate pushdown, columnar execution, hash joins, benchmarked on TPC-H-shaped workloads

We integrate data chunk compaction methods into duckdb, for an end-to-end benchmark.

A scalable single-GPU engine for subgraph query processing on large labeled graphs, featuring vectorized execution and workload decomposition.

C++17 vector-at-a-time analytical engine with null bitmaps, selection vectors, hash aggregation, joins, and scalar differential tests

A relational database engine written from scratch in C++17 — B+ tree indexing, buffer pool, Volcano-model executor, rule-based optimizer, Strict 2PL, and WAL recovery in ~2,000 lines with zero dependencies. Vectorized execution gives 10.4× scan throughput.

Vectorized executor to speed up PostgreSQL

Basalt — a from-scratch hybrid OLTP/OLAP (HTAP) database engine in C++17, columnar by nature, with a browser Workbench and WASM playground; benchmarked vs SQLite, DuckDB, and MySQL

Low-level systems engineering analysis of DuckDB - query execution path, vectorized execution, benchmarking, and failure analysis.

A vectorized in-memory analytical query engine in C++20 - columnar storage, rule-based optimizer with predicate/projection pushdown, hash joins. Measured, not asserted.