finds.dev← search

// the find

alibaba/AliSQL

★ 5,819 · C++ · NOASSERTION · updated May 2026

AliSQL is a MySQL branch originated from Alibaba Group. Fetch document from Release Notes at bottom.

AliSQL is Alibaba's MySQL 8.0.44 fork that embeds DuckDB as a pluggable storage engine and adds native HNSW vector search — both features accessible via standard MySQL SQL syntax. Open-sourced in December 2025, it targets teams running OLTP on MySQL who want analytics or AI workloads on the same database without adding a separate system. If you're already on MySQL and reaching for ClickHouse or pgvector, this is the direct competitor.

- DuckDB-as-a-storage-engine is the right architectural call: you get columnar execution inside MySQL without ETL pipelines or a separate process. CREATE TABLE ... ENGINE=DuckDB is all it takes to route a table to the columnar path.

- Vector index (VIDX) supports up to 16,383 dimensions with HNSW and exposes COSINE_DISTANCE/EUCLIDEAN_DISTANCE as SQL functions — the same pattern pgvector uses, so the mental model is familiar and the SQL is portable.

- Based on MySQL 8.0.44 LTS, not some ancient 5.7 fork. Alibaba is likely to track upstream security patches since they run this in production at scale.

- Production provenance matters for a database fork. This has reportedly run millions of instances in Alibaba Cloud — that's a different confidence level than a research prototype.

- The '200x faster analytics' claim has no methodology attached. DuckDB will beat InnoDB on aggregation — that's expected — but the actual number depends entirely on query shape, selectivity, and whether data lives in DuckDB or InnoDB tables. Benchmark without workload description is noise.

- Just open-sourced in December 2025. No official binary packages, no Docker image in the repo, build-from-source only for a C++ MySQL fork. The build script works but standing up a test instance is a half-day job on a good day.

- DDL optimization, crash recovery (RTO), and replication improvements are all marked Planned for 2026 — which means the features that differentiate a MySQL fork for operational teams aren't there yet. You're adopting ahead of the feature set.

- Data synchronization between InnoDB and DuckDB tables is not explained anywhere in the README or linked docs. If you have a DuckDB analytical table, how do you keep it in sync with the InnoDB OLTP table? Manual INSERT? Triggers? This is the core operational question and it's completely unanswered.

View on GitHub →

// want more like this?

We dig through GitHub every week and send a few repos picked for what you actually care about — each with an honest take like this one.

Get finds in your inbox → Search again →