datalevin

Huahai 2026-02-01T22:35:42.693109Z

Update: I am currently working on adding document database features: automatically indexing structural documents by paths, supporting edn, json and markdown values. This feature is only available for Datalog DB, with an additional data type :db.type/idoc , "idoc" means "indexed document". Two new built-in query functions idoc-get to get a value by path, and idoc-match to search documents by path expression.

10
🚀 1
Anton Shastun 2026-02-03T09:13:27.597309Z

Sounds cool!

Huahai 2026-02-04T00:22:11.139899Z

It is now released in 0.10.4, give it a try!

2026-02-02T10:05:01.686609Z

Is it principally a matter of ergonomics - for example, convenience of storing and retrieving ordered lists - or does it also unlock some efficiencies?

Huahai 2026-02-02T15:31:51.628379Z

It is more of a new capability: automatically indexing structured document by paths, so it unlocks a lot of functionalities. The main benefit of document databases is their flexibility, i.e. no need to define a schema, just drop in your documents, the DB breaks them down and indexes the pieces for you, so you can update/retrieve/search the values easily. No schema migration to worry about.

Huahai 2026-02-02T15:33:38.696369Z

The implementation of document DB of course requires some new indices, mainly inverted index of path -> doc id.

Huahai 2026-02-02T15:36:05.467479Z

idoc index complements full text and vector index, to allow one to store and retrieve content in new way.