Downloads n/athis week

GitHub stars 0

Extension repository on GitHub

Extension descriptor (YAML)

Analytics-optimized JSON storage — parse one time and read many times, and shred hot paths into typed columns for pushdown and row-group pruning

Installing and Loading

INSTALL jsono FROM community;
LOAD jsono;

Example

INSTALL jsono FROM community;
LOAD jsono;

-- A JSON document becomes a pre-parsed binary value. Queries navigate the
-- binary structure and do not parse the text again.
SELECT jsono_transform(
    jsono('{"id": 42, "name": "duck", "active": true}'),
    {id: 'BIGINT', name: 'VARCHAR', active: 'BOOLEAN'}
);
-- {'id': 42, 'name': duck, 'active': true}

-- With a shredding spec, the hot paths go into typed lane columns.
-- The conversion is lossless: to_json gives the initial document back.
SELECT to_json(jsono('{"kind":"commit","time_us":1700,"extra":"e1"}',
                     shredding := '{"$.kind": "VARCHAR", "$.time_us": "BIGINT"}'));
-- {"extra":"e1","kind":"commit","time_us":1700}

About jsono

jsono is analytics-optimized storage for JSON in DuckDB.

jsono(payload) converts a JSON document into a pre-parsed binary value. Queries navigate that binary structure and do not parse the text again. The value is a plain nested STRUCT, so it goes through Parquet and DuckLake with no custom type.

With shredding := spec, the same constructor also moves the hot paths (the paths that queries read most) into typed lane columns. An extraction or a filter on a shredded path becomes a direct read of a narrow typed column, which the planner pushes down and which prunes Parquet row groups. The conversion is lossless: ->>, to_json, and the casts give the same answers as on a plain value, and a path with no lane stays in the residual.

Functions

• Build and convert: jsono(json) / try_jsono(json), jsono(struct), to_json(value).
• Extract and project: jsono_extract (->), jsono_extract_string (->>), jsono_transform (many fields into a typed STRUCT, in one pass), jsono_entries, jsono_array_elements.
• Merge and aggregate: jsono_merge_patch (RFC 7396), jsono_group_merge, jsono_group_merge_max / jsono_group_merge_min, jsono_diff, jsono_group_array / jsono_group_object.
• Shredded storage: jsono(value, shredding := spec), jsono_suggest_shredding (a spec from a plain column), jsono_shred_stats.
• Introspection of the document (jsono_type, jsono_keys, jsono_array_length), of the row (jsono_validate, jsono_storage_size, jsono_shred_manifest), and of the type (jsono_layout_diagnose, jsono_layout_lanes, jsono_storage_type).

Stability. Stored values stay readable across upgrades: a newer release reads the values that an older release wrote. The SQL API is not frozen — function names and signatures can change while the version is 0.x.

Maintenance is best-effort, by one maintainer. Give feedback and contributions through GitHub.

Full documentation, including the storage format and the pruning conditions: https://github.com/Flamefork/duckdb-jsono#readme

Added Functions

Overloaded Functions

Added Types

This extension does not add any types.

Added Settings

This extension does not add any settings.