Downloads 317this week

GitHub stars 0

Extension repository on GitHub

Extension descriptor (YAML)

Storage extension wrapping LevelDB with pivot semantics for structured relational tables over key-value pairs

Installing and Loading

INSTALL level_pivot FROM community;
LOAD level_pivot;

Example

-- Attach a LevelDB database
ATTACH '/tmp/mydb' AS kv (TYPE level_pivot, CREATE_IF_MISSING true);

-- Create a pivot table: rows decompose into LevelDB key-value pairs
-- Key pattern: users##<group>##<id>##<attribute_name> → attribute_value
CALL level_pivot_create_table('kv', 'users', 'users##{group}##{id}##{attr}', ['group', 'id', 'name', 'email']);

-- Insert data
INSERT INTO kv.users VALUES ('admins', 'u1', 'Alice', '[email protected]');
INSERT INTO kv.users VALUES ('admins', 'u2', 'Bob', '[email protected]');

-- Query with standard SQL (filter pushdown on identity columns)
SELECT * FROM kv.users;
SELECT * FROM kv.users WHERE "group" = 'admins' AND id = 'u1';

About level_pivot

LevelPivot is a DuckDB storage extension that wraps LevelDB with pivot semantics. It maps relational rows to LevelDB key-value pairs using configurable key patterns, enabling structured multi-column SQL tables on top of a key-value store.

Tables are created via level_pivot_create_table() with a key pattern like 'users##{group}##{id}##{attr}' that defines how rows decompose into LevelDB entries. Identity columns ({group}, {id}) form the key prefix; attribute columns (name, email) each become a separate LevelDB key-value pair.

Key Features:

• Pivot mode: decomposes rows into per-attribute LevelDB entries
• Raw mode: simple 2-column key-value access
• Configurable key patterns with identity columns and delimiters
• Typed columns: VARCHAR, BIGINT, INTEGER, DOUBLE, BOOLEAN
• Filter pushdown on identity columns for efficient prefix scans
• Full DML support: INSERT, INSERT INTO…SELECT, UPDATE, DELETE
• SIMD-optimized key parsing (AVX2, SSE2, NEON)
• Data persistence across DETACH/ATTACH cycles

Added Functions