rbtr-lang-sql

SQL support for rbtr. Optional plugin — install with pip install rbtr[sql].

What it ingests

One chunk per top-level SQL statement (plus one per CTE). SQL has no classes or functions in the object-oriented sense and no native import mechanism, so statements map onto rbtr's capture conventions by shape: structural definitions become classes, routines and executable statements become functions, and standalone named objects become variables.

• Structural definitions — CREATE TABLE, CREATE VIEW / MATERIALIZED VIEW, CREATE TYPE ... AS ENUM / composite types.
• Routines & statements — CREATE FUNCTION, SELECT, INSERT, UPDATE, DELETE, WITH ... AS (CTEs), ALTER / DROP.
• Named objects — CREATE INDEX / SEQUENCE / SCHEMA / ROLE / TRIGGER.

CREATE PROCEDURE and PRAGMA are not extracted — the generic grammar has no node for them.

Chunks produced

name is the statement's target object; scope is always empty (SQL has no nesting). A statement with no nameable target (SELECT 1, a UNION) carries no name.

CREATE TABLE users (id INT, name TEXT);   -- class    "users"
CREATE VIEW active AS SELECT * FROM users; -- class    "active"
CREATE TYPE mood AS ENUM ('sad','happy');  -- class    "mood"
CREATE FUNCTION add(a INT) ...             -- function "add"
SELECT id, name FROM users;                -- function "users"
INSERT INTO logs (msg) VALUES ('hi');      -- function "logs"
WITH ranked AS (...) SELECT ...            -- function "ranked" (per CTE)
CREATE INDEX idx_name ON users (name);     -- variable "idx_name"
CREATE SEQUENCE order_id START 1;          -- variable "order_id"

Embedded / injected chunks

None. SQL does not embed other languages.

Grammar & dependencies

Uses the tree-sitter-sql grammar (one generic grammar for all dialects). No dependency on other language plugins.