SQL

July 11, 2026 ยท View on GitHub

MongrelDB ships a DataFusion-backed SQL engine at POST /sql. From Odin, run SQL with sql:

import m "mdb:mongreldb"

rows, err := m.sql(db, "SELECT 1")

This guide covers the SQL surface - DDL, DML, CREATE TABLE AS SELECT, recursive CTEs, and window functions - and when to reach for SQL versus the native query builder.

How sql behaves

sql(db, sql_text) sends {"sql": "...", "format": "json"} to /sql. It returns the decoded row slice and .None_ on a 2xx response.

In practice:

  • DDL and DML (CREATE TABLE, INSERT, UPDATE, DELETE) reply with a non-JSON status body. sql returns .None_ with an empty slice - success is the signal.
  • SELECT in daemon builds that honor the requested JSON format returns a JSON array of row objects keyed by column name, decoded into the returned []JSONValue. In older builds the server streams Arrow IPC bytes rather than JSON; sql detects a non-array body and returns an empty slice rather than a .Json error. Use the native query builder for typed row retrieval in application code, and SQL for statements whose execution is the goal.

Errors are mapped to the same Mongrel_Error values as everything else: an HTTP 400 or 5xx is .Query/.Http; 409 is .Conflict; and so on. See errors.md.

_, err := m.sql(db,
	"INSERT INTO orders (id, customer, amount) VALUES (99, 'Zoe', 999.0)")
if err == .Conflict {
	fmt.eprintln("duplicate row")
}

CREATE TABLE

Define a table in SQL instead of via create_table. Column ids are assigned by the server when not stated.

_, _ = m.sql(db,
	"CREATE TABLE products (" +
	"  id INT64 PRIMARY KEY," +
	"  name VARCHAR," +
	"  price FLOAT64," +
	"  category VARCHAR," +
	"  in_stock BOOLEAN)")

INSERT

_, _ = m.sql(db,
	"INSERT INTO products (id, name, price, category, in_stock) " +
	"VALUES (1, 'Widget', 9.99, 'tools', true)")
_, _ = m.sql(db,
	"INSERT INTO products VALUES (2, 'Gadget', 19.99, 'tools', true)")

For bulk inserts, the native batch transaction (commit) is usually faster because it stages ops in one round trip without re-parsing SQL.

UPDATE

_, _ = m.sql(db, "UPDATE products SET price = 14.99 WHERE id = 1")
_, _ = m.sql(db, "UPDATE orders SET amount = 200.0 WHERE customer = 'Bob'")

DELETE

_, _ = m.sql(db, "DELETE FROM products WHERE in_stock = false")
_, _ = m.sql(db, "DELETE FROM products WHERE id = 2")

SELECT

rows, _ := m.sql(db, "SELECT id, name FROM products WHERE category = 'tools' ORDER BY price")
rows, _ = m.sql(db, "SELECT category, COUNT(*) AS n FROM products GROUP BY category")

Each returned row is a JSONValue object keyed by column name. Read a field with json_object_get:

for row in rows {
	obj, ok := row.(m.JSONObject)
	if !ok do continue
	name_any, has := m.json_object_get(obj, "name")
	if has {
		name, _ := name_any.(m.JSONString)
		fmt.printf("name = %s\n", name)
	}
}

Remember SELECT bodies may arrive as Arrow IPC on older servers, in which case sql returns an empty slice. To read rows back into typed values reliably, mirror the same lookup with the native query builder.

CREATE TABLE AS SELECT

Materialize a query result into a new table. Great for snapshots, rollups, and denormalized aggregates.

// Snapshot all high-value orders into a new table.
_, _ = m.sql(db, "CREATE TABLE archive AS SELECT * FROM orders WHERE amount > 500")

// Roll up sales by customer.
_, _ = m.sql(db,
	"CREATE TABLE sales_by_customer AS " +
	"SELECT customer, SUM(amount) AS total FROM orders GROUP BY customer")

The new table inherits column types from the query. Query it afterward with the native builder or SQL.

Recursive CTEs

WITH RECURSIVE is fully supported. Classic use cases: series generation, hierarchy/graph traversal.

// Generate the numbers 1..10.
_, _ = m.sql(db,
	"WITH RECURSIVE r(n) AS (" +
	"  SELECT 1 UNION ALL SELECT n + 1 FROM r WHERE n < 10" +
	") SELECT n FROM r")

A common practical example is walking an adjacency list:

_, _ = m.sql(db,
	"WITH RECURSIVE descendants(id) AS (" +
	"  SELECT id FROM categories WHERE id = 1" +
	"  UNION ALL" +
	"  SELECT c.id FROM categories c JOIN descendants d ON c.parent_id = d.id" +
	") SELECT id FROM descendants")

Window functions

Window functions compute aggregates/rankings across a moving window without collapsing rows. Useful for top-N-per-group, running totals, and row numbers.

// Row number within each customer, ordered by amount descending.
_, _ = m.sql(db,
	"SELECT id, customer, amount, " +
	"ROW_NUMBER() OVER (PARTITION BY customer ORDER BY amount DESC) AS rn " +
	"FROM orders")

// Running total per customer.
_, _ = m.sql(db,
	"SELECT id, customer, amount, " +
	"SUM(amount) OVER (PARTITION BY customer ORDER BY id) AS running_total " +
	"FROM orders")

RANK(), DENSE_RANK(), LAG(), LEAD(), NTILE(), and the usual window-frame clauses are available through DataFusion.

When to use SQL vs. the query builder

Both read from the same tables, but they are optimized for different jobs.

Reach forWhen
query builderPoint lookups, range scans, bitmap filters, and full-text that map to a native index. Sub-millisecond, no parser overhead, and rows decode into typed values directly.
SQLDDL (CREATE TABLE, schemas, materialized views), multi-statement setup, joins, recursive CTEs, window functions, and arbitrary aggregates. Also the natural choice for admin scripts and one-off analysis.

Rules of thumb:

  • Need typed rows of matching values? Use the query builder.
  • Building/dropping tables, or running a CREATE TABLE AS SELECT? Use SQL.
  • Joining multiple tables, computing rankings, or walking a graph? Use SQL.
  • Filtering by one or more indexed columns? Use the query builder - it is faster and avoids Arrow-to-typed decoding.

Mix freely: create tables with SQL, write rows with put, read them back with the query builder, and run analytics with SQL.

Next steps