SQL impostor

An in-process engine. Syntax is SQLite-ish.

Query: SELECT … FROM … [JOIN …]* [WHERE …] [GROUP BY …] [HAVING …] [ORDER BY …] [LIMIT …] [OFFSET …].
DML: INSERT INTO … [(cols)] VALUES (rows…), UPDATE … SET … [, …] [WHERE …], DELETE FROM … [WHERE …].
DDL: CREATE TABLE [@COLUMNAR] … (columns… [, PRIMARY KEY(col)]), CREATE INDEX name ON table (col), ALTER TABLE … ADD COLUMN …, RENAME COLUMN … TO …, DROP COLUMN …, DROP TABLE [IF EXISTS] ….
Transactions: BEGIN, COMMIT, ROLLBACK.

Projections: *, qualified names, expressions (arithmetic, unary -), optional AS / implicit aliases.
Joins: INNER/JOIN, LEFT JOIN, RIGHT JOIN, CROSS JOIN; ON join condition; table/column aliases.
Subqueries: IN (SELECT …), derived tables (SELECT …) AS alias, including nested IN and inner ORDER BY/LIMIT.
Aggregates: COUNT(*) / COUNT(expr), SUM, AVG - with GROUP BY, HAVING.
Not implemented SELECT DISTINCT, UNION, window functions, CASE.

Comparisons: =, <>, !=, <, >, <=, >=.
Boolean combos: AND, OR, NOT (with precedence).
Sets / ranges: IN (…), NOT IN, BETWEEN, NOT BETWEEN.
String matching: LIKE / NOT LIKE with SQL % and _ wildcards (not full POSIX/regex).
Literals: integers, floats, single-quoted strings (with escaping rules), NULL, TRUE, FALSE.

Declared types: INTEGER/INT, FLOAT/DOUBLE/REAL, BOOL/BOOLEAN, TEXT/STRING.
Primary key required per table.
NOT NULL on column definitions where supported by CREATE/ALTER.
DROP TABLE IF EXISTS.

Unique index on PK; CREATE INDEX builds a non-unique B+ tree on one column.
Planner can use PK or secondary indexes for selective WHERE and for index-backed ORDER BY on PK or indexed columns when the query allows.

WASM playground

./build_wasm.sh

python -m http.server 8000

odin test . -sanitize:address -debug