syntaqlite 0.2.13

SQLite SQL tools — parser, formatter, validator, and MCP server

syntaqlite

Parse, format, and validate SQLite SQL from Python using SQLite's own grammar and tokenizer. No approximations: if SQLite accepts it, syntaqlite parses it.

Docs · Playground · GitHub

pip install syntaqlite

Requires Python 3.10+. Wheels for Linux (x86_64, aarch64), macOS (x86_64, arm64), and Windows (x86_64).

Library

The package includes a C extension linked directly against syntaqlite. No subprocesses, no shelling out.

Formatting

import syntaqlite

sql = "select u.id, u.name, p.title from users u join posts p on u.id = p.user_id where u.active = 1 and p.published = true order by p.created_at desc limit 10"
print(syntaqlite.format_sql(sql))

SELECT u.id, u.name, p.title
FROM users AS u
JOIN posts AS p ON u.id = p.user_id
WHERE
  u.active = 1
  AND p.published = true
ORDER BY
  p.created_at DESC
LIMIT 10;

Raises syntaqlite.FormatError on invalid input.

Parsing

syntaqlite.parse() returns a full AST as typed Python objects, one per statement:

import syntaqlite

stmts = syntaqlite.parse("SELECT 1 + 2 FROM foo")
stmt = stmts[0]  # SelectStmt

print(type(stmt).__name__)       # SelectStmt
print(stmt.columns[0].expr)      # BinaryExpr(...)
print(stmt.from_clause)          # TableRef(...)
print(stmt.where_clause)         # None

Every node type is a __slots__ class with typed attributes, so you get IDE autocomplete and isinstance checks:

from syntaqlite.nodes import SelectStmt, BinaryExpr

assert isinstance(stmt, SelectStmt)
assert isinstance(stmt.columns[0].expr, BinaryExpr)

Enum and flag fields are wrapped as IntEnum/IntFlag from syntaqlite.enums:

from syntaqlite.enums import BinaryOp

expr = stmt.columns[0].expr
print(BinaryOp(expr.op).name)  # PLUS

For performance-sensitive code, use syntaqlite.parse_raw() to get plain dicts instead of typed objects:

import syntaqlite, json

stmts = syntaqlite.parse_raw("SELECT 1 + 2; SELECT 3")
print(json.dumps(stmts[0], indent=2))

{
  "type": "SelectStmt",
  "flags": 0,
  "columns": [
    {
      "type": "ResultColumn",
      "flags": 0,
      "alias": null,
      "expr": {
        "type": "BinaryExpr",
        "op": 0,
        "left": { "type": "Literal", "literal_type": 0, "source": "1" },
        "right": { "type": "Literal", "literal_type": 0, "source": "2" }
      }
    }
  ],
  "from_clause": null,
  "where_clause": null,
  ...
}

Error nodes have "type": "Error" with a "message" field.

Tokenizing

import syntaqlite

for tok in syntaqlite.tokenize("SELECT 1 + 2"):
    print(tok["text"], tok["type"])

SELECT 161
  185
1 110
  185
+ 97
  185
2 110

Each token is a dict with text, offset, length, and type fields.

Validation

Check SQL against a schema without touching a database. Catches unknown tables, columns, functions, CTE column mismatches, and more.

import syntaqlite

result = syntaqlite.validate(
    "SELECT nme FROM users",
    tables=[syntaqlite.Table("users", columns=["id", "name", "email"])],
)
for d in result.diagnostics:
    print(f"{d.severity}: {d.message}")

error: unknown column 'nme'

Pass render=True to get formatted diagnostics with source locations and suggestions:

print(syntaqlite.validate("SELECT nme FROM users",
    tables=[syntaqlite.Table("users", columns=["id", "name", "email"])],
    render=True))

error: unknown column 'nme'
 --> <input>:1:8
  |
1 | SELECT nme FROM users
  |        ^~~
  = help: did you mean 'name'?

Schema can come from syntaqlite.Table/syntaqlite.View objects or raw DDL:

result = syntaqlite.validate(
    "SELECT * FROM orders",
    schema_ddl="CREATE TABLE orders (id INTEGER, total REAL);",
)

Column lineage

For SELECT statements, validation results include column lineage tracing each output column back to its source:

import syntaqlite

result = syntaqlite.validate(
    "SELECT id, name FROM users",
    tables=[syntaqlite.Table("users", columns=["id", "name", "email"])],
)
for col in result.lineage.columns:
    print(f"{col.name} <- {col.origin}")

id <- users.id
name <- users.name

CLI

The pip package also bundles the syntaqlite binary:

syntaqlite fmt -e "select 1, 2, 3"
syntaqlite validate query.sql
syntaqlite parse -e "SELECT * FROM users"

The CLI supports pinning to a specific SQLite version or enabling compile-time flags to match your target environment:

syntaqlite --sqlite-version 3.32.0 validate query.sql
syntaqlite --sqlite-cflag SQLITE_ENABLE_MATH_FUNCTIONS validate query.sql

See the CLI reference for all commands and flags.

License

Apache 2.0. SQLite components are public domain under the SQLite blessing.