Convert JSON to SQL Test Fixture INSERT Statements

A two-row user fixture for a Postgres test DB. Booleans become BOOLEAN, the apostrophe in the name is escaped — ready to pass to db.exec() in beforeEach.

Input (fixtures/users.json):
[
  {"id": 1, "email": "ada@x.com", "is_admin": true},
  {"id": 2, "email": "o'neil@x.com", "is_admin": false}
]

Output (PostgreSQL):
CREATE TABLE IF NOT EXISTS "users" (
  "id" INTEGER,
  "email" TEXT,
  "is_admin" BOOLEAN
);

INSERT INTO "users" ("id", "email", "is_admin")
VALUES
  (1, 'ada@x.com', TRUE),
  (2, 'o''neil@x.com', FALSE);

Loading the same fixture before every test would fail on duplicate keys. Enable the conflict toggle so re-seeds skip existing rows instead of erroring.

Config: ON CONFLICT DO NOTHING enabled (PostgreSQL)

Output:
INSERT INTO "roles" ("id", "name")
VALUES
  (1, 'admin'),
  (2, 'user') ON CONFLICT DO NOTHING;

MySQL equivalent (same toggle, MySQL dialect):
INSERT IGNORE INTO `roles` (`id`, `name`)
VALUES
  (1, 'admin'),
  (2, 'user');

Test only cares about email and status. Omitting other fields in some rows yields NULL — no need to specify every column in every fixture row.

Input:
[
  {"email": "a@x.com", "status": "active", "plan": "pro"},
  {"email": "b@x.com", "status": "active"}
]

Output (PostgreSQL):
CREATE TABLE IF NOT EXISTS "users" (
  "email" TEXT,
  "status" TEXT,
  "plan" TEXT
);

INSERT INTO "users" ("email", "status", "plan")
VALUES
  ('a@x.com', 'active', 'pro'),
  ('b@x.com', 'active', NULL);

The database fills id and created_at itself. Exclude them so the INSERT only sets fixture-controlled columns and doesn't fight the schema defaults.

Input:
[{"id": 1, "email": "a@x.com", "created_at": "2026-01-01"}]

Config: Exclude columns = id, created_at, dialect = SQLite

Output (SQLite):
CREATE TABLE IF NOT EXISTS "users" (
  "email" TEXT
);

INSERT INTO "users" ("email")
VALUES
  ('a@x.com');

-- id (AUTOINCREMENT) and created_at (DEFAULT now) are left
-- to the database.

When migrations already created the tables with real constraints, turn off Include CREATE TABLE so you load only the fixture rows.

Config: Include CREATE TABLE = off, dialect = MySQL

Output:
INSERT INTO `products` (`sku`, `price`)
VALUES
  ('A1', 9.99),
  ('B2', 14.5);

-- No CREATE TABLE: your migrated schema, keys, and
-- foreign keys are untouched. Wrap in a transaction in
-- your test and roll back in teardown.

The whole fixture array goes into the single table named in Table name. There's no multi-table fixture support in one pass. For a fixture spanning users + orders + line_items, split the JSON by table, run the converter once per table, and concatenate the outputs in foreign-key dependency order before loading.

The tool only generates inserts (and optionally CREATE TABLE). It never emits TRUNCATE, DELETE, or DROP. Handle teardown in your test framework — a transaction-per-test with rollback, or an explicit afterEach truncate — to keep tests isolated.

ON CONFLICT DO NOTHING / INSERT IGNORE skips conflicting rows; it does not update them. If a test expects a fixture row to overwrite an existing one, this won't do it — there's no upsert. Truncate first, or hand-edit the SQL into an upsert form for your DB.

A fixture field holding an object or array is JSON-stringified into a TEXT column, not separate columns. If your assertion reads row.settings.theme, store the JSON and parse it in the test, or flatten the fixture with json-flattener so settings_theme is its own column.

If you exclude an auto-increment id so the DB assigns it, related fixtures that reference that id by a hardcoded value may not match the DB-assigned one. Either keep explicit ids in the fixtures (don't exclude id), or insert and capture generated ids in the test before seeding dependents.

Postgres/SQLite emit BOOLEAN with TRUE/FALSE; MySQL emits TINYINT(1) with 1/0. If a test asserts the literal value, match it to the dialect you generated. Switching dialects changes both the column type and the boolean literal.

Only .json and spreadsheet files (.xlsx/.xls/.ods) become rows. A .csv fixture fails JSON.parse. Convert it with csv-to-json first, then generate the fixture SQL here.

An empty [] produces -- Empty array, no statements generated rather than an error. If your fixture loader runs that, it's a no-op — useful as a deliberately-empty fixture, but check the stats row count if you expected data.

Single quotes in fixture values are escaped by doubling ('o''neil@x.com'), so realistic names and emails with apostrophes load cleanly. No backslash escaping is used, matching standard SQL string literals across all three dialects.