Test Data Generator

functional

Deterministic, dialect-correct INSERT scripts built from a table's real column metadata.

Generate deterministic, dialect-correct INSERT scripts for a table, built from its real column metadata - types, nullability, and which columns the engine fills itself, so the sample data actually fits the schema you have. Name the table, choose a row count (up to 10,000), a NULL percentage for nullable columns, and a seed: the same seed always produces the same data, so a test fixture stays reproducible.

Load Columns reads the table's catalog and lists every column with an inferred generator - names, words, integers, decimals, dates, GUIDs, and friends - chosen from the column's type and name; click any column to cycle it to a different generator. Identity and computed columns are skipped automatically, values are clamped to each column's declared length and precision so nothing truncates or overflows, and Script INSERTs emits the batch (500 rows per statement, headed by a comment recording the row count and seed) into a query tab. It never writes to the server itself.

SQLly
SQLly Test Data GeneratorSQLly Test Data Generator
Test Data Generator: choose a table and row count; SQLly emits INSERT statements matching the column types.

Filling a customers table

  1. Right-click the database and choose Tools › Test Data Generator....
  2. Enter Schema (optional) and Table, for example sales and Customers. A blank schema means the engine's default: dbo on SQL Server, public on PostgreSQL, main on DuckDB, the current database on MySQL and ClickHouse, and the current schema on Oracle.
  3. Set Rows (default 100), NULL percent for nullable columns (default 10), and Seed (default 1).
  4. Press Load Columns. Each column shows its type, a nullable marker, and its generator, or skipped (identity/computed).
  5. Click a generator button to cycle it, for instance to turn a Notes column from Word into Text.
  6. Press Script INSERTs, read the script in its new tab, and run it when you are happy.
-- Test data generated by SQLLY: 100 rows, seed 1. Review before running.
-- The same table, generators, and seed always produce identical data.
INSERT INTO [sales].[Customers] ([Name], [Email], [Joined]) VALUES
  (N'Maria Chen', N'[email protected]', N'2023-04-17'),
  ...

Generators

There are thirteen: Name, Word, Text, Email, Integer, Decimal, Boolean, Date, Date & time, Time, GUID, JSON, and Binary. The starting choice comes from the type first - GUID, JSON, binary, boolean, date and time, integer, and decimal types each get their own generator, and, for text columns, from the name: a column containing email gets emails, name gets person names, and description, comment, or note get a few words of text. Everything else gets a single word.

  • Strings are cut to the column's declared length, decimals respect precision and scale, small integer types stay in range, and binary values fit the declared byte length.
  • Emails use the example.com domain, and dates fall between 2023 and 2025.
  • Boolean columns are recognized by type - including SQL Server's bit, MySQL's tinyint(1), and Oracle's NUMBER(1) - and get the engine's own true/false literal.
  • ClickHouse Enum8 / Enum16 columns get one of their declared members; IP address columns get valid addresses.

Per-engine details

  • Skipped columns: identity and computed columns on SQL Server; identity, generated, and nextval(...) defaults on PostgreSQL and DuckDB; auto_increment and generated columns on MySQL and MariaDB; an INTEGER primary key on SQLite and libSQL; MATERIALIZED and ALIAS columns on ClickHouse; identity and virtual columns on Oracle.
  • On Oracle, each batch is a single INSERT ... SELECT ... FROM DUAL UNION ALL ..., with typed DATE and TIMESTAMP literals.

Limits

  • The generator does not read foreign keys or unique constraints. Values can repeat, and a foreign-key column gets ordinary integers, so fill parent tables first or point those columns at real ids by editing the script.
  • Rows above 10,000 are capped at 10,000. If every column is identity or computed, there is nothing to insert and the script button stays disabled.
  • The script opens on the current connection; running it is up to you, and your normal safety settings apply when you do.

Supported on SQL Server and Azure SQL, PostgreSQL, MySQL/MariaDB, SQLite, libSQL / Turso, DuckDB, ClickHouse, and Oracle Database; each engine's own catalog and quoting rules are used for the script (on Oracle, typed DATE, TIMESTAMP and HEXTORAW literals in batched INSERT … SELECT … FROM DUAL statements). Redis has no tables to fill, write keys with SET / HSET in a query tab.