Random Test-Data Generator
Build reproducible sample records, inspect them and export complete test datasets.
Generation stays in this tab and its local worker. No schema, names, seeds or results are uploaded, logged or saved.Build reproducible sample data
Choose fields, enter a row count and select Generate data. Review the captured preview, then copy or download every record. Use Advanced for related tables and extra field settings.
Synthetic test fixtures only. Seeds and UUID values are predictable, including automatically created seeds after they are shown. Do not use this tool for passwords, security tokens, anonymization or real identities.
Up to 5 tables, 20 fields per table, 10,000 rows per table and 200,000 cells across the whole project. Complete JSON and all CSV files together allow 8 MiB; SQL allows 8 MiB. Nothing is silently truncated.
Fields for records
Each name identifies a field. With nested objects on, use paths such as profile.city. Decimal values keep exact fixed-point text in JSON.
Null rate gives the exact rounded number of base nulls per field, shuffled separately. Correlated fields also inherit nulls from their source, so inspect their actual counts. Nulls may repeat in unique fields.
Active options: nested objects off; boundary-first valid values off; 1 tables. Related fields use only earlier tables or earlier fields. Boundary mode favors finite-pool lower, upper and middle values; it does not promise invalid cases or natural frequencies.
Advanced — field rules, related tables, schemas and projects
Object shape and boundary cases
Paths use up to six ASCII identifier segments, such as address.city. A leaf cannot also contain another leaf. With this off, names remain literal, including dots.
First non-null records favor distinct lower, upper and middle pool values while preserving uniqueness. Foreign keys remain actual parent values, weighted choices never use zero-weight values, and correlations are calculated from the final source. UUID test boundaries retain version/variant bits. Small datasets may have room for only some boundaries.
Field rules for records
Related table management
Use distinct ASCII identifiers. Renaming/removing a referenced table requires reviewing its references; earlier captured output stays available. Table order determines which keys are available.
Import JSON Schema
Import replaces all current tables with one table using the selected table’s row count, preserves the seed/CSV policy, generates every declared property, and sets nulls to zero. Supported draft 2020-12 subset: objects, safe inclusive integers, booleans, string enums, bounded anchored patterns, bounded lengths and date/email/uuid formats. Exact decimals remain JSON strings, so number schemas are rejected. Unsupported keywords, references, arrays, composition and combined enum/format constraints are rejected rather than ignored. Files allow 2 MiB.
Reusable schema project files
Version 1 project files retain every table, field option, seed and switch, including inactive options, without generated records. Save the actual captured project when exact reproduction matters. Import is checked completely before replacement and supports Undo; it never uploads or automatically saves data.
All source settings, seeds, results and Undo stay in this tab and clear on reload. Downloaded files can contain sensitive text you entered; keep them private. No external references or database connections are opened.
How to generate and reuse test fixtures
Choose names and types, enter bounds for numbers or dates, and set a row count and null rate. Use a seed to reproduce a complete configuration. A blank seed creates a visible seed locally. Generate data keeps a full captured result and a first-30-row preview; JSON and CSV include every accepted row.
Open Advanced to enable uniqueness, dotted nested paths, valid boundary-first values, extra field types or related tables. Foreign keys choose actual unique values in an earlier table. Correlated fields copy, calculate from or map an earlier field. Every job validates the whole dataset before replacing output. Invalid settings and Cancel retain previous captured exports; Undo restores one complete workspace.
Import the supported JSON Schema subset to start a schema, or save/load a versioned schema project to preserve all tables, field options and the seed. Import replaces the working project only after validation. Closing Advanced keeps every setting. Earlier captured results remain clearly labeled when settings change.
Example: repeat IDs and link two tables
For the original three-row fixture, use one unique Whole number field named id, minimum 1, maximum 3, three rows, zero nulls and seed tulip-demo. With the original tt-fixture-v1 algorithm the IDs are 3, 2, 1 and complete JSON/CSV bytes repeat.
Load related-table example in Advanced for parent and child records. Parents have unique IDs, nested fictional names, codes, weighted categories and seeded test UUIDs. Child keys come from the actual parent IDs; totals use units × 10 + 5 and labels map the units. Boundary mode keeps valid endpoints visible without breaking keys, correlations or uniqueness. Replace the practice schema before use.
About determinism, boundaries and privacy
Keep the actual seed, full captured schema and generator version together. The original single-table native mode retains tt-fixture-v1 bytes; extended modes use tt-fixture-plan-v1. Seeds, generated UUIDs and all records are predictable test data, unsuitable for secrets, anonymization or population claims. Fictional names and reserved-domain emails remain synthetic.
Inclusive safe integer bounds, exact decimal strings at 0–6 places, strict UTC dates from 1970 through 2100, exact base null masks and unique non-null pools retain the native rules. Correlations can inherit extra nulls. Boundary mode selects valid lower/upper/middle pool values first; it does not certify complete edge coverage or natural frequencies. Pattern generation uses a small bounded subset, never arbitrary regex code.
Nested JSON uses dotted paths; CSV and SQLite INSERT columns remain flat. Formula-safe CSV can convert risky cells, including negative numbers, to apostrophe-prefixed text. JSON keeps original types; CSV empty cells cannot distinguish empty strings from null. SQL retains decimal strings as quoted text, quotes names/literals and contains no schema creation or automatic execution. Review a compatible empty test database before importing it.
Everything stays in browser memory and local workers; reload clears the workspace without touching unrelated browser data. Downloaded JSON, CSV, SQL and project files remain on your device. No schema, seed or result is uploaded, logged or automatically stored.
Stars and feedback
Loading…
What should we improve next?
Help shape the next update to Random test-data generator. Tell us what would make it better for you.
Prefer email? feedback@tooltulip.com
Tool reference #181