Test data generator with a seed you can write down

Reproducibility. A generator that produces different values on every run gives you a test that fails randomly and a bug report nobody else can reproduce. Here every value is a function of the seed, the row number and the column, so “seed 7, 200 rows, this schema” is a complete description of a dataset — write it in the ticket and anyone can rebuild the exact rows that broke the import. Press the New rows button and the seed goes up by one, which is why it is shown on the button rather than hidden: the number is the point.

Because a value that follows the format of a real identifier is indistinguishable from a real one the moment it leaves this page. Test rows end up in databases, screenshots, spreadsheets and support tickets, and at that point a correctly formed identity card number looks exactly like somebody’s. There is no way to generate one that is obviously not real, so none is generated here — and the same reasoning rules out bank account numbers, payment card numbers and vehicle registration plates. If you need a placeholder for such a column, a UUID column or your own list of values does the job without the risk.

Because those are the only ranges we could establish as officially reserved for fiction. The North American Numbering Plan sets aside 555-0100 through 555-0199 in every area code, and Ofcom reserves 07700 900000–900999 and 020 7946 0000–0999 for use in drama; numbers in those ranges cannot be assigned to a subscriber, so they cannot ring a real person. No equivalent Malaysian range could be established, and generating a well-formed +60 number would mean generating a number that may well belong to someone. If a reserved Malaysian range exists and can be cited, adding it is a small change.

No. Names are common given names and family names from small built-in lists — 24 Malay personal names, 16 patronyms, 16 Chinese family names, 18 Chinese personal names, 17 Indian personal names, 14 Indian patronyms, and 28 English names with 24 surnames — recombined at random. A row is a combination, not a record, and nothing about a row comes from anywhere but those lists. Postcodes are different in kind: they are published delivery-area codes, so real ones are used and matched to their city, because a fixture that puts Kuching in Selangor fails every address validation test ever written. A postcode identifies an area, never a person.

Enough to plan around, and the arithmetic is simple. On the Malaysia setting there are 1,920 distinct Malay full names (12 personal names per gender, in five forms once the optional Mohd, Muhammad, Nur or Siti prefix is counted, against 16 patronyms), 288 Chinese and 238 Indian — about 2,450 in total. By the birthday bound a repeat becomes more likely than not somewhere around 60 rows, so a 500-row set will certainly contain duplicate names. Emails carry an extra two or three digit number, which makes them far less likely to collide but does not guarantee uniqueness; if you need a genuinely unique key, add the auto-increment id column, which is unique by construction.

PostgreSQL, MySQL and SQLite, without editing. Column and table names are reduced to letters, digits and underscores and written without quotes, because Postgres and SQLite quote identifiers with double quotes while MySQL uses backticks unless ANSI_QUOTES is on — an unquoted identifier is the one form all three accept. Text values are wrapped in single quotes with any internal quote doubled, which is standard SQL. One caveat worth knowing: a backslash inside a string is a literal in Postgres and an escape character in MySQL unless NO_BACKSLASH_ESCAPES is set. Nothing generated here contains a backslash, so this only matters if you type one into a list column.

No, and this is the one misuse worth spelling out. Every value on this page, the UUIDs included, is derived from the seed by an ordinary arithmetic generator — that is exactly why the output is reproducible, and it also means anyone holding the seed can recreate it. That is the correct trade for a fixture and completely wrong for a secret. A UUID here is a stand-in for an identifier column, not a token. Anything that has to be unguessable should come from a cryptographic source; the UUID generator on this site uses one.

Because everything runs in the tab you have open, and the honest ceiling is the point at which building the string starts to feel slow rather than the point at which the browser gives up. A thousand rows across twenty columns is already a few megabytes of SQL. If you need more, generate several files with different seeds — 1, 2, 3 — and load them one after another; because the seed fully determines the contents, the sets are different from each other and each one is still reproducible on its own.

Nothing is uploaded — the word lists are inside the page rather than fetched, which is why the tool keeps working with the network off. Your schema is saved to this browser’s local storage so a set of columns you spent time building is still there tomorrow; it stays on your device and reaches us at no point, because there is no server for it to reach. The generated rows are not saved, since they can be rebuilt from the seed in a fraction of a second, and writing a thousand rows to disk for something recomputable would be storing data for no reason. Clearing this site’s data removes the saved schema.