HippoDatamusA million rows before your coffee lands

Fast enough that you stop trimming your test data - and reproducible, so the same configuration and seed always produce the same rows.

Download

Documentation

Release notes

Fast, on measured numbers

Over 150,000 rows a second on a twelve-column order table - a million rows of CSV in a few seconds, with memory that stays flat however many rows you ask for. Generate the volume you actually need instead of a sample you have to explain.

See the benchmarks

Reproducible by construction

A configuration and a seed pin the output exactly. Regenerate it next month, on another machine, and get the same bytes - so a generated dataset is something you can cite in a bug report rather than something you hope looks the same.

How it works

59 column types

Names, addresses, emails, identifiers, dates, money, IP addresses, pick lists, values looked up from a spreadsheet you already have, and templates composed from other columns.

See them all

Constraints that actually hold

Unique columns and unique combinations are checked against the number of values that exist before a run starts, so you find out immediately rather than at row 400,000.

Uniqueness

Bring your own data

Point a column at a column of an existing spreadsheet, or import a header row to get a starting schema. Real names and real product codes, generated volumes.

Reading from a file

One file to hand over

A .hippo bundle seals a schema together with every file it reads. The command-line tool takes a bundle directly, so a whole scenario travels as one artefact.

Bundles

Desktop app and CLI

Build the schema in the application, then run it in CI with the command-line tool. Both read the same configuration file and produce the same data.

Command line

Free

HippoDatamus is 0.7.0 and free to use. No account, no licence key, no trial period, and nothing to sign up for.

Future versions may be paid. A version you have already downloaded stays licensed under the terms it came with, so a price on a later release is not a bill for the one you are running.

Below 1.0 the configuration format and the data a given configuration produces may still change between releases. Changes to generated output are always called out in the release notes — an existing schema never quietly starts producing something else.

What a configuration looks like

yaml

Output:
  Path: customers.csv
  Format: CSV

Schema:
  RowCount: 1000
  GlobalSeed: 20260822
  Columns:
    - Id: 11111111-0000-0000-0000-000000000001
      Name: customer_id
      Generator:
        Type: Guid

    - Id: 11111111-0000-0000-0000-000000000002
      Name: name
      Generator:
        Type: Full Name

    - Id: 11111111-0000-0000-0000-000000000003
      Name: status
      Generator:
        Type: Pick List
        Settings:
          ValueType: String
          Values:
            - Active
            - Pending review
            - Closed

customer_id,name,status
6d71b4ad-6907-46ab-a14b-75a820889519,Nia Krajcik,Active
9b839a94-3fb7-6d84-7b4d-0e5bd47bee71,Candice Blick,Closed

Run it again in a year and you get those same two rows. That is the whole idea.

How fast, exactly

Shape	Rows/second
Sensor readings, 8 columns	~250,000
Order lines, 12 columns	~150,000
Ledger entries, 10 columns	~125,000
Contacts with a spreadsheet lookup, 14 columns	~80,000

Measured, not estimated, and regenerated from a real run on every release. The performance page has the full grid, the machine it ran on, and what actually costs time.