Show HN: Generate test data for your database tables | Hacker News Reader

Show HN: Generate test data for your database tables

databasetestdata.com

57 pointsvasco17 comments

Creating rnd data on a single table is easy, I don't really need a tool for that. However, it would be extremely useful to have a tool able to creata test data using actual relations between tables in the DB (e.g. generate fake users, then use their User IDs to generate 1-10 random posts for each user. Then for each posts I want a couple of random comments attached. And so on and on...)

I think a tool like this should be able to inspect the real database and model the columns statistically. If my salary info is distributed a certain way I would like the fake data to have the same distribution. A tool could also build language models of my text columns to generate random names that seem like real names.

Ideally the tool could start to analyze relationships between tables and columns and produce data that preserved those relationships in a statistical way. For example if in production each customer has 5 orders on average then so could the test data.

I'm not sure what you could do with clobs or xml data in columns but I guess some stuff would remain opaque.

I could see this tool being useful to enable you to move your test code to the cloud w/o revealing any private data to third parties. Studying a model like this could also help new developers come up to speed on the system.

This is easily done with a few lines of Python. I personally think that's a more efficient way to go -- the Web interface doesn't really provide anything essential to the process, and the Python approach generates more data more quickly.

Also, Python knows how to talk directly to the user's MySQL database if it exists.

To be fair, you could say that about most tools on the web though.

Yes, that's fair, and I could. It's true -- there are any number of advertising-supported Web-based applications whose only reason for existence is to recruit eyeballs. Not a comment on the merits of the present example, just the pattern.

I think this needs a SQL export format. It is 'databasetestdata' after all :)

Will look into it, thanks for the tip. Also had an idea of accepting an SQL schema dump to auto-generate the right columns. There's definitely stuff to improve.

Generate SQL is really easy using Squel.js, especially when all you need is insert statement.

Pretty cool! I think a good addition would be adding the ability specify data type for random numbers. So if I want to test 64-bit integers, I can have it generate numbers in the 64-bit range that exceed the 32-bit range.

Yeah, one of the things I had in mind were special option fields for each kind of data. Like specifying range of numbers, number of words, formats for data/time/zipcodes... But I wanted something out of the door quickly and had to leave that out for now. Thanks for the feedback :)

Interesting, some of the 'test' data emails it gives me seem awfully real. For example: Lue_Schiller@shanie.net

The domain is real, the email probably not...but if they have a catch all someone else using this could get pretty annoying.

use @example.com for emails. see http://www.iana.org/domains/example for more info.

I know its the standard, but how will that help validating, creating statistics, or otherwise interacting with the domain portion of an email address?

You could use @prefix.example.com.

Clicking on "Load Recipe" gives "No LocalStorage support detected. Please upgrade your browser to enable." on both Firefox 19.0.2 and Chrome 25.0.1364.172 on my Mac. I presume that's a bug because a web site that I wrote which uses HTML5 local storage works fine in those same browsers.

Edit: I ought to have said, cool idea, and I like being able to save and load recipes. It would be neat to be able to share recipes on this, like you can do with web pages on JsFiddle.

You have a bug. It returns null for a random number.

Ah, there was a typo. It's fixed, thanks.

Reply on news.ycombinator.com