On Friday, January 17, 2003, 4:22:16 AM, Travis wrote:

> I'm just starting to get into unit testing, and I can already see how powerful
> a tool it can be to preemptively find bugs in my programs; however, a good
> portion of the classes I write revolve around databases and CGI. A lot of the
> methods implemented in these classes return highly variable results. Is there
> a way for me to leverge the benefits of unit testing when working with this
> type of data?   [code snipped]

> Would there be a way to use unit test my get_pct_solids method, or is it only
> suited for environments where functions return predictable values?

Several times in the past, I've created test data in test tables (or
perhaps tables of the same name in a different database).  This can be
quite arduous, and has indeed taken a long time to get right.

To do it properly, you need to create the test data in #setup, and
remove it in #teardown.

I haven't used mock objects, as Hugh suggests, but based on my
experiences with hardcoding test data in a database, I'll certainly
look into them next time.

Cheers,
Gavin