January 9, 2018
Test Granularity Matters. Ask Any Accountant.It's that time of year when I have to make sure my company's accounts are all up to date and tickety-boo, and I got a useful reminder about why the granularity of our tests really matters.
In my spreadsheet for bank payments and receipts, I have a formula for calculating what the closing balance at the end of the financial year is. Today, I realised that calculated balance was about £1200 short. Evidently, I had either entered one or more payments incorrectly, or one or more receipts.
I had to go back through all the bank statements for the year double-checking every line item against the spreadsheet.
Now, if I'd had a formula for the balance at the end of every line item, I could simply have checked the closing balances on each statement to see where they diverged.
I've experienced similar pain when relying on tests that check logic at too high a level (e.g., system tests or API tests). When a test fails, I have to go rummage through the call stack to figure out where it went wrong - the equivalent of reading all my bank statements looking for the line item that doesn't match. Much time is spent in the debugger: a red flag.
I strongly encourage teams to rely more on small, focused tests that - ideally - have only reason to fail, and to write those tests as close to the module that's doing that piece of work as they can. So when a test fails it's easy to deduce that "the problem is this, and the problem is here".
Posted 21 hours, 8 minutes ago on January 9, 2018