Re:Ask HN: Why don't you write functional tests?
by jacobian on Wednesday July 11, 2012 @04:40PM
(#4231349)
My experience has been that functional tests are excellent and discovering that "something" is wrong, but less good at finding what is actually broken.
Say my "log in" test fails: OK, I know auth is broken... but what part of it? Is there some JavaScript hijacking the form submission? Is my login view busted? Do I have an error in my authorization system itself? Is the LDAP server I'm authorizing against down? Has someone changed the schema? Etc., etc., etc.
I do find functional tests to be a pretty valuable part of my testing toolkit, but unit tests seem to get me a bigger "bang for my buck", so functional tests tend to be something I don't spend a ton of time working on.
(I wish I did write more functional tests, but clearly they're not valuable enough for me to prioritize them. Otherwise I'd already be using 'em more.)
Re:Ask HN: Why don't you write functional tests?
by hugs on Wednesday July 11, 2012 @04:46PM
(#4231374)
Thanks, Jacob. :-) Follow-up question: Do you do monitoring in production? Does it have the same issues in that it tells you that /something/ is broken, but doesn't tell you what's broken?
Say my "log in" test fails: OK, I know auth is broken... but what part of it? Is there some JavaScript hijacking the form submission? Is my login view busted? Do I have an error in my authorization system itself? Is the LDAP server I'm authorizing against down? Has someone changed the schema? Etc., etc., etc.
I do find functional tests to be a pretty valuable part of my testing toolkit, but unit tests seem to get me a bigger "bang for my buck", so functional tests tend to be something I don't spend a ton of time working on.
(I wish I did write more functional tests, but clearly they're not valuable enough for me to prioritize them. Otherwise I'd already be using 'em more.)