[dojo-contributors] unit tests: t.assertTrue("foo") fails

Brian Douglas Skinner skinner at dojotoolkit.org
Mon Jun 11 18:30:52 EDT 2007

In writing some unit tests, I got tripped up by the behavior of 

Doing this works fine:

But doing this causes an error:

That feels counter-intuitive to me.  I can see where maybe you might 
argue that assertTrue should accept only actual boolean values (true, 
false).  But given that assertTrue is designed to accept non-boolean 
values and treat "falsey" values as false, then it seems like "foo" 
should count as non-falsey.

The assertTrue("foo") call causes an error because assertTrue calls eval 
on the arg.  I'm not sure why it does that, but I'm guessing it does 
that so that it can give better error messages.  For example, the line:
   t.assertTrue("3 == 8/2");
results in
   doh._AssertFailure: assertTrue('3 == 8/2') failed

But, looking at some actual unit tests, it seems like nobody is taking 
advantage of that feature -- people do assertTrue(3 == 8/2) instead of 
assertTrue("3 == 8/2").

Would it be a good idea to change assertTrue so that it does not call eval?

Or, if we do want to keep the eval behavior, would it make sense to 
expose that more explicitly?  We could have two separate methods: 
doh.assertEval(condition) and doh.assert(condition), where assertEval() 
does if(eval(condition)), and assert() just does if(condition).

:o) Brian

More information about the dojo-contributors mailing list