From: Nathaniel Talbott <nathaniel@...>
Date: 2002-03-16T09:54:06+09:00
Subject: RE: testunit - setup -> set_up ?

Ron Jeffries [mailto:ronjeffries@REMOVEacm.org] wrote:

> On Fri, 15 Mar 2002 02:10:16 GMT, "Nathaniel Talbott" 
> <nathaniel@talbott.ws> wrote:
> 
> >> In message "RE: testunit - setup -> set_up ?"
> >>     on 02/03/14, "Nathaniel Talbott" <nathaniel@talbott.ws> writes:
> >> 
> >> |Ah, but JUnit uses setUp... two words. So which is the "xUnit
way?"
> >> 
> >> I just miss "setup" just for its conciseness.
> >
> >Woah... pressure comes to bear ;-)
> 
> Doesn't sUnit use setup?

I have no earthly idea if it does now, but looking at Kent's original
write-up (http://www.xprogramming.com/testfram.htm), it originally used
setUp.


> >To avoid more floundering, let me bring up a couple of other name 
> >changes I'm pondering:
> >
> >#assert_raises -> #assert_raised
> >#assert_nothing_raised -> #assert_not_raised
> >#assert_throws -> #assert_thrown
> >#assert_nothing_thrown -> #assert_not_thrown
> 
> in many cases I might just allow either, but that's a style 
> that some disagree with.

I would be one of them, at least at this time and in this place.


> assert_raises seems right to me because the next thing is the 
> name of the exception.

Interesting... this for me touches on the whole #assert_equal vs.
#assert_equals thing, and has to do with how we read code. In Smalltalk,
I would say,

	assert: aBlock raises: anExceptionList

At least, if I wrote Smalltalk, I think that's what I'd say. However, in
Ruby, all the arguments are at the end:

	assert_raised(*args) {}

So for me, the sentence in Ruby reads more like, "Assert that these
exceptions are thrown by this block: exceptions, block". Does anyone
else read code like this? Perhaps the different sentence perception has
to do with your Smalltalk background and my Java one? Am I stuck on an
issue that doesn't really matter?

I'm decided on #assert_equal. I'm still not sure about #assert_raised.
I'm glad to be convinced one way or the other.


> assert_nothing_raised seems right because the next thing is 
> /not/ the name of an exception.

False. Exceptions can be specified as arguments to the method, and if
they are, only those exceptions are rescued. Does that change your mind?


> Ditto the throwns.

I wonder if the behavior and naming of the assertions for exceptions and
signals should be symmetrical. Currently #assert_nothing_thrown does not
allow the specification of specific signals to be caught in the way
#assert_nothing_raised does for exceptions. If it did, I'd definitely
keep the names symmetrical. If it stays the same, I'll probably break
the naming symmetry as a small mnemonic of the behavioral difference.


> >I'll keep thinking about #setup vs. #set_up, and get back to ya'll.
> 
> Get it right, Nathaniel, this one is for the species!

I've been through enough iterations of it to hopefully be getting close
:-)


Nathaniel

<:((><
+ - -
| RoleModel Software, Inc.
| EQUIP VI