Re: [Nmock-general] MethodSignature and friends

[Steve F. <st...@m3...>]

Have you looked at how jmock works? Nat will be rolling onto a .Net 
project soon, so I expect we'll finally have someone with the motivation 
to do something about a port (making promises on his behalf).

S.

Levi Khatskevitch wrote:

> I had the same idea a while ago, but how would you handle argument 
> constraints like NotNull() that do not specify the argument type? Losing 
> them will make expectations more verbose but on the other hand requiring 
> too precise expectations often results in unit tests that follow real code 
> too closely - a common problem with interaction base testing that I've 
> seen.
> 
> Recently I was thinking about resolving methods to the first expectation 
> in the call sequence where argument types match. This preserves the 
> flexibility but also supports overloading in a more consistent way. For 
> example:
> 
> mock.Expect("Foo", new IsAnything()); // expect either overload
> mock.Expect("Foo", (int)1); // expect Foo(int) overload
> mock.Expect("Foo", (double)1.2); // expect Foo(double) overload
> 
> mock.Invoke("Foo", 2); // ok, resolved to the first expectation
> mock.Invoke("Foo", 1.0); // fail, resolves to the third expectation and 
> thus should be 1.2
> 
> Any thoughts on this?
> 
> Regards,
>  - Levi
> 
> 
> 
> 
> Jim Arnold <JA...@th...> 
> Sent by: nmo...@li...
> 11/01/2004 08:41 AM
> 
> To
> nmo...@li...
> cc
> 
> Subject
> [Nmock-general] MethodSignature and friends
> 
> 
> 
> 
> 
> 
> The method resolution in NMock is really starting to annoy me :-)  It 
> works for simple cases (find a method named "Foo" with a return type of 
> "Bar"), but complex scenarios with overloads  and indexed properties are 
> still buggy.  The fundamental problem is that the user is not forced to 
> specify the attributes necessary to find a method.  SetupResult() can just 
> take a method name and a return value, which is really not enough to 
> resolve overloaded properties.
>  
> We should just get rid of the whole MethodSignature thing and start using 
> real MethodInfos and PropertyInfos instead.  It might make the public API 
> slightly more complex, but would close a whole class of bugs.
>  
> Can I get a +/- 1 from anybody?
>  
> Jim
> ------------------------------------------------------- This SF.Net email 
> is sponsored by: Sybase ASE Linux Express Edition - download now for FREE 
> LinuxWorld Reader's Choice Award Winner for best database on Linux. 
> http://ads.osdn.com/?ad_idU88&alloc_id065&op=click 
> _______________________________________________ Nmock-general mailing list 
> Nmo...@li... 
> https://lists.sourceforge.net/lists/listinfo/nmock-general 
>