Beta 2.5 and TestCase with array parameters

[Mark]

Hello,
I have a question about the Beta of 2.5.

I am attempting to pass an array of objects to my paramaterized test
so that the parameter of the test is an array of objects but I can't
quite seem to figure out how it is supposed to work.

Example
[TestCase(1, "s" , Result = "Int32;String;")]
[TestCase(1, "s" , true, Result = "Int32;String;Boolean")]
[TestCase(1, "s", 1.1, false , Result =
"Int32;String;Double;Boolean")]
public object ParametersArePassedCorrectly(object[] args)
{
var f = new Function();
return f.Call(2, args);
}

I have even tried something like
[TestCase(new object[] { 1, "s" } , Result = "Int32;String;")]
public object ParametersArePassedCorrectly(object[] args)
{
var f = new Function();
return f.Call(2, args);
}

In all cases NUnit indicates the test was not executed becase "Wrong
number of arguments provided".

Any ideas or is this a bug/limitation of the api?

[Charlie Poole]

Hi Mark,

> I am attempting to pass an array of objects to my
> paramaterized test so that the parameter of the test is an
> array of objects but I can't quite seem to figure out how it
> is supposed to work.
>
> Example
> [TestCase(1, "s" , Result = "Int32;String;")]
> [TestCase(1, "s" , true, Result = "Int32;String;Boolean")]
> [TestCase(1, "s", 1.1, false , Result =
> "Int32;String;Double;Boolean")]
> public object ParametersArePassedCorrectly(object[] args)
> {
> var f = new Function();
> return f.Call(2, args);
> }

This should not work, because you are passing, respectively, two,
three and four arguments to a method that only takes one.

> I have even tried something like
> [TestCase(new object[] { 1, "s" } , Result = "Int32;String;")]
> public object ParametersArePassedCorrectly(object[] args)
> {
> var f = new Function();
> return f.Call(2, args);
> }

This is intended to work, so we have a bug. I'm guessing that
new object[] { new object[] { 1, "s"} } will do what you want,
but that's not a satisfactory solution.

Could you file a bug on this - I'd like to fix it before releasing
the second beta. If you try the above workaround and let me know
what happens, that will be helpful.

Charlie

[Mark]

Hi Charlie,

> > I have even tried something like
> >         [TestCase(new object[] { 1, "s" } , Result = "Int32;String;")]
> >         public object ParametersArePassedCorrectly(object[] args)
> >         {
> >             var f = new Function();
> >             return f.Call(2, args);
> >         }
>
> This is intended to work, so we have a bug. I'm guessing that
> new object[] { new object[] { 1, "s"} } will do what you want,
> but that's not a satisfactory solution.
>
> Could you file a bug on this - I'd like to fix it before releasing
> the second beta. If you try the above workaround and let me know
> what happens, that will be helpful.

This workaround does work.

I tried to enter a bug on SF but it appears that anon reports are not
allowed and I don't have the time right now to setup a SF account.

-mark

[Simone Busoli]

I'd like to give this a look, I think it depends on the way the params keyword works in C#. I tried creating a TestCase with an object[] but the NUnit GUI marks that as invalid. I don't know the code except for the framework, any hits on where to look?

[Charlie Poole]

Hi Mark,

> > Could you file a bug on this - I'd like to fix it before
> releasing the
> > second beta. If you try the above workaround and let me know what
> > happens, that will be helpful.
>
> This workaround does work.

OK, then it's clear why it happens.

> I tried to enter a bug on SF but it appears that anon reports
> are not allowed and I don't have the time right now to setup
> a SF account.

I'll file it then. Thanks for the info.

Charlie

> -mark
> >
>

[Charlie Poole]

Sure... it's in the constructor of TestAttribute, which has overloads:

 

1:  TestAttribute( params object[] )

2:  TestAttribute( object )

3:  TestAttribute( object, object )

4:  TestAttribute( object, object, object )

 

2, 3 and 4 are for languages that don't know about params

 

In addition, 2 was intended to pick up the case where a single object[] parameter

was passed in. Obviously, it ain't working, which is understandable since

object[] is a closer match for object[] than object is.

 

I suggest looking at how RowAttribute handles this, because I remember that

Andreas had to deal with exactly the same issue. Then, write a test, make

it fail, make it pass. :-)

 

If you have time for this, I'm glad to leave it for you!

 

Charlie

---

From: nunit-...@googlegroups.com [mailto:nunit-...@googlegroups.com] On Behalf Of Simone Busoli
Sent: Monday, January 12, 2009 1:30 PM
To: nunit-...@googlegroups.com
Subject: [nunit-discuss] Re: Beta 2.5 and TestCase with array parameters

[Charlie Poole]

PS: If you are fixing this, please file the bug as well, so we remember to put it into the list of bugs fixed.

 

Charlie

---

From: nunit-...@googlegroups.com [mailto:nunit-...@googlegroups.com] On Behalf Of Charlie Poole
Sent: Monday, January 12, 2009 1:44 PM

[Charlie Poole]

Mark wrote:

> I tried to enter a bug on SF but it appears that anon reports
> are not allowed and I don't have the time right now to setup
> a SF account.

Note to self: allow anon bugs on our next bug tracker system :-)

Charlie

[Simone Busoli]

I was there already, what I see is the gui not being able to load that test.

[Simone Busoli]

I think sf does that already if you enable it.

[Olof Bjarnason]

Maybe choose an easier-to-use-issue tracker is the real deal ;)

2009/1/12 Simone Busoli <simone...@gmail.com>:

--
Min blogg:
http://olofb.wordpress.com
[My blog, in Swedish]

[Olof Bjarnason]

2009/1/12 Olof Bjarnason <olof.bj...@gmail.com>:

> Maybe choose an easier-to-use-issue tracker is the real deal ;)

And with that I simply mean I think sf.net is hard to use.. are there
any plans on changing to another issue-tracking-system? Eg.
launchpads?

[Kelly Anderson]

I'll second that one... :-)

-Kelly

[Simone Busoli]

I think the constructor which accepts a parameter array should be removed because once you're there you cannot distinguish between several arguments and a single argument of type object[].

Removing that constructor fixes the reported issue but breaks one test which supplies more than 3 arguments to the TestCaseAttribute constructor.

[Simone Busoli]

Here's the sourceforge issue: https://sourceforge.net/tracker2/?func=detail&aid=2503011&group_id=10749&atid=110749

Attached to the issue is a patch. The resolution ha not been to remove the constructor accepting a parameter array, but instead coalesce the attribute parameters in an object[] in case the test method has an argument of type object[].

[Charlie Poole]

I just enabled it. Of course, if we start getting spam, I'll have to revert to requiring login.

 

Charlie

---

From: nunit-...@googlegroups.com [mailto:nunit-...@googlegroups.com] On Behalf Of Simone Busoli

Sent: Monday, January 12, 2009 1:48 PM

To: nunit-...@googlegroups.com
Subject: [nunit-discuss] Re: Beta 2.5 and TestCase with array parameters

[Charlie Poole]

Hi Simone,

 

I'll have to take a look at your fix. We may need to do more if other tests are breaking.

 

If I remember correctly, we want to take a simple object[] as an argument by itself, IFF

the method takes an object[]. However - again IIRC - TestCaseAttribute doesn't know

anything about the method being called, so it can't do that. I'll look deeper.

 

Charlie

---

From: nunit-...@googlegroups.com [mailto:nunit-...@googlegroups.com] On Behalf Of Simone Busoli

Sent: Monday, January 12, 2009 4:09 PM

To: nunit-...@googlegroups.com
Subject: [nunit-discuss] Re: Beta 2.5 and TestCase with array parameters

[Simone Busoli]

No other tests are breaking :) Actually, with the approach I followed we get the benefit that even if you specify arguments individually in the constructor of the attribute and you have a test method which accepts an obejct[], the arguments are wrapped in the array. I added some tests to show this.

[Charlie Poole]

OK - I must have misunderstood you then. I'll look again. I see what you're saying about being able

to specify it either way.

 

Charlie

---

From: nunit-...@googlegroups.com [mailto:nunit-...@googlegroups.com] On Behalf Of Simone Busoli

Sent: Tuesday, January 13, 2009 12:17 AM

[Simone Busoli]

Oh I see what you mean,  previously I meant that I had to comment a test which supplied 4 arguments to the attribute constructor, because in the first place I had removed the constructor which takes a parameters array, but now it's all fine.

[Charlie Poole]

It's checked in. Thanks.

 

Charlie

 

---

From: nunit-...@googlegroups.com [mailto:nunit-...@googlegroups.com] On Behalf Of Simone Busoli

Sent: Tuesday, January 13, 2009 1:39 AM

[Simone Busoli]

Did you have to make any changes?

[Charlie Poole]

Hi Simone,

Yes, I changed

if (method.GetParameters().Length == 1 &&
method.GetParameters()[0].ParameterType == typeof(object[]))
parms.Arguments = new object[]{parms.Arguments};

to this

if (argsNeeded == 1 && method.GetParameters()[0].ParameterType ==
typeof(object[]))
{
if (parms.Arguments.Length > 1 ||
parms.Arguments.Length == 1 && parms.Arguments[0].GetType() !=
typeof(object[]))
{
parms.Arguments = new object[] { parms.Arguments };
}
}

making this new test pass as well...

[TestCase(new object[] { 1, "two", 3.0 })]
[TestCase(new object[] { "zip" })]
public void CanPassObjectArrayAsFirstArgument(object[] a)
{
}

The test does nothing but proves we can load and execute it
without an error. I like the convenience of your approach,
but some folks may insist on creating the array themselves.

Charlie