Optional (or null) output parameters

  • Thread starter Thread starter Ken Durden
  • Start date Start date
K

Ken Durden

I'm looking for a way to simulate the following kind of C code:

--- C++ ---
void F( double * array, double * fMean, double * fMedian, double *
fMax, double * fMin );

--- C# ---
void F( double[] array, out double fMean, out double fMedian, out
double fMax, out double fMin );

I've read some of the examples on using overrides to simulate default
values, but that's not exactly what I'm doing here. On the client
side, NULL would be specified to indicate that (for example) mean and
median aren't needed, but max and min are.

Now, I can certainly use use overloads to have 2^4 -1 versions of this
function. This would provide a maximally efficient way to calculate
all these statistics; however, I don't need that, and the code
duplication would kill me. Furthermore, the function above is just an
example of the concept, the real code is more complex, of course.

I could also pass a boolean next to each out parameter indicating
whether to calculate it or not.

I could also pass a structure, etc...

I was kinda hoping I could pass in null for one of those output
parameters, but then I realized there wouldn't be any way to check on
the other side.

I could also declare all the parameters as objects, pass in nulls for
the ones I don't want, unbox the ones which were requested.

Any other options? Which one of the above makes the most sense?

Many Thanks,
-ken
 
Ken,
I could also declare all the parameters as objects, pass in nulls for
the ones I don't want, unbox the ones which were requested.

I would consider using a single Parameter Object. Either an actual object
with multiple fields (Mean, Median, Max, Min, WantMean, WantMedian, WantMax,
WantMin) or an extra enum flags parameter that indicates which values I am
interested in.

Something like:
[Flags]
enum StatsWanted
{
Mean = 1,
Median = 2,
Max = 4,
Min = 8
};

void F( double[] array, StatsWanted statsWanted,
out double fMean, out double fMedian,
out double fMax, out double fMin );

Then when I called it I would include the appropriate bits for what I was
interested in.

Hope this helps
Jay

Ken Durden said:
I'm looking for a way to simulate the following kind of C code:

--- C++ ---
void F( double * array, double * fMean, double * fMedian, double *
fMax, double * fMin );

--- C# ---
void F( double[] array, out double fMean, out double fMedian, out
double fMax, out double fMin );

I've read some of the examples on using overrides to simulate default
values, but that's not exactly what I'm doing here. On the client
side, NULL would be specified to indicate that (for example) mean and
median aren't needed, but max and min are.

Now, I can certainly use use overloads to have 2^4 -1 versions of this
function. This would provide a maximally efficient way to calculate
all these statistics; however, I don't need that, and the code
duplication would kill me. Furthermore, the function above is just an
example of the concept, the real code is more complex, of course.

I could also pass a boolean next to each out parameter indicating
whether to calculate it or not.

I could also pass a structure, etc...

I was kinda hoping I could pass in null for one of those output
parameters, but then I realized there wouldn't be any way to check on
the other side.

I could also declare all the parameters as objects, pass in nulls for
the ones I don't want, unbox the ones which were requested.

Any other options? Which one of the above makes the most sense?

Many Thanks,
-ken
 
Hi,
you do not have to use out or ref if all your arguments are reference type.
In C# all reference types are passed by reference by default.
So in the caller you have to declare some objects and assign them values.
For some of the objects you may assign null.
And in your code you may check for null, change the values, etc.
And no overolads :)

I mean:

void caller()
{
int[] res, par1, par2, par3, par4;

par1 = new int[10];
// fill values
par2 = new int[5];
par3 = null;
par4 = null;
res = null;

myfunc(res, par1, par2, par, par4);
if (res != null)
//display results
{}
}

void myfunc(int[] r, int[] a1, int[] a2, int[] a3, int[] a4)
{
if ((a1 != null) && (a2 != null))
{
r = new int[15];
// do calcs
}
else
{
r = new int[10];
//other calcs, or check for a3 and a4, etc...
}
}

home that helps.
Sunny

P.S. watch out for typos, I've write directly in the reader, haven't tested,
but it shoul work that way.
 
Unfortunately, fundamental types like double are value types and hence
cannot be null. There are several ways round it.

1) Pass the doubles as objects. An object can be null, or you can pass a
double which will be boxed into the object. On the downside this is messy
and not type safe.

2) Create a class (hence a reference type) that encapsulates a double (a bit
like the Java class Double). This will give you type safety, and the ability
to pass a null object.

Those are the most direct translations of the C++. Personally I'd return a
structure and pass flags to say what needs to be calculated.

Regards,

Jasper Kent.
 
Sunny said:
you do not have to use out or ref if all your arguments are
reference type.

You do sometimes.
In C# all reference types are passed by reference by default.

No they're not. There's a difference between passing a reference by
value and passing a value by reference. (Or a reference by reference.)

Calling the default C# reference type parameter passing mechanism "pass
by reference" is just wrong, and can be a source of confusion.

See http://www.pobox.com/~skeet/csharp/parameters.html
 
Back
Top