Obtaining real types from variants

[therealwill]

I'm looping through an image, reading the pixel values and manipulating them. The pixel value can only be returned as a variant, therefore I'm going to need to use variant.dblVal or .iVal etc to get the value. My program will however encounter most real types, therefore I will need to use .dblVal or .iVal depending on the image. Is there a way of doing this without a hefty switch statement inside my image loop getting in the way? i.e. obtaining the value just using a single statement?

Any help appreciated.

Will

long i, j;
    long ij[2];
    VARIANT v;
    double getvalM;
    ipInRasProps->get_NoDataValue(&vInnodatavalue);
    ipOutputRasProps->get_NoDataValue(&vOutNodatavalue);
	
	for (i=0; i<(lHeight - 1); i++)
	{
	ij[0]=i;
                  for (j=0; j<(lWidth - 1); j++)
		 {
		 ij[1]=j;
                 ipInputBlock3->GetVal(0, i, j, &v);
		 getvalM = v.iVal;  //here I'd have a switch statement 
 
		if ((j + 1) < (lWidth))
	        SafeArrayPutElement(vSrcPixelArray.parray, ij, &v);
			
							
		}
                
		}
	}
    ipOutputBlock = ipOutputBlock3;
	ipOutRawPixel->Write(pPnt, ipOutputBlock);

[iMalc]

What on earth kind of ultra-slow API are you using that forces you to painstakingly retrieve individual pixel values as VARIANTs one at a time?!?!

[Ancient Dragon]

>>Is there a way of doing this without a hefty switch statement inside my image loop getting in the way? i.e. obtaining the value just using a single statement?

Yes, but it would not save any time at all because you still must determine from .vt whether it is VT_REAL, VT_I4 or whatever. If you know there are only two possible data types then you can do like below. If other data types are possible then a switch statement would probably be more efficient.

 getvalM = (vt.vt == VT_I4) ? v.iVal : v.dblVal;

If you are looking at ways to speed up that algorithm you posted, your best efforts would be spent by replacing SafeArrayPutElement() because it locks the array, puts the data into the array, the unlocks it again. That is awful time consuming. You can lock it at the beginning of the loop, add each element in your code then unlock it at the end.

[ElastoManiac]

very slow algorithm ... if you could even call it that.
But anyway you could try using unions. Google it.

[Ancient Dragon]

very slow algorithm ... if you could even call it that.
But anyway you could try using unions. Google it.

A VATIANT is a union.

[therealwill]

your best efforts would be spent by replacing SafeArrayPutElement() because it locks the array, puts the data into the array, the unlocks it again. That is awful time consuming. You can lock it at the beginning of the loop, add each element in your code then unlock it at the end.

Thanks, I'll look into that. Any further tips on safearray functions?

I'll know the variant type before the loop if that makes any difference.

What on earth kind of ultra-slow API are you using that forces you to painstakingly retrieve individual pixel values as VARIANTs one at a time?!?!

ESRI's ArcGIS. You should see the amount of code required to unwrap the raw pixels.

Thanks for the help.

Will

[therealwill]

I Can't find a function to input data to the safearray once it is locked, other than safearrayputelement - any ideas?