Rookie query about short and byte arithmetic/bitwise ops

VS19 16.9.688.6828
Writing a simple C# app as an exercise to try to get my brain working after six years I hit a problem I don't recall. I ran some simple code and it seems that C# doesn't allow arithmetic and bitwise ops on 8 and 16 bit numbers viz:

Code:

---

        ushort a = 0xf;
        ushort b = 0x7;
        ushort c = 0x3;
        a = b + c; // error
        a = b & c; // error
        a = a + b; // error
        a += b;    // no error
        a = (ushort)((int)a + (int)b);  // no error
//implicit cast
        uint d = 0x1;
        d = (uint)(a + b);  // no error
        d = d + a;  // no error

---

I can't believe this is a bug so can somebody please explain the thinking behind this apparent inconsistency?
No luck finding an explanation in VS19 help!

It seems that the result of all these arithmetic operations with int/uint, short/ushort, byte types will be always of the type int.
Therefore you will have to explicitly cast the result to the type you need to:

Code:

---

        ushort a = 0xf;
        ushort b = 0x7;
        ushort c = 0x3;
        a = b + c; // error
        a = (ushort)(b + c);  // no error

---

Strange!
I have gone back to VC# 2010 and that was the same so M.S. must know and presumably have a reason so I suppose I will have to ask them but thanks VictorN anyway.

Quote:

---

Originally Posted by RogerD

M.S. must know and presumably have a reason

---

They know, and they call it the C# type promotion rules. It can give "somewhat unexpected results" mentioned here,

https://prasadiapsara.wordpress.com/...omotion-rules/

To avoid such problems, I prefer to use only either int or double in expressions, whichever best fits the bill. It is seldom necessary to use something else.

Thanks Wolle. The project I was working on (some six years back) is for processing and displaying 16 bit astronomical images which are stored in 4x 6Mpixel 16 bit arrays. Process involves stretching, stacking, registering, filtering (hot pixel, unsharp, gaussian, con/satellite trail removal etc.) and of course diplay. I seem to remember that it got rather slow when I converted from C++ and if C# is converting everything to 32 bit that could explain it. From my perspective if there's no way round this it's a big downside to C#.