Ascii Bytes Array To Int32 or Double

Question

I'm re-writing alibrary with a mandate to make it totally allocation free. The goal is to have 0 collections after the app's startup phase is done.

Previously, there were a lot of calls like this:

Int32 foo = Int32.Parse(ASCIIEncoding.ASCII.GetString(bytes, start, length));

Which I believe is allocating a string. I couldn't find a C# library function that would do the same thing automatically. I looked at the BitConverter class, but it looks like that is only if your Int32 is encoded with the actual bytes that represent it. Here, I have an array of bytes representing Ascii characters that represent an Int32.

Here's what I did

public static Int32 AsciiBytesToInt32(byte[] bytes, int start, int length)
{
     Int32 Temp = 0;
     Int32 Result = 0;
     Int32 j = 1;

     for (int i = start + length - 1; i >= start; i--)
     {
          Temp = ((Int32)bytes[i]) - 48;

          if (Temp < 0 || Temp > 9)
          {
               throw new Exception("Bytes In AsciiBytesToInt32 Are Not An Int32");
          }

          Result += Temp * j;
          j *= 10;
     }

     return Result;
}

Does anyone know of a C# library function that already does this in a more optimal way? Or an improvement to make the above run faster (its going to be called millions of times during the day probably). Thanks!

Answer 1

Millions of times per day shouldn't be a problem - I'd expect that to be able to run hundreds of thousands of times per second. Personally I'd rewrite the above to only declare "temp" within the loop (and get rid of the Pascal-cases local variable names - urgh) but it should be okay.

The code would be more immediately understandable as:

int digit = bytes[i] - '0';

which does the same as your

Temp = ((Int32)bytes[i]) - 48;

line, but in a simpler way (IMO). They should behave exactly the same way.

On a general note, trying to write C# without any allocations is pretty harsh, and fights against the way the language and framework are designed. Do you believe this is actually a reasonable requirement? Admittedly I've heard about it being the way some games are written in managed code... but it does seem a bit odd.

Of course, you're going to allocate an exception if the bytes are inappropriate...

EDIT: Note that your code doesn't allow for negative numbers. Is that okay?