C#•2y ago

Parsing a string into a Vector of bytes most efficiently

Given a string with an even amount of characters, what's the most efficient way to parse that into a Vector<byte> along with another Vector<byte> of bit masks? A sample input is 12 34 ?? 78 87 ?5 ?3 21 -- these are meant to be hex bytes -- which would get turned into < 0x12 0x34 0x00 0x78 0x87 0x05 0x03 0x21 > for the values, and < 0xFF 0xFF 0x00 0xFF 0xFF 0x0F 0x0F 0xFF > for the masks. Currently I'm doing the input.Length % 2 != 0 check first, of course, followed by Regex.Matches(input, @"..").Select(match => match.Value).ToList();. From here I do this to parse all of the bytes;

int length = bytes.Length;
if (length < Vector<byte>.Count)
    length = Vector<byte>.Count;

byte[] values = new byte[length], masks = new byte[length];

for (int i = 0; i < bytes.Length; i++)
{
    var currByte = bytes[i];

    var hasUpper = byte.TryParse(currByte[0].ToString(), NumberStyles.HexNumber, null, out byte upper);
    var hasLower = byte.TryParse(currByte[1].ToString(), NumberStyles.HexNumber, null, out byte lower);

    switch ((hasUpper, hasLower))
    {
        case (true, true):
        {
            values[i] = (byte)((upper << 4) + lower);
            masks[i] = 0xFF;
            break;
        }

        case (true, false):
        {
            values[i] = (byte)(upper << 4);
            masks[i] = 0xF0;
            break;
        }

        case (false, true):
        {
            values[i] = lower;
            masks[i] = 0x0F;
            break;
        }

        case (false, false):
        {
            values[i] = 0x00;
            masks[i] = 0x00;
            break;
        }
    }
}

return (length, new(values), new(masks));

int length = bytes.Length;
if (length < Vector<byte>.Count)
    length = Vector<byte>.Count;

byte[] values = new byte[length], masks = new byte[length];

for (int i = 0; i < bytes.Length; i++)
{
    var currByte = bytes[i];

    var hasUpper = byte.TryParse(currByte[0].ToString(), NumberStyles.HexNumber, null, out byte upper);
    var hasLower = byte.TryParse(currByte[1].ToString(), NumberStyles.HexNumber, null, out byte lower);

    switch ((hasUpper, hasLower))
    {
        case (true, true):
        {
            values[i] = (byte)((upper << 4) + lower);
            masks[i] = 0xFF;
            break;
        }

        case (true, false):
        {
            values[i] = (byte)(upper << 4);
            masks[i] = 0xF0;
            break;
        }

        case (false, true):
        {
            values[i] = lower;
            masks[i] = 0x0F;
            break;
        }

        case (false, false):
        {
            values[i] = 0x00;
            masks[i] = 0x00;
            break;
        }
    }
}

return (length, new(values), new(masks));

Obviously the big one is using Regex. Any ideas?

6 Replies

TimberStalker•2y ago

what type is the bytes variable?

ero•2y ago

string[]. Using the example, the contents are [ "12", "34", "??", "78", "87", "?5", "?3", "21" ] watch me optimizing removing whitespace from a string

TimberStalker•2y ago

why use question marks instead of 0. Is that just for the mask?

ero•2y ago

for the mask, yeah it should be any non-hexdigit character

char* output = stackalloc char[INPUT.Length];

fixed (char* pInput = INPUT)
{
  for (int i = 0, j = 0; i < INPUT.Length; i++)
  {
    char c = *(pInput + i);

    if (!char.IsWhiteSpace(c))
      output[j++] = c;
  }

  return new string(output);
}

char* output = stackalloc char[INPUT.Length];

fixed (char* pInput = INPUT)
{
  for (int i = 0, j = 0; i < INPUT.Length; i++)
  {
    char c = *(pInput + i);

    if (!char.IsWhiteSpace(c))
      output[j++] = c;
  }

  return new string(output);
}

🙃

TimberStalker•2y ago

you can change the logic inside the for loop to this

var currByte = bytes[i];
for(int j = currByte.Length-1; j >= 0; j++)
{
  if(char.IsDigit(currByte[^(j + 1)])
  {
    values[i] += (currByte[^(j + 1)] - '0') << (4 * j);
    mask[i] += 0xF << (4 * j);
  }
}

var currByte = bytes[i];
for(int j = currByte.Length-1; j >= 0; j++)
{
  if(char.IsDigit(currByte[^(j + 1)])
  {
    values[i] += (currByte[^(j + 1)] - '0') << (4 * j);
    mask[i] += 0xF << (4 * j);
  }
}

if you want to be funky you can do

var currByte = bytes[i];
for(int j = 0; j < currByte.Length; j++)
{
  values[i] <<= 4;
  mask[i] <<= 4;
  if(char.IsDigit(currByte[j])
  {
    values[i] += (currByte[j] - '0');
    mask[i] += 0xF;
  }
}

var currByte = bytes[i];
for(int j = 0; j < currByte.Length; j++)
{
  values[i] <<= 4;
  mask[i] <<= 4;
  if(char.IsDigit(currByte[j])
  {
    values[i] += (currByte[j] - '0');
    mask[i] += 0xF;
  }
}

Instead

ero•2y ago

char.IsAsciiHexDigit if anything

Gaming

Programming

Parsing a string into a Vector of bytes most efficiently