How to get the line count in a string using .NET (with any line break)

Question

I need to count the number of lines in a string. Any line break can be character can be present in the string (CR, LF or CRLF).

So possible new line chars:
* \n
* \r
* \r\n

For example, with the following input:

This is [\n]
an string that [\r]
has four [\r\n]
lines

The method should return 4 lines. Do you know any built in function, or someone already implemented it?

static int GetLineCount(string input)
{
   // could you provide a good implementation for this method?
   // I want to avoid string.split since it performs really bad
}

NOTE: Performance is important for me, because I could read large strings.

Answer 1

int count = 0;
int len = input.Length;
for(int i = 0; i != len; ++i)
  switch(input[i])
  {
    case '\r':
      ++count;
      if (i + 1 != len && input[i + 1] == '\n')
        ++i;
      break;
    case '\n':
    // Uncomment below to include all other line break sequences
    // case '\u000A':
    // case '\v':
    // case '\f':
    // case '\u0085':
    // case '\u2028':
    // case '\u2029':
      ++count;
      break;
  }

Simply scan through, counting the line-breaks, and in the case of \r test if the next character is \n and skip it if it is.

Performance is important for me, because I could read large strings.

If at all possible then, avoid reading large strings at all. E.g. if they come from streams this is pretty easy to do directly on a stream as there is no more than one-character read-ahead ever needed.

Here's another variant that doesn't count newlines at the very end of a string:

int count = 1;
int len = input.Length - 1;
for(int i = 0; i < len; ++i)
  switch(input[i])
  {
    case '\r':
    if (input[i + 1] == '\n')
    {
      if (++i >= len)
      {
        break;
      }
    }
    goto case '\n';
        case '\n':
        // Uncomment below to include all other line break sequences
        // case '\u000A':
        // case '\v':
        // case '\f':
        // case '\u0085':
        // case '\u2028':
        // case '\u2029':
          ++count;
          break;      
  }

This therefore considers "", "a line", "a line\n" and "a line\r\n" to each be one line only, and so on.