When the dust settles, I propose to rewrite UnicodeUtils.ToCharArray as follows:
public static char[] ToCharArray(int[] codePoints, int offset, int
count)
{
if (count < 0)
{
throw new System.ArgumentException();
}
// as a first approximation, assume each codepoint
is 1 character
char[] chars = new char[count];
int w = 0;
for (int r = offset, e = offset + count; r < e; ++r)
{
int cp = codePoints[r];
if (cp < 0 || cp > 0x10ffff)
{
throw new System.ArgumentException();
}
if (cp < 0x010000)
{
chars[w++] = (char)cp;
}
else
{
chars[w++] = (char)(LEAD_SURROGATE_OFFSET_ + (cp >>
LEAD_SURROGATE_SHIFT_));
// if we need more room,
add enough to hold 2-character codepoints for the rest of
// the string. Normally,
this resize operation will happen at most once.
if (w >= chars.Length)
Array.Resize(ref
chars, chars.Length + (e - r) * 2 - 1);
chars[w++] =
(char)(TRAIL_SURROGATE_MIN_VALUE + (cp & TRAIL_SURROGATE_MASK_));
}
}
// resize to the exact length: it's slightly faster
to check if the resize is needed
if (w != chars.Length)
Array.Resize(ref chars, w);
return chars;
}
This avoids exception overhead and at least one heap allocation. If no code
points generate 2 characters, the method doesn't allocate anything extra than
the result array.
Vincent
