/// <summary>
/// Given an ICU normalizer, enumerate the limit indices of the "segments" of this string.
/// A "segment" is defined as a group of characters that interact with each other in this
/// normalization, and which therefore can't be split apart and normalized separately without
/// changing the result of the normalization. For example, under NFC, if LATIN SMALL LETTER C (U+0063)
/// is followed by COMBINING CEDILLA (U+0327) which is followed by LATIN SMALL LETTER D (U+0064),
/// then the c and cedilla will form one "segment": splitting them apart and normalizing them
/// separately would produce a different result than normalizing them together. So this function
/// would yield (among other values) the index of LATIN SMALL LETTER D, the first index that is
/// not part of the segment (that is, the limit index).
///
/// The last index yielded by this function will be equal to the length of the string, and it
/// will never yield the index 0. (If the string is empty, it will return an empty enumerable).
/// Therefore, it is always safe to do GetChars(previousIndex, thisIndex) in a foreach loop to get
/// the "current" segment (assuming previousIndex is set to 0 the first time through the loop).
/// </summary>
/// <param name="icuNormalizer">IntPtr to the ICU normalizer to use (get this from Icu.GetIcuNormalizer)</param>
/// <returns>An enumerable of indexes into "this" TsString, at all the normalization "segment" boundaries, suitable for passing into GetChars(prevIdx, thisIdx)</returns>
private IEnumerable <int> EnumerateSegmentLimits(IntPtr icuNormalizer)
{
if (String.IsNullOrEmpty(Text))
{
yield break;
}
int i = 0;
while (i < Text.Length)
{
int codepoint = Char.ConvertToUtf32(Text, i);
if (Icu.HasNormalizationBoundaryBefore(icuNormalizer, codepoint) && i > 0)
{
yield return(i);
}
i += codepoint > 0xffff ? 2 : 1;
}
yield return(Text.Length);
}
