/// <summary>
/// Constructs for all variants of <see cref="Span(string, int, SpanCondition)"/>, or only for any one variant.
/// Initializes as little as possible, for single use.
/// </summary>
public UnicodeSetStringSpan(UnicodeSet set, IList <string> setStrings, int which)
{
spanSet = new UnicodeSet(0, 0x10ffff);
// TODO: With Java 6, just take the parent set's strings as is,
// as a NavigableSet<String>, rather than as an ArrayList copy of the set of strings.
// Then iterate via the first() and higher() methods.
// (We do not want to create multiple Iterator objects in each span().)
// See ICU ticket #7454.
strings = setStrings;
all = (which == All);
spanSet.RetainAll(set);
if (0 != (which & NotContained))
{
// Default to the same sets.
// addToSpanNotSet() will create a separate set if necessary.
spanNotSet = spanSet;
}
offsets = new OffsetList();
// Determine if the strings even need to be taken into account at all for span() etc.
// If any string is relevant, then all strings need to be used for
// span(longest match) but only the relevant ones for span(while contained).
// TODO: Possible optimization: Distinguish CONTAINED vs. LONGEST_MATCH
// and do not store UTF-8 strings if !thisRelevant and CONTAINED.
// (Only store irrelevant UTF-8 strings for LONGEST_MATCH where they are relevant after all.)
// Also count the lengths of the UTF-8 versions of the strings for memory allocation.
int stringsLength = strings.Count;
int i, spanLength;
int maxLength16 = 0;
someRelevant = false;
for (i = 0; i < stringsLength; ++i)
{
string str = strings[i];
int length16 = str.Length;
spanLength = spanSet.Span(str, SpanCondition.Contained);
if (spanLength < length16)
{ // Relevant string.
someRelevant = true;
}
if (/* (0 != (which & UTF16)) && */ length16 > maxLength16)
{
maxLength16 = length16;
}
}
this.maxLength16 = maxLength16;
if (!someRelevant && (which & WithCount) == 0)
{
return;
}
// Freeze after checking for the need to use strings at all because freezing
// a set takes some time and memory which are wasted if there are no relevant strings.
if (all)
{
spanSet.Freeze();
}
int spanBackLengthsOffset;
// Allocate a block of meta data.
int allocSize;
if (all)
{
// 2 sets of span lengths
allocSize = stringsLength * (2);
}
else
{
allocSize = stringsLength; // One set of span lengths.
}
spanLengths = new short[allocSize];
if (all)
{
// Store span lengths for all span() variants.
spanBackLengthsOffset = stringsLength;
}
else
{
// Store span lengths for only one span() variant.
spanBackLengthsOffset = 0;
}
// Set the meta data and spanNotSet and write the UTF-8 strings.
for (i = 0; i < stringsLength; ++i)
{
string str = strings[i];
int length16 = str.Length;
spanLength = spanSet.Span(str, SpanCondition.Contained);
if (spanLength < length16)
{ // Relevant string.
if (true /* 0 != (which & UTF16) */)
{
if (0 != (which & Contained))
{
if (0 != (which & Forward))
{
spanLengths[i] = MakeSpanLengthByte(spanLength);
}
if (0 != (which & Backward))
{
spanLength = length16
- spanSet.SpanBack(str, length16, SpanCondition.Contained);
spanLengths[spanBackLengthsOffset + i] = MakeSpanLengthByte(spanLength);
}
}
else /* not CONTAINED, not all, but NOT_CONTAINED */
{
spanLengths[i] = spanLengths[spanBackLengthsOffset + i] = 0; // Only store a relevant/irrelevant
// flag.
}
}
if (0 != (which & NotContained))
{
// Add string start and end code points to the spanNotSet so that
// a span(while not contained) stops before any string.
int c;
if (0 != (which & Forward))
{
c = str.CodePointAt(0);
AddToSpanNotSet(c);
}
if (0 != (which & Backward))
{
c = str.CodePointBefore(length16);
AddToSpanNotSet(c);
}
}
}
else
{ // Irrelevant string.
if (all)
{
spanLengths[i] = spanLengths[spanBackLengthsOffset + i] = ALL_CP_CONTAINED;
}
else
{
// All spanXYZLengths pointers contain the same address.
spanLengths[i] = ALL_CP_CONTAINED;
}
}
}
// Finish.
if (all)
{
spanNotSet.Freeze();
}
}