private void HandlePrefixes(int start, int end, int ce32)
{
int index = Collation.IndexFromCE32(ce32);
ce32 = data.GetCE32FromContexts(index); // Default if no prefix match.
HandleCE32(start, end, ce32);
if (!addPrefixes)
{
return;
}
using (CharsTrie.Enumerator prefixes = new CharsTrie(data.contexts, index + 2).GetEnumerator())
{
while (prefixes.MoveNext())
{
var e = prefixes.Current;
SetPrefix(e.Chars);
// Prefix/pre-context mappings are special kinds of contractions
// that always yield expansions.
AddStrings(start, end, contractions);
AddStrings(start, end, expansions);
HandleCE32(start, end, e.Value);
}
}
ResetPrefix();
}
private bool GetCEsFromContractionCE32(CollationData data, int ce32)
{
int trieIndex = Collation.IndexFromCE32(ce32);
ce32 = data.GetCE32FromContexts(trieIndex); // Default if no suffix match.
// Since the original ce32 is not a prefix mapping,
// the default ce32 must not be another contraction.
Debug.Assert(!Collation.IsContractionCE32(ce32));
int contractionIndex = contractionCEs.Count;
if (GetCEsFromCE32(data, Collation.SentinelCodePoint, ce32))
{
AddContractionEntry(CollationFastLatin.CONTR_CHAR_MASK, ce0, ce1);
}
else
{
// Bail out for c-without-contraction.
AddContractionEntry(CollationFastLatin.CONTR_CHAR_MASK, Collation.NoCE, 0);
}
// Handle an encodable contraction unless the next contraction is too long
// and starts with the same character.
int prevX = -1;
bool addContraction = false;
using (CharsTrieEnumerator suffixes = CharsTrie.GetEnumerator(data.contexts, trieIndex + 2, 0))
{
while (suffixes.MoveNext())
{
CharsTrieEntry entry = suffixes.Current;
ICharSequence suffix = entry.Chars;
int x = CollationFastLatin.GetCharIndex(suffix[0]);
if (x < 0)
{
continue;
} // ignore anything but fast Latin text
if (x == prevX)
{
if (addContraction)
{
// Bail out for all contractions starting with this character.
AddContractionEntry(x, Collation.NoCE, 0);
addContraction = false;
}
continue;
}
if (addContraction)
{
AddContractionEntry(prevX, ce0, ce1);
}
ce32 = entry.Value;
if (suffix.Length == 1 && GetCEsFromCE32(data, Collation.SentinelCodePoint, ce32))
{
addContraction = true;
}
else
{
AddContractionEntry(x, Collation.NoCE, 0);
addContraction = false;
}
prevX = x;
}
}
if (addContraction)
{
AddContractionEntry(prevX, ce0, ce1);
}
// Note: There might not be any fast Latin contractions, but
// we need to enter contraction handling anyway so that we can bail out
// when there is a non-fast-Latin character following.
// For example: Danish &Y<<u+umlaut, when we compare Y vs. u\u0308 we need to see the
// following umlaut and bail out, rather than return the difference of Y vs. u.
ce0 = (Collation.NO_CE_PRIMARY << 32) | CONTRACTION_FLAG | (uint)contractionIndex;
ce1 = 0;
return(true);
}
private void Compare(int c, int ce32, int baseCE32)
{
if (Collation.IsPrefixCE32(ce32))
{
int dataIndex = Collation.IndexFromCE32(ce32);
ce32 = data.GetFinalCE32(data.GetCE32FromContexts(dataIndex));
if (Collation.IsPrefixCE32(baseCE32))
{
int baseIndex = Collation.IndexFromCE32(baseCE32);
baseCE32 = baseData.GetFinalCE32(baseData.GetCE32FromContexts(baseIndex));
ComparePrefixes(c, data.contexts, dataIndex + 2, baseData.contexts, baseIndex + 2);
}
else
{
AddPrefixes(data, c, data.contexts, dataIndex + 2);
}
}
else if (Collation.IsPrefixCE32(baseCE32))
{
int baseIndex = Collation.IndexFromCE32(baseCE32);
baseCE32 = baseData.GetFinalCE32(baseData.GetCE32FromContexts(baseIndex));
AddPrefixes(baseData, c, baseData.contexts, baseIndex + 2);
}
if (Collation.IsContractionCE32(ce32))
{
int dataIndex = Collation.IndexFromCE32(ce32);
if ((ce32 & Collation.CONTRACT_SINGLE_CP_NO_MATCH) != 0)
{
ce32 = Collation.NO_CE32;
}
else
{
ce32 = data.GetFinalCE32(data.GetCE32FromContexts(dataIndex));
}
if (Collation.IsContractionCE32(baseCE32))
{
int baseIndex = Collation.IndexFromCE32(baseCE32);
if ((baseCE32 & Collation.CONTRACT_SINGLE_CP_NO_MATCH) != 0)
{
baseCE32 = Collation.NO_CE32;
}
else
{
baseCE32 = baseData.GetFinalCE32(baseData.GetCE32FromContexts(baseIndex));
}
CompareContractions(c, data.contexts, dataIndex + 2, baseData.contexts, baseIndex + 2);
}
else
{
AddContractions(c, data.contexts, dataIndex + 2);
}
}
else if (Collation.IsContractionCE32(baseCE32))
{
int baseIndex = Collation.IndexFromCE32(baseCE32);
baseCE32 = baseData.GetFinalCE32(baseData.GetCE32FromContexts(baseIndex));
AddContractions(c, baseData.contexts, baseIndex + 2);
}
int tag;
if (Collation.IsSpecialCE32(ce32))
{
tag = Collation.TagFromCE32(ce32);
Debug.Assert(tag != Collation.PREFIX_TAG);
Debug.Assert(tag != Collation.CONTRACTION_TAG);
// Currently, the tailoring data builder does not write offset tags.
// They might be useful for saving space,
// but they would complicate the builder,
// and in tailorings we assume that performance of tailored characters is more important.
Debug.Assert(tag != Collation.OFFSET_TAG);
}
else
{
tag = -1;
}
int baseTag;
if (Collation.IsSpecialCE32(baseCE32))
{
baseTag = Collation.TagFromCE32(baseCE32);
Debug.Assert(baseTag != Collation.PREFIX_TAG);
Debug.Assert(baseTag != Collation.CONTRACTION_TAG);
}
else
{
baseTag = -1;
}
// Non-contextual mappings, expansions, etc.
if (baseTag == Collation.OFFSET_TAG)
{
// We might be comparing a tailoring CE which is a copy of
// a base offset-tag CE, via the [optimize [set]] syntax
// or when a single-character mapping was copied for tailored contractions.
// Offset tags always result in long-primary CEs,
// with common secondary/tertiary weights.
if (!Collation.IsLongPrimaryCE32(ce32))
{
Add(c);
return;
}
long dataCE = baseData.ces[Collation.IndexFromCE32(baseCE32)];
long p = Collation.GetThreeBytePrimaryForOffsetData(c, dataCE);
if (Collation.PrimaryFromLongPrimaryCE32(ce32) != p)
{
Add(c);
return;
}
}
if (tag != baseTag)
{
Add(c);
return;
}
if (tag == Collation.EXPANSION32_TAG)
{
int length = Collation.LengthFromCE32(ce32);
int baseLength = Collation.LengthFromCE32(baseCE32);
if (length != baseLength)
{
Add(c);
return;
}
int idx0 = Collation.IndexFromCE32(ce32);
int idx1 = Collation.IndexFromCE32(baseCE32);
for (int i = 0; i < length; ++i)
{
if (data.ce32s[idx0 + i] != baseData.ce32s[idx1 + i])
{
Add(c);
break;
}
}
}
else if (tag == Collation.EXPANSION_TAG)
{
int length = Collation.LengthFromCE32(ce32);
int baseLength = Collation.LengthFromCE32(baseCE32);
if (length != baseLength)
{
Add(c);
return;
}
int idx0 = Collation.IndexFromCE32(ce32);
int idx1 = Collation.IndexFromCE32(baseCE32);
for (int i = 0; i < length; ++i)
{
if (data.ces[idx0 + i] != baseData.ces[idx1 + i])
{
Add(c);
break;
}
}
}
else if (tag == Collation.HANGUL_TAG)
{
StringBuilder jamos = new StringBuilder();
int length = Hangul.Decompose(c, jamos);
if (tailored.Contains(jamos[0]) || tailored.Contains(jamos[1]) ||
(length == 3 && tailored.Contains(jamos[2])))
{
Add(c);
}
}
else if (ce32 != baseCE32)
{
Add(c);
}
}
