public override bool Equals(object other)
{
if (other == null)
{
return(false);
}
if (!this.GetType().Equals(other.GetType()))
{
return(false);
}
CollationSettings o = (CollationSettings)other;
if (options != o.options)
{
return(false);
}
if ((options & AlternateMask) != 0 && variableTop != o.variableTop)
{
return(false);
}
if (!ArrayEqualityComparer <int> .OneDimensional.Equals(reorderCodes, o.reorderCodes))
{
return(false);
}
return(true);
}
public override bool Equals(object other)
{
if (other == null)
{
return(false);
}
if (!this.GetType().Equals(other.GetType()))
{
return(false);
}
CollationSettings o = (CollationSettings)other;
if (options != o.options)
{
return(false);
}
if ((options & ALTERNATE_MASK) != 0 && variableTop != o.variableTop)
{
return(false);
}
if (!Arrays.Equals(reorderCodes, o.reorderCodes))
{
return(false);
}
return(true);
}
public override object Clone()
{
CollationSettings newSettings = (CollationSettings)base.Clone();
// Note: The reorderTable, reorderRanges, and reorderCodes need not be cloned
// because, in Java, they only get replaced but not modified.
newSettings.fastLatinPrimaries = (char[])fastLatinPrimaries.Clone();
return(newSettings);
}
public void CopyReorderingFrom(CollationSettings other)
{
if (!other.HasReordering)
{
ResetReordering();
return;
}
minHighNoReorder = other.minHighNoReorder;
reorderTable = other.reorderTable;
reorderRanges = other.reorderRanges;
reorderCodes = other.reorderCodes;
}
/**
* Writes the sort key bytes for minLevel up to the iterator data's strength. Optionally writes
* the case level. Stops writing levels when callback.needToWrite(level) returns false.
* Separates levels with the LEVEL_SEPARATOR_BYTE but does not write a TERMINATOR_BYTE.
*/
public static void WriteSortKeyUpToQuaternary(CollationIterator iter, bool[] compressibleBytes,
CollationSettings settings, SortKeyByteSink sink, int minLevel, LevelCallback callback,
bool preflight)
{
int options = settings.Options;
// Set of levels to process and write.
int levels = levelMasks[(int)CollationSettings.GetStrength(options)];
if ((options & CollationSettings.CASE_LEVEL) != 0)
{
levels |= Collation.CASE_LEVEL_FLAG;
}
// Minus the levels below minLevel.
levels &= ~((1 << minLevel) - 1);
if (levels == 0)
{
return;
}
long variableTop;
if ((options & CollationSettings.ALTERNATE_MASK) == 0)
{
variableTop = 0;
}
else
{
// +1 so that we can use "<" and primary ignorables test out early.
variableTop = settings.VariableTop + 1;
}
int tertiaryMask = CollationSettings.GetTertiaryMask(options);
byte[] p234 = new byte[3];
SortKeyLevel cases = GetSortKeyLevel(levels, Collation.CASE_LEVEL_FLAG);
SortKeyLevel secondaries = GetSortKeyLevel(levels, Collation.SECONDARY_LEVEL_FLAG);
SortKeyLevel tertiaries = GetSortKeyLevel(levels, Collation.TERTIARY_LEVEL_FLAG);
SortKeyLevel quaternaries = GetSortKeyLevel(levels, Collation.QUATERNARY_LEVEL_FLAG);
long prevReorderedPrimary = 0; // 0==no compression
int commonCases = 0;
int commonSecondaries = 0;
int commonTertiaries = 0;
int commonQuaternaries = 0;
int prevSecondary = 0;
int secSegmentStart = 0;
for (; ;)
{
// No need to keep all CEs in the buffer when we write a sort key.
iter.ClearCEsIfNoneRemaining();
long ce = iter.NextCE();
long p = ce.TripleShift(32);
if (p < variableTop && p > Collation.MERGE_SEPARATOR_PRIMARY)
{
// Variable CE, shift it to quaternary level.
// Ignore all following primary ignorables, and shift further variable CEs.
if (commonQuaternaries != 0)
{
--commonQuaternaries;
while (commonQuaternaries >= QUAT_COMMON_MAX_COUNT)
{
quaternaries.AppendByte(QUAT_COMMON_MIDDLE);
commonQuaternaries -= QUAT_COMMON_MAX_COUNT;
}
// Shifted primary weights are lower than the common weight.
quaternaries.AppendByte(QUAT_COMMON_LOW + commonQuaternaries);
commonQuaternaries = 0;
}
do
{
if ((levels & Collation.QUATERNARY_LEVEL_FLAG) != 0)
{
if (settings.HasReordering)
{
p = settings.Reorder(p);
}
if (((int)p.TripleShift(24)) >= QUAT_SHIFTED_LIMIT_BYTE)
{
// Prevent shifted primary lead bytes from
// overlapping with the common compression range.
quaternaries.AppendByte(QUAT_SHIFTED_LIMIT_BYTE);
}
quaternaries.AppendWeight32(p);
}
do
{
ce = iter.NextCE();
p = ce.TripleShift(32);
} while (p == 0);
} while (p < variableTop && p > Collation.MERGE_SEPARATOR_PRIMARY);
}
// ce could be primary ignorable, or NO_CE, or the merge separator,
// or a regular primary CE, but it is not variable.
// If ce==NO_CE, then write nothing for the primary level but
// terminate compression on all levels and then exit the loop.
if (p > Collation.NO_CE_PRIMARY && (levels & Collation.PRIMARY_LEVEL_FLAG) != 0)
{
// Test the un-reordered primary for compressibility.
bool isCompressible = compressibleBytes[(int)p.TripleShift(24)];
if (settings.HasReordering)
{
p = settings.Reorder(p);
}
int p1 = (int)p.TripleShift(24);
if (!isCompressible || p1 != ((int)prevReorderedPrimary.TripleShift(24)))
{
if (prevReorderedPrimary != 0)
{
if (p < prevReorderedPrimary)
{
// No primary compression terminator
// at the end of the level or merged segment.
if (p1 > Collation.MERGE_SEPARATOR_BYTE)
{
sink.Append(Collation.PRIMARY_COMPRESSION_LOW_BYTE);
}
}
else
{
sink.Append(Collation.PRIMARY_COMPRESSION_HIGH_BYTE);
}
}
sink.Append(p1);
if (isCompressible)
{
prevReorderedPrimary = p;
}
else
{
prevReorderedPrimary = 0;
}
}
byte p2 = (byte)(p.TripleShift(16));
if (p2 != 0)
{
p234[0] = p2;
p234[1] = (byte)(p.TripleShift(8));
p234[2] = (byte)p;
sink.Append(p234, (p234[1] == 0) ? 1 : (p234[2] == 0) ? 2 : 3);
}
// Optimization for internalNextSortKeyPart():
// When the primary level overflows we can stop because we need not
// calculate (preflight) the whole sort key length.
if (!preflight && sink.Overflowed)
{
// not used in Java -- if (!sink.IsOk()) {
// Java porting note: U_MEMORY_ALLOCATION_ERROR is set here in
// C implementation. IsOk() in Java always returns true, so this
// is a dead code.
return;
}
}
int lower32 = (int)ce;
if (lower32 == 0)
{
continue;
} // completely ignorable, no secondary/case/tertiary/quaternary
if ((levels & Collation.SECONDARY_LEVEL_FLAG) != 0)
{
int s = lower32.TripleShift(16); // 16 bits
if (s == 0)
{
// secondary ignorable
}
else if (s == Collation.COMMON_WEIGHT16 &&
((options & CollationSettings.BACKWARD_SECONDARY) == 0 ||
p != Collation.MERGE_SEPARATOR_PRIMARY))
{
// s is a common secondary weight, and
// backwards-secondary is off or the ce is not the merge separator.
++commonSecondaries;
}
else if ((options & CollationSettings.BACKWARD_SECONDARY) == 0)
{
if (commonSecondaries != 0)
{
--commonSecondaries;
while (commonSecondaries >= SEC_COMMON_MAX_COUNT)
{
secondaries.AppendByte(SEC_COMMON_MIDDLE);
commonSecondaries -= SEC_COMMON_MAX_COUNT;
}
int b;
if (s < Collation.COMMON_WEIGHT16)
{
b = SEC_COMMON_LOW + commonSecondaries;
}
else
{
b = SEC_COMMON_HIGH - commonSecondaries;
}
secondaries.AppendByte(b);
commonSecondaries = 0;
}
secondaries.AppendWeight16(s);
}
else
{
if (commonSecondaries != 0)
{
--commonSecondaries;
// Append reverse weights. The level will be re-reversed later.
int remainder = commonSecondaries % SEC_COMMON_MAX_COUNT;
int b;
if (prevSecondary < Collation.COMMON_WEIGHT16)
{
b = SEC_COMMON_LOW + remainder;
}
else
{
b = SEC_COMMON_HIGH - remainder;
}
secondaries.AppendByte(b);
commonSecondaries -= remainder;
// commonSecondaries is now a multiple of SEC_COMMON_MAX_COUNT.
while (commonSecondaries > 0)
{ // same as >= SEC_COMMON_MAX_COUNT
secondaries.AppendByte(SEC_COMMON_MIDDLE);
commonSecondaries -= SEC_COMMON_MAX_COUNT;
}
// commonSecondaries == 0
}
if (0 < p && p <= Collation.MERGE_SEPARATOR_PRIMARY)
{
// The backwards secondary level compares secondary weights backwards
// within segments separated by the merge separator (U+FFFE).
byte[] secs = secondaries.Data();
int last = secondaries.Length - 1;
while (secSegmentStart < last)
{
byte b = secs[secSegmentStart];
secs[secSegmentStart++] = secs[last];
secs[last--] = b;
}
secondaries.AppendByte(p == Collation.NO_CE_PRIMARY ?
Collation.LEVEL_SEPARATOR_BYTE : Collation.MERGE_SEPARATOR_BYTE);
prevSecondary = 0;
secSegmentStart = secondaries.Length;
}
else
{
secondaries.AppendReverseWeight16(s);
prevSecondary = s;
}
}
}
if ((levels & Collation.CASE_LEVEL_FLAG) != 0)
{
if ((CollationSettings.GetStrength(options) == (int)CollationStrength.Primary) ? p == 0
: (lower32.TripleShift(16)) == 0)
{
// Primary+caseLevel: Ignore case level weights of primary ignorables.
// Otherwise: Ignore case level weights of secondary ignorables.
// For details see the comments in the CollationCompare class.
}
else
{
int c = (lower32.TripleShift(8)) & 0xff; // case bits & tertiary lead byte
Debug.Assert((c & 0xc0) != 0xc0);
if ((c & 0xc0) == 0 && c > Collation.LEVEL_SEPARATOR_BYTE)
{
++commonCases;
}
else
{
if ((options & CollationSettings.UPPER_FIRST) == 0)
{
// lowerFirst: Compress common weights to nibbles 1..7..13, mixed=14,
// upper=15.
// If there are only common (=lowest) weights in the whole level,
// then we need not write anything.
// Level length differences are handled already on the next-higher level.
if (commonCases != 0 &&
(c > Collation.LEVEL_SEPARATOR_BYTE || !cases.IsEmpty))
{
--commonCases;
while (commonCases >= CASE_LOWER_FIRST_COMMON_MAX_COUNT)
{
cases.AppendByte(CASE_LOWER_FIRST_COMMON_MIDDLE << 4);
commonCases -= CASE_LOWER_FIRST_COMMON_MAX_COUNT;
}
int b;
if (c <= Collation.LEVEL_SEPARATOR_BYTE)
{
b = CASE_LOWER_FIRST_COMMON_LOW + commonCases;
}
else
{
b = CASE_LOWER_FIRST_COMMON_HIGH - commonCases;
}
cases.AppendByte(b << 4);
commonCases = 0;
}
if (c > Collation.LEVEL_SEPARATOR_BYTE)
{
c = (CASE_LOWER_FIRST_COMMON_HIGH + (c.TripleShift(6))) << 4; // 14 or 15
}
}
else
{
// upperFirst: Compress common weights to nibbles 3..15, mixed=2,
// upper=1.
// The compressed common case weights only go up from the "low" value
// because with upperFirst the common weight is the highest one.
if (commonCases != 0)
{
--commonCases;
while (commonCases >= CASE_UPPER_FIRST_COMMON_MAX_COUNT)
{
cases.AppendByte(CASE_UPPER_FIRST_COMMON_LOW << 4);
commonCases -= CASE_UPPER_FIRST_COMMON_MAX_COUNT;
}
cases.AppendByte((CASE_UPPER_FIRST_COMMON_LOW + commonCases) << 4);
commonCases = 0;
}
if (c > Collation.LEVEL_SEPARATOR_BYTE)
{
c = (CASE_UPPER_FIRST_COMMON_LOW - (c.TripleShift(6))) << 4; // 2 or 1
}
}
// c is a separator byte 01,
// or a left-shifted nibble 0x10, 0x20, ... 0xf0.
cases.AppendByte(c);
}
}
}
if ((levels & Collation.TERTIARY_LEVEL_FLAG) != 0)
{
int t = lower32 & tertiaryMask;
Debug.Assert((lower32 & 0xc000) != 0xc000);
if (t == Collation.COMMON_WEIGHT16)
{
++commonTertiaries;
}
else if ((tertiaryMask & 0x8000) == 0)
{
// Tertiary weights without case bits.
// Move lead bytes 06..3F to C6..FF for a large common-weight range.
if (commonTertiaries != 0)
{
--commonTertiaries;
while (commonTertiaries >= TER_ONLY_COMMON_MAX_COUNT)
{
tertiaries.AppendByte(TER_ONLY_COMMON_MIDDLE);
commonTertiaries -= TER_ONLY_COMMON_MAX_COUNT;
}
int b;
if (t < Collation.COMMON_WEIGHT16)
{
b = TER_ONLY_COMMON_LOW + commonTertiaries;
}
else
{
b = TER_ONLY_COMMON_HIGH - commonTertiaries;
}
tertiaries.AppendByte(b);
commonTertiaries = 0;
}
if (t > Collation.COMMON_WEIGHT16)
{
t += 0xc000;
}
tertiaries.AppendWeight16(t);
}
else if ((options & CollationSettings.UPPER_FIRST) == 0)
{
// Tertiary weights with caseFirst=lowerFirst.
// Move lead bytes 06..BF to 46..FF for the common-weight range.
if (commonTertiaries != 0)
{
--commonTertiaries;
while (commonTertiaries >= TER_LOWER_FIRST_COMMON_MAX_COUNT)
{
tertiaries.AppendByte(TER_LOWER_FIRST_COMMON_MIDDLE);
commonTertiaries -= TER_LOWER_FIRST_COMMON_MAX_COUNT;
}
int b;
if (t < Collation.COMMON_WEIGHT16)
{
b = TER_LOWER_FIRST_COMMON_LOW + commonTertiaries;
}
else
{
b = TER_LOWER_FIRST_COMMON_HIGH - commonTertiaries;
}
tertiaries.AppendByte(b);
commonTertiaries = 0;
}
if (t > Collation.COMMON_WEIGHT16)
{
t += 0x4000;
}
tertiaries.AppendWeight16(t);
}
else
{
// Tertiary weights with caseFirst=upperFirst.
// Do not change the artificial uppercase weight of a tertiary CE (0.0.ut),
// to keep tertiary CEs well-formed.
// Their case+tertiary weights must be greater than those of
// primary and secondary CEs.
//
// Separator 01 -> 01 (unchanged)
// Lowercase 02..04 -> 82..84 (includes uncased)
// Common weight 05 -> 85..C5 (common-weight compression range)
// Lowercase 06..3F -> C6..FF
// Mixed case 42..7F -> 42..7F
// Uppercase 82..BF -> 02..3F
// Tertiary CE 86..BF -> C6..FF
if (t <= Collation.NO_CE_WEIGHT16)
{
// Keep separators unchanged.
}
else if ((lower32.TripleShift(16)) != 0)
{
// Invert case bits of primary & secondary CEs.
t ^= 0xc000;
if (t < (TER_UPPER_FIRST_COMMON_HIGH << 8))
{
t -= 0x4000;
}
}
else
{
// Keep uppercase bits of tertiary CEs.
Debug.Assert(0x8600 <= t && t <= 0xbfff);
t += 0x4000;
}
if (commonTertiaries != 0)
{
--commonTertiaries;
while (commonTertiaries >= TER_UPPER_FIRST_COMMON_MAX_COUNT)
{
tertiaries.AppendByte(TER_UPPER_FIRST_COMMON_MIDDLE);
commonTertiaries -= TER_UPPER_FIRST_COMMON_MAX_COUNT;
}
int b;
if (t < (TER_UPPER_FIRST_COMMON_LOW << 8))
{
b = TER_UPPER_FIRST_COMMON_LOW + commonTertiaries;
}
else
{
b = TER_UPPER_FIRST_COMMON_HIGH - commonTertiaries;
}
tertiaries.AppendByte(b);
commonTertiaries = 0;
}
tertiaries.AppendWeight16(t);
}
}
if ((levels & Collation.QUATERNARY_LEVEL_FLAG) != 0)
{
int q = lower32 & 0xffff;
if ((q & 0xc0) == 0 && q > Collation.NO_CE_WEIGHT16)
{
++commonQuaternaries;
}
else if (q == Collation.NO_CE_WEIGHT16 &&
(options & CollationSettings.ALTERNATE_MASK) == 0 &&
quaternaries.IsEmpty)
{
// If alternate=non-ignorable and there are only common quaternary weights,
// then we need not write anything.
// The only weights greater than the merge separator and less than the common
// weight
// are shifted primary weights, which are not generated for
// alternate=non-ignorable.
// There are also exactly as many quaternary weights as tertiary weights,
// so level length differences are handled already on tertiary level.
// Any above-common quaternary weight will compare greater regardless.
quaternaries.AppendByte(Collation.LEVEL_SEPARATOR_BYTE);
}
else
{
if (q == Collation.NO_CE_WEIGHT16)
{
q = Collation.LEVEL_SEPARATOR_BYTE;
}
else
{
q = 0xfc + ((q.TripleShift(6)) & 3);
}
if (commonQuaternaries != 0)
{
--commonQuaternaries;
while (commonQuaternaries >= QUAT_COMMON_MAX_COUNT)
{
quaternaries.AppendByte(QUAT_COMMON_MIDDLE);
commonQuaternaries -= QUAT_COMMON_MAX_COUNT;
}
int b;
if (q < QUAT_COMMON_LOW)
{
b = QUAT_COMMON_LOW + commonQuaternaries;
}
else
{
b = QUAT_COMMON_HIGH - commonQuaternaries;
}
quaternaries.AppendByte(b);
commonQuaternaries = 0;
}
quaternaries.AppendByte(q);
}
}
if ((lower32.TripleShift(24)) == Collation.LEVEL_SEPARATOR_BYTE)
{
break;
} // ce == NO_CE
}
// Append the beyond-primary levels.
// not used in Java -- boolean ok = true;
if ((levels & Collation.SECONDARY_LEVEL_FLAG) != 0)
{
if (!callback.NeedToWrite(Collation.SECONDARY_LEVEL))
{
return;
}
// not used in Java -- ok &= secondaries.isOk();
sink.Append(Collation.LEVEL_SEPARATOR_BYTE);
secondaries.AppendTo(sink);
}
if ((levels & Collation.CASE_LEVEL_FLAG) != 0)
{
if (!callback.NeedToWrite(Collation.CASE_LEVEL))
{
return;
}
// not used in Java -- ok &= cases.isOk();
sink.Append(Collation.LEVEL_SEPARATOR_BYTE);
// Write pairs of nibbles as bytes, except separator bytes as themselves.
int length = cases.Length - 1; // Ignore the trailing NO_CE.
byte b = 0;
for (int i = 0; i < length; ++i)
{
byte c = cases.GetAt(i);
Debug.Assert((c & 0xf) == 0 && c != 0);
if (b == 0)
{
b = c;
}
else
{
sink.Append(b | ((c >> 4) & 0xf));
b = 0;
}
}
if (b != 0)
{
sink.Append(b);
}
}
if ((levels & Collation.TERTIARY_LEVEL_FLAG) != 0)
{
if (!callback.NeedToWrite(Collation.TERTIARY_LEVEL))
{
return;
}
// not used in Java -- ok &= tertiaries.isOk();
sink.Append(Collation.LEVEL_SEPARATOR_BYTE);
tertiaries.AppendTo(sink);
}
if ((levels & Collation.QUATERNARY_LEVEL_FLAG) != 0)
{
if (!callback.NeedToWrite(Collation.QUATERNARY_LEVEL))
{
return;
}
// not used in Java -- ok &= quaternaries.isOk();
sink.Append(Collation.LEVEL_SEPARATOR_BYTE);
quaternaries.AppendTo(sink);
}
// not used in Java -- if (!ok || !sink.IsOk()) {
// Java porting note: U_MEMORY_ALLOCATION_ERROR is set here in
// C implementation. IsOk() in Java always returns true, so this
// is a dead code.
}
public static int CompareUTF16(char[] table, char[] primaries, int options,
ICharSequence left, ICharSequence right, int startIndex)
{
// This is a modified copy of CollationCompare.compareUpToQuaternary(),
// optimized for common Latin text.
// Keep them in sync!
int variableTop = options >> 16; // see getOptions()
options &= 0xffff; // needed for CollationSettings.getStrength() to work
// Check for supported characters, fetch mini CEs, and compare primaries.
int leftIndex = startIndex, rightIndex = startIndex;
// Single mini CE or a pair.
// The current mini CE is in the lower 16 bits, the next one is in the upper 16 bits.
// If there is only one, then it is in the lower bits, and the upper bits are 0.
int leftPair = 0, rightPair = 0;
for (; ;)
{
// We fetch CEs until we get a non-ignorable primary or reach the end.
while (leftPair == 0)
{
if (leftIndex == left.Length)
{
leftPair = EOS;
break;
}
int c = left[leftIndex++];
if (c <= LatinMax)
{
leftPair = primaries[c];
if (leftPair != 0)
{
break;
}
if (c <= 0x39 && c >= 0x30 && (options & CollationSettings.Numeric) != 0)
{
return(BailOutResult);
}
leftPair = table[c];
}
else if (PUNCT_START <= c && c < PUNCT_LIMIT)
{
leftPair = table[c - PUNCT_START + LatinLimit];
}
else
{
leftPair = Lookup(table, c);
}
if (leftPair >= MIN_SHORT)
{
leftPair &= SHORT_PRIMARY_MASK;
break;
}
else if (leftPair > variableTop)
{
leftPair &= LONG_PRIMARY_MASK;
break;
}
else
{
long pairAndInc = NextPair(table, c, leftPair, left, leftIndex);
if (pairAndInc < 0)
{
++leftIndex;
pairAndInc = ~pairAndInc;
}
leftPair = (int)pairAndInc;
if (leftPair == BAIL_OUT)
{
return(BailOutResult);
}
leftPair = GetPrimaries(variableTop, leftPair);
}
}
while (rightPair == 0)
{
if (rightIndex == right.Length)
{
rightPair = EOS;
break;
}
int c = right[rightIndex++];
if (c <= LatinMax)
{
rightPair = primaries[c];
if (rightPair != 0)
{
break;
}
if (c <= 0x39 && c >= 0x30 && (options & CollationSettings.Numeric) != 0)
{
return(BailOutResult);
}
rightPair = table[c];
}
else if (PUNCT_START <= c && c < PUNCT_LIMIT)
{
rightPair = table[c - PUNCT_START + LatinLimit];
}
else
{
rightPair = Lookup(table, c);
}
if (rightPair >= MIN_SHORT)
{
rightPair &= SHORT_PRIMARY_MASK;
break;
}
else if (rightPair > variableTop)
{
rightPair &= LONG_PRIMARY_MASK;
break;
}
else
{
long pairAndInc = NextPair(table, c, rightPair, right, rightIndex);
if (pairAndInc < 0)
{
++rightIndex;
pairAndInc = ~pairAndInc;
}
rightPair = (int)pairAndInc;
if (rightPair == BAIL_OUT)
{
return(BailOutResult);
}
rightPair = GetPrimaries(variableTop, rightPair);
}
}
if (leftPair == rightPair)
{
if (leftPair == EOS)
{
break;
}
leftPair = rightPair = 0;
continue;
}
int leftPrimary = leftPair & 0xffff;
int rightPrimary = rightPair & 0xffff;
if (leftPrimary != rightPrimary)
{
// Return the primary difference.
return((leftPrimary < rightPrimary) ? Collation.Less : Collation.Greater);
}
if (leftPair == EOS)
{
break;
}
//leftPair >>>= 16;
//rightPair >>>= 16;
leftPair = leftPair.TripleShift(16);
rightPair = rightPair.TripleShift(16);
}
// In the following, we need to re-fetch each character because we did not buffer the CEs,
// but we know that the string is well-formed and
// only contains supported characters and mappings.
// We might skip the secondary level but continue with the case level
// which is turned on separately.
if (CollationSettings.GetStrength(options) >= CollationStrength.Secondary)
{
leftIndex = rightIndex = startIndex;
leftPair = rightPair = 0;
for (; ;)
{
while (leftPair == 0)
{
if (leftIndex == left.Length)
{
leftPair = EOS;
break;
}
int c = left[leftIndex++];
if (c <= LatinMax)
{
leftPair = table[c];
}
else if (PUNCT_START <= c && c < PUNCT_LIMIT)
{
leftPair = table[c - PUNCT_START + LatinLimit];
}
else
{
leftPair = Lookup(table, c);
}
if (leftPair >= MIN_SHORT)
{
leftPair = GetSecondariesFromOneShortCE(leftPair);
break;
}
else if (leftPair > variableTop)
{
leftPair = COMMON_SEC_PLUS_OFFSET;
break;
}
else
{
long pairAndInc = NextPair(table, c, leftPair, left, leftIndex);
if (pairAndInc < 0)
{
++leftIndex;
pairAndInc = ~pairAndInc;
}
leftPair = GetSecondaries(variableTop, (int)pairAndInc);
}
}
while (rightPair == 0)
{
if (rightIndex == right.Length)
{
rightPair = EOS;
break;
}
int c = right[rightIndex++];
if (c <= LatinMax)
{
rightPair = table[c];
}
else if (PUNCT_START <= c && c < PUNCT_LIMIT)
{
rightPair = table[c - PUNCT_START + LatinLimit];
}
else
{
rightPair = Lookup(table, c);
}
if (rightPair >= MIN_SHORT)
{
rightPair = GetSecondariesFromOneShortCE(rightPair);
break;
}
else if (rightPair > variableTop)
{
rightPair = COMMON_SEC_PLUS_OFFSET;
break;
}
else
{
long pairAndInc = NextPair(table, c, rightPair, right, rightIndex);
if (pairAndInc < 0)
{
++rightIndex;
pairAndInc = ~pairAndInc;
}
rightPair = GetSecondaries(variableTop, (int)pairAndInc);
}
}
if (leftPair == rightPair)
{
if (leftPair == EOS)
{
break;
}
leftPair = rightPair = 0;
continue;
}
int leftSecondary = leftPair & 0xffff;
int rightSecondary = rightPair & 0xffff;
if (leftSecondary != rightSecondary)
{
if ((options & CollationSettings.BackwardSecondary) != 0)
{
// Full support for backwards secondary requires backwards contraction matching
// and moving backwards between merge separators.
return(BailOutResult);
}
return((leftSecondary < rightSecondary) ? Collation.Less : Collation.Greater);
}
if (leftPair == EOS)
{
break;
}
//leftPair >>>= 16;
//rightPair >>>= 16;
leftPair = leftPair.TripleShift(16);
rightPair = rightPair.TripleShift(16);
}
}
if ((options & CollationSettings.CaseLevel) != 0)
{
bool strengthIsPrimary = CollationSettings.GetStrength(options) == CollationStrength.Primary;
leftIndex = rightIndex = startIndex;
leftPair = rightPair = 0;
for (; ;)
{
while (leftPair == 0)
{
if (leftIndex == left.Length)
{
leftPair = EOS;
break;
}
int c = left[leftIndex++];
leftPair = (c <= LatinMax) ? table[c] : Lookup(table, c);
if (leftPair < MIN_LONG)
{
long pairAndInc = NextPair(table, c, leftPair, left, leftIndex);
if (pairAndInc < 0)
{
++leftIndex;
pairAndInc = ~pairAndInc;
}
leftPair = (int)pairAndInc;
}
leftPair = GetCases(variableTop, strengthIsPrimary, leftPair);
}
while (rightPair == 0)
{
if (rightIndex == right.Length)
{
rightPair = EOS;
break;
}
int c = right[rightIndex++];
rightPair = (c <= LatinMax) ? table[c] : Lookup(table, c);
if (rightPair < MIN_LONG)
{
long pairAndInc = NextPair(table, c, rightPair, right, rightIndex);
if (pairAndInc < 0)
{
++rightIndex;
pairAndInc = ~pairAndInc;
}
rightPair = (int)pairAndInc;
}
rightPair = GetCases(variableTop, strengthIsPrimary, rightPair);
}
if (leftPair == rightPair)
{
if (leftPair == EOS)
{
break;
}
leftPair = rightPair = 0;
continue;
}
int leftCase = leftPair & 0xffff;
int rightCase = rightPair & 0xffff;
if (leftCase != rightCase)
{
if ((options & CollationSettings.UpperFirst) == 0)
{
return((leftCase < rightCase) ? Collation.Less : Collation.Greater);
}
else
{
return((leftCase < rightCase) ? Collation.Greater : Collation.Less);
}
}
if (leftPair == EOS)
{
break;
}
//leftPair >>>= 16;
//rightPair >>>= 16;
leftPair = leftPair.TripleShift(16);
rightPair = rightPair.TripleShift(16);
}
}
if (CollationSettings.GetStrength(options) <= CollationStrength.Secondary)
{
return(Collation.Equal);
}
// Remove the case bits from the tertiary weight when caseLevel is on or caseFirst is off.
bool withCaseBits = CollationSettings.IsTertiaryWithCaseBits(options);
leftIndex = rightIndex = startIndex;
leftPair = rightPair = 0;
for (; ;)
{
while (leftPair == 0)
{
if (leftIndex == left.Length)
{
leftPair = EOS;
break;
}
int c = left[leftIndex++];
leftPair = (c <= LatinMax) ? table[c] : Lookup(table, c);
if (leftPair < MIN_LONG)
{
long pairAndInc = NextPair(table, c, leftPair, left, leftIndex);
if (pairAndInc < 0)
{
++leftIndex;
pairAndInc = ~pairAndInc;
}
leftPair = (int)pairAndInc;
}
leftPair = GetTertiaries(variableTop, withCaseBits, leftPair);
}
while (rightPair == 0)
{
if (rightIndex == right.Length)
{
rightPair = EOS;
break;
}
int c = right[rightIndex++];
rightPair = (c <= LatinMax) ? table[c] : Lookup(table, c);
if (rightPair < MIN_LONG)
{
long pairAndInc = NextPair(table, c, rightPair, right, rightIndex);
if (pairAndInc < 0)
{
++rightIndex;
pairAndInc = ~pairAndInc;
}
rightPair = (int)pairAndInc;
}
rightPair = GetTertiaries(variableTop, withCaseBits, rightPair);
}
if (leftPair == rightPair)
{
if (leftPair == EOS)
{
break;
}
leftPair = rightPair = 0;
continue;
}
int leftTertiary = leftPair & 0xffff;
int rightTertiary = rightPair & 0xffff;
if (leftTertiary != rightTertiary)
{
if (CollationSettings.SortsTertiaryUpperCaseFirst(options))
{
// Pass through EOS and MERGE_WEIGHT
// and keep real tertiary weights larger than the MERGE_WEIGHT.
// Tertiary CEs (secondary ignorables) are not supported in fast Latin.
if (leftTertiary > MERGE_WEIGHT)
{
leftTertiary ^= CASE_MASK;
}
if (rightTertiary > MERGE_WEIGHT)
{
rightTertiary ^= CASE_MASK;
}
}
return((leftTertiary < rightTertiary) ? Collation.Less : Collation.Greater);
}
if (leftPair == EOS)
{
break;
}
//leftPair >>>= 16;
//rightPair >>>= 16;
leftPair = leftPair.TripleShift(16);
rightPair = rightPair.TripleShift(16);
}
if (CollationSettings.GetStrength(options) <= CollationStrength.Tertiary)
{
return(Collation.Equal);
}
leftIndex = rightIndex = startIndex;
leftPair = rightPair = 0;
for (; ;)
{
while (leftPair == 0)
{
if (leftIndex == left.Length)
{
leftPair = EOS;
break;
}
int c = left[leftIndex++];
leftPair = (c <= LatinMax) ? table[c] : Lookup(table, c);
if (leftPair < MIN_LONG)
{
long pairAndInc = NextPair(table, c, leftPair, left, leftIndex);
if (pairAndInc < 0)
{
++leftIndex;
pairAndInc = ~pairAndInc;
}
leftPair = (int)pairAndInc;
}
leftPair = GetQuaternaries(variableTop, leftPair);
}
while (rightPair == 0)
{
if (rightIndex == right.Length)
{
rightPair = EOS;
break;
}
int c = right[rightIndex++];
rightPair = (c <= LatinMax) ? table[c] : Lookup(table, c);
if (rightPair < MIN_LONG)
{
long pairAndInc = NextPair(table, c, rightPair, right, rightIndex);
if (pairAndInc < 0)
{
++rightIndex;
pairAndInc = ~pairAndInc;
}
rightPair = (int)pairAndInc;
}
rightPair = GetQuaternaries(variableTop, rightPair);
}
if (leftPair == rightPair)
{
if (leftPair == EOS)
{
break;
}
leftPair = rightPair = 0;
continue;
}
int leftQuaternary = leftPair & 0xffff;
int rightQuaternary = rightPair & 0xffff;
if (leftQuaternary != rightQuaternary)
{
return((leftQuaternary < rightQuaternary) ? Collation.Less : Collation.Greater);
}
if (leftPair == EOS)
{
break;
}
//leftPair >>>= 16;
//rightPair >>>= 16;
leftPair = leftPair.TripleShift(16);
rightPair = rightPair.TripleShift(16);
}
return(Collation.Equal);
}
/// <summary>
/// Computes the options value for the compare functions
/// and writes the precomputed primary weights.
/// Returns -1 if the Latin fastpath is not supported for the data and settings.
/// The capacity must be <see cref="LatinLimit"/>.
/// </summary>
public static int GetOptions(CollationData data, CollationSettings settings,
char[] primaries)
{
char[] header = data.fastLatinTableHeader;
if (header == null)
{
return(-1);
}
Debug.Assert((header[0] >> 8) == Version);
if (primaries.Length != LatinLimit)
{
Debug.Assert(false);
return(-1);
}
int miniVarTop;
if ((settings.Options & CollationSettings.AlternateMask) == 0)
{
// No mini primaries are variable, set a variableTop just below the
// lowest long mini primary.
miniVarTop = MIN_LONG - 1;
}
else
{
int headerLength = header[0] & 0xff;
int i = 1 + settings.MaxVariable;
if (i >= headerLength)
{
return(-1); // variableTop >= digits, should not occur
}
miniVarTop = header[i];
}
bool digitsAreReordered = false;
if (settings.HasReordering)
{
long prevStart = 0;
long beforeDigitStart = 0;
long digitStart = 0;
long afterDigitStart = 0;
for (int group = ReorderCodes.First;
group < ReorderCodes.First + CollationData.MAX_NUM_SPECIAL_REORDER_CODES;
++group)
{
long start = data.GetFirstPrimaryForGroup(group);
start = settings.Reorder(start);
if (group == ReorderCodes.Digit)
{
beforeDigitStart = prevStart;
digitStart = start;
}
else if (start != 0)
{
if (start < prevStart)
{
// The permutation affects the groups up to Latin.
return(-1);
}
// In the future, there might be a special group between digits & Latin.
if (digitStart != 0 && afterDigitStart == 0 && prevStart == beforeDigitStart)
{
afterDigitStart = start;
}
prevStart = start;
}
}
long latinStart = data.GetFirstPrimaryForGroup(UScript.Latin);
latinStart = settings.Reorder(latinStart);
if (latinStart < prevStart)
{
return(-1);
}
if (afterDigitStart == 0)
{
afterDigitStart = latinStart;
}
if (!(beforeDigitStart < digitStart && digitStart < afterDigitStart))
{
digitsAreReordered = true;
}
}
char[] table = data.FastLatinTable; // skip the header
for (int c = 0; c < LatinLimit; ++c)
{
int p = table[c];
if (p >= MIN_SHORT)
{
p &= SHORT_PRIMARY_MASK;
}
else if (p > miniVarTop)
{
p &= LONG_PRIMARY_MASK;
}
else
{
p = 0;
}
primaries[c] = (char)p;
}
if (digitsAreReordered || (settings.Options & CollationSettings.Numeric) != 0)
{
// Bail out for digits.
for (int c = 0x30; c <= 0x39; ++c)
{
primaries[c] = (char)0;
}
}
// Shift the miniVarTop above other options.
return((miniVarTop << 16) | settings.Options);
}
public static int CompareUpToQuaternary(CollationIterator left, CollationIterator right,
CollationSettings settings)
{
int options = settings.Options;
long variableTop;
if ((options & CollationSettings.ALTERNATE_MASK) == 0)
{
variableTop = 0;
}
else
{
// +1 so that we can use "<" and primary ignorables test out early.
variableTop = settings.VariableTop + 1;
}
bool anyVariable = false;
// Fetch CEs, compare primaries, store secondary & tertiary weights.
for (; ;)
{
// We fetch CEs until we get a non-ignorable primary or reach the end.
long leftPrimary;
do
{
long ce = left.NextCE();
leftPrimary = ce.TripleShift(32);
if (leftPrimary < variableTop && leftPrimary > Collation.MERGE_SEPARATOR_PRIMARY)
{
// Variable CE, shift it to quaternary level.
// Ignore all following primary ignorables, and shift further variable CEs.
anyVariable = true;
do
{
// Store only the primary of the variable CE.
left.SetCurrentCE(ce & unchecked ((long)0xffffffff00000000L));
for (; ;)
{
ce = left.NextCE();
leftPrimary = ce.TripleShift(32);
if (leftPrimary == 0)
{
left.SetCurrentCE(0);
}
else
{
break;
}
}
} while (leftPrimary < variableTop && leftPrimary > Collation.MERGE_SEPARATOR_PRIMARY);
}
} while (leftPrimary == 0);
long rightPrimary;
do
{
long ce = right.NextCE();
rightPrimary = ce.TripleShift(32);
if (rightPrimary < variableTop && rightPrimary > Collation.MERGE_SEPARATOR_PRIMARY)
{
// Variable CE, shift it to quaternary level.
// Ignore all following primary ignorables, and shift further variable CEs.
anyVariable = true;
do
{
// Store only the primary of the variable CE.
right.SetCurrentCE(ce & unchecked ((long)0xffffffff00000000L));
for (; ;)
{
ce = right.NextCE();
rightPrimary = ce.TripleShift(32);
if (rightPrimary == 0)
{
right.SetCurrentCE(0);
}
else
{
break;
}
}
} while (rightPrimary < variableTop && rightPrimary > Collation.MERGE_SEPARATOR_PRIMARY);
}
} while (rightPrimary == 0);
if (leftPrimary != rightPrimary)
{
// Return the primary difference, with script reordering.
if (settings.HasReordering)
{
leftPrimary = settings.Reorder(leftPrimary);
rightPrimary = settings.Reorder(rightPrimary);
}
return((leftPrimary < rightPrimary) ? Collation.LESS : Collation.GREATER);
}
if (leftPrimary == Collation.NO_CE_PRIMARY)
{
break;
}
}
// Compare the buffered secondary & tertiary weights.
// We might skip the secondary level but continue with the case level
// which is turned on separately.
if (CollationSettings.GetStrength(options) >= CollationStrength.Secondary)
{
if ((options & CollationSettings.BACKWARD_SECONDARY) == 0)
{
int leftIndex2 = 0;
int rightIndex2 = 0;
for (; ;)
{
int leftSecondary;
do
{
leftSecondary = ((int)left.GetCE(leftIndex2++)).TripleShift(16);
} while (leftSecondary == 0);
int rightSecondary;
do
{
rightSecondary = ((int)right.GetCE(rightIndex2++)).TripleShift(16);
} while (rightSecondary == 0);
if (leftSecondary != rightSecondary)
{
return((leftSecondary < rightSecondary) ? Collation.LESS : Collation.GREATER);
}
if (leftSecondary == Collation.NO_CE_WEIGHT16)
{
break;
}
}
}
else
{
// The backwards secondary level compares secondary weights backwards
// within segments separated by the merge separator (U+FFFE, weight 02).
int leftStart = 0;
int rightStart = 0;
for (; ;)
{
// Find the merge separator or the NO_CE terminator.
long p;
int leftLimit = leftStart;
while ((p = left.GetCE(leftLimit).TripleShift(32)) > Collation.MERGE_SEPARATOR_PRIMARY ||
p == 0)
{
++leftLimit;
}
int rightLimit = rightStart;
while ((p = right.GetCE(rightLimit).TripleShift(32)) > Collation.MERGE_SEPARATOR_PRIMARY ||
p == 0)
{
++rightLimit;
}
// Compare the segments.
int leftIndex3 = leftLimit;
int rightIndex3 = rightLimit;
for (; ;)
{
int leftSecondary = 0;
while (leftSecondary == 0 && leftIndex3 > leftStart)
{
leftSecondary = ((int)left.GetCE(--leftIndex3)).TripleShift(16);
}
int rightSecondary = 0;
while (rightSecondary == 0 && rightIndex3 > rightStart)
{
rightSecondary = ((int)right.GetCE(--rightIndex3)).TripleShift(16);
}
if (leftSecondary != rightSecondary)
{
return((leftSecondary < rightSecondary) ? Collation.LESS : Collation.GREATER);
}
if (leftSecondary == 0)
{
break;
}
}
// Did we reach the end of either string?
// Both strings have the same number of merge separators,
// or else there would have been a primary-level difference.
Debug.Assert(left.GetCE(leftLimit) == right.GetCE(rightLimit));
if (p == Collation.NO_CE_PRIMARY)
{
break;
}
// Skip both merge separators and continue.
leftStart = leftLimit + 1;
rightStart = rightLimit + 1;
}
}
}
if ((options & CollationSettings.CASE_LEVEL) != 0)
{
CollationStrength strength = CollationSettings.GetStrength(options);
int leftIndex4 = 0;
int rightIndex4 = 0;
for (; ;)
{
int leftCase, leftLower32, rightCase;
if (strength == CollationStrength.Primary)
{
// Primary+caseLevel: Ignore case level weights of primary ignorables.
// Otherwise we would get a-umlaut > a
// which is not desirable for accent-insensitive sorting.
// Check for (lower 32 bits) == 0 as well because variable CEs are stored
// with only primary weights.
long ce;
do
{
ce = left.GetCE(leftIndex4++);
leftCase = (int)ce;
} while ((ce.TripleShift(32)) == 0 || leftCase == 0);
leftLower32 = leftCase;
leftCase &= 0xc000;
do
{
ce = right.GetCE(rightIndex4++);
rightCase = (int)ce;
} while ((ce.TripleShift(32)) == 0 || rightCase == 0);
rightCase &= 0xc000;
}
else
{
// Secondary+caseLevel: By analogy with the above,
// ignore case level weights of secondary ignorables.
//
// Note: A tertiary CE has uppercase case bits (0.0.ut)
// to keep tertiary+caseFirst well-formed.
//
// Tertiary+caseLevel: Also ignore case level weights of secondary ignorables.
// Otherwise a tertiary CE's uppercase would be no greater than
// a primary/secondary CE's uppercase.
// (See UCA well-formedness condition 2.)
// We could construct a special case weight higher than uppercase,
// but it's simpler to always ignore case weights of secondary ignorables,
// turning 0.0.ut into 0.0.0.t.
// (See LDML Collation, Case Parameters.)
do
{
leftCase = (int)left.GetCE(leftIndex4++);
} while ((leftCase & 0xffff0000) == 0);
leftLower32 = leftCase;
leftCase &= 0xc000;
do
{
rightCase = (int)right.GetCE(rightIndex4++);
} while ((rightCase & 0xffff0000) == 0);
rightCase &= 0xc000;
}
// No need to handle NO_CE and MERGE_SEPARATOR specially:
// There is one case weight for each previous-level weight,
// so level length differences were handled there.
if (leftCase != rightCase)
{
if ((options & CollationSettings.UPPER_FIRST) == 0)
{
return((leftCase < rightCase) ? Collation.LESS : Collation.GREATER);
}
else
{
return((leftCase < rightCase) ? Collation.GREATER : Collation.LESS);
}
}
if ((leftLower32.TripleShift(16)) == Collation.NO_CE_WEIGHT16)
{
break;
}
}
}
if (CollationSettings.GetStrength(options) <= CollationStrength.Secondary)
{
return(Collation.EQUAL);
}
int tertiaryMask = CollationSettings.GetTertiaryMask(options);
int leftIndex = 0;
int rightIndex = 0;
int anyQuaternaries = 0;
for (; ;)
{
int leftLower32, leftTertiary;
do
{
leftLower32 = (int)left.GetCE(leftIndex++);
anyQuaternaries |= leftLower32;
Debug.Assert((leftLower32 & Collation.ONLY_TERTIARY_MASK) != 0 || (leftLower32 & 0xc0c0) == 0);
leftTertiary = leftLower32 & tertiaryMask;
} while (leftTertiary == 0);
int rightLower32, rightTertiary;
do
{
rightLower32 = (int)right.GetCE(rightIndex++);
anyQuaternaries |= rightLower32;
Debug.Assert((rightLower32 & Collation.ONLY_TERTIARY_MASK) != 0 || (rightLower32 & 0xc0c0) == 0);
rightTertiary = rightLower32 & tertiaryMask;
} while (rightTertiary == 0);
if (leftTertiary != rightTertiary)
{
if (CollationSettings.SortsTertiaryUpperCaseFirst(options))
{
// Pass through NO_CE and keep real tertiary weights larger than that.
// Do not change the artificial uppercase weight of a tertiary CE (0.0.ut),
// to keep tertiary CEs well-formed.
// Their case+tertiary weights must be greater than those of
// primary and secondary CEs.
if (leftTertiary > Collation.NO_CE_WEIGHT16)
{
if ((leftLower32 & 0xffff0000) != 0)
{
leftTertiary ^= 0xc000;
}
else
{
leftTertiary += 0x4000;
}
}
if (rightTertiary > Collation.NO_CE_WEIGHT16)
{
if ((rightLower32 & 0xffff0000) != 0)
{
rightTertiary ^= 0xc000;
}
else
{
rightTertiary += 0x4000;
}
}
}
return((leftTertiary < rightTertiary) ? Collation.LESS : Collation.GREATER);
}
if (leftTertiary == Collation.NO_CE_WEIGHT16)
{
break;
}
}
if (CollationSettings.GetStrength(options) <= CollationStrength.Tertiary)
{
return(Collation.EQUAL);
}
if (!anyVariable && (anyQuaternaries & 0xc0) == 0)
{
// If there are no "variable" CEs and no non-zero quaternary weights,
// then there are no quaternary differences.
return(Collation.EQUAL);
}
leftIndex = 0;
rightIndex = 0;
for (; ;)
{
long leftQuaternary;
do
{
long ce = left.GetCE(leftIndex++);
leftQuaternary = ce & 0xffff;
if (leftQuaternary <= Collation.NO_CE_WEIGHT16)
{
// Variable primary or completely ignorable or NO_CE.
leftQuaternary = ce.TripleShift(32);
}
else
{
// Regular CE, not tertiary ignorable.
// Preserve the quaternary weight in bits 7..6.
leftQuaternary |= 0xffffff3fL;
}
} while (leftQuaternary == 0);
long rightQuaternary;
do
{
long ce = right.GetCE(rightIndex++);
rightQuaternary = ce & 0xffff;
if (rightQuaternary <= Collation.NO_CE_WEIGHT16)
{
// Variable primary or completely ignorable or NO_CE.
rightQuaternary = ce.TripleShift(32);
}
else
{
// Regular CE, not tertiary ignorable.
// Preserve the quaternary weight in bits 7..6.
rightQuaternary |= 0xffffff3fL;
}
} while (rightQuaternary == 0);
if (leftQuaternary != rightQuaternary)
{
// Return the difference, with script reordering.
if (settings.HasReordering)
{
leftQuaternary = settings.Reorder(leftQuaternary);
rightQuaternary = settings.Reorder(rightQuaternary);
}
return((leftQuaternary < rightQuaternary) ? Collation.LESS : Collation.GREATER);
}
if (leftQuaternary == Collation.NO_CE_PRIMARY)
{
break;
}
}
return(Collation.EQUAL);
}
internal void Parse(string ruleString, CollationSettings outSettings)
{
settings = outSettings;
Parse(ruleString);
}
internal static void Read(CollationTailoring @base, ByteBuffer inBytes,
CollationTailoring tailoring)
{
tailoring.Version = ICUBinary.ReadHeader(inBytes, DATA_FORMAT, IS_ACCEPTABLE);
if (@base != null && @base.GetUCAVersion() != tailoring.GetUCAVersion())
{
throw new ICUException("Tailoring UCA version differs from base data UCA version");
}
int inLength = inBytes.Remaining;
if (inLength < 8)
{
throw new ICUException("not enough bytes");
}
int indexesLength = inBytes.GetInt32(); // inIndexes[IX_INDEXES_LENGTH]
if (indexesLength < 2 || inLength < indexesLength * 4)
{
throw new ICUException("not enough indexes");
}
int[] inIndexes = new int[IX_TOTAL_SIZE + 1];
inIndexes[0] = indexesLength;
for (int i = 1; i < indexesLength && i < inIndexes.Length; ++i)
{
inIndexes[i] = inBytes.GetInt32();
}
for (int i = indexesLength; i < inIndexes.Length; ++i)
{
inIndexes[i] = -1;
}
if (indexesLength > inIndexes.Length)
{
ICUBinary.SkipBytes(inBytes, (indexesLength - inIndexes.Length) * 4);
}
// Assume that the tailoring data is in initial state,
// with null pointers and 0 lengths.
// Set pointers to non-empty data parts.
// Do this in order of their byte offsets. (Should help porting to Java.)
int index; // one of the indexes[] slots
int offset; // byte offset for the index part
int length; // number of bytes in the index part
if (indexesLength > IX_TOTAL_SIZE)
{
length = inIndexes[IX_TOTAL_SIZE];
}
else if (indexesLength > IX_REORDER_CODES_OFFSET)
{
length = inIndexes[indexesLength - 1];
}
else
{
length = 0; // only indexes, and inLength was already checked for them
}
if (inLength < length)
{
throw new ICUException("not enough bytes");
}
CollationData baseData = @base == null ? null : @base.Data;
int[] reorderCodes;
int reorderCodesLength;
index = IX_REORDER_CODES_OFFSET;
offset = inIndexes[index];
length = inIndexes[index + 1] - offset;
if (length >= 4)
{
if (baseData == null)
{
// We assume for collation settings that
// the base data does not have a reordering.
throw new ICUException("Collation base data must not reorder scripts");
}
reorderCodesLength = length / 4;
reorderCodes = ICUBinary.GetInts(inBytes, reorderCodesLength, length & 3);
// The reorderRanges (if any) are the trailing reorderCodes entries.
// Split the array at the boundary.
// Script or reorder codes do not exceed 16-bit values.
// Range limits are stored in the upper 16 bits, and are never 0.
int reorderRangesLength = 0;
while (reorderRangesLength < reorderCodesLength &&
(reorderCodes[reorderCodesLength - reorderRangesLength - 1] & 0xffff0000) != 0)
{
++reorderRangesLength;
}
Debug.Assert(reorderRangesLength < reorderCodesLength);
reorderCodesLength -= reorderRangesLength;
}
else
{
reorderCodes = new int[0];
reorderCodesLength = 0;
ICUBinary.SkipBytes(inBytes, length);
}
// There should be a reorder table only if there are reorder codes.
// However, when there are reorder codes the reorder table may be omitted to reduce
// the data size.
byte[] reorderTable = null;
index = IX_REORDER_TABLE_OFFSET;
offset = inIndexes[index];
length = inIndexes[index + 1] - offset;
if (length >= 256)
{
if (reorderCodesLength == 0)
{
throw new ICUException("Reordering table without reordering codes");
}
reorderTable = new byte[256];
inBytes.Get(reorderTable);
length -= 256;
}
else
{
// If we have reorder codes, then build the reorderTable at the end,
// when the CollationData is otherwise complete.
}
ICUBinary.SkipBytes(inBytes, length);
if (baseData != null && baseData.numericPrimary != (inIndexes[IX_OPTIONS] & 0xff000000L))
{
throw new ICUException("Tailoring numeric primary weight differs from base data");
}
CollationData data = null; // Remains null if there are no mappings.
index = IX_TRIE_OFFSET;
offset = inIndexes[index];
length = inIndexes[index + 1] - offset;
if (length >= 8)
{
tailoring.EnsureOwnedData();
data = tailoring.OwnedData;
data.Base = baseData;
data.numericPrimary = inIndexes[IX_OPTIONS] & 0xff000000L;
data.trie = tailoring.Trie = Trie2_32.CreateFromSerialized(inBytes);
int trieLength = data.trie.GetSerializedLength();
if (trieLength > length)
{
throw new ICUException("Not enough bytes for the mappings trie"); // No mappings.
}
length -= trieLength;
}
else if (baseData != null)
{
// Use the base data. Only the settings are tailored.
tailoring.Data = baseData;
}
else
{
throw new ICUException("Missing collation data mappings"); // No mappings.
}
ICUBinary.SkipBytes(inBytes, length);
index = IX_RESERVED8_OFFSET;
offset = inIndexes[index];
length = inIndexes[index + 1] - offset;
ICUBinary.SkipBytes(inBytes, length);
index = IX_CES_OFFSET;
offset = inIndexes[index];
length = inIndexes[index + 1] - offset;
if (length >= 8)
{
if (data == null)
{
throw new ICUException("Tailored ces without tailored trie");
}
data.ces = ICUBinary.GetLongs(inBytes, length / 8, length & 7);
}
else
{
ICUBinary.SkipBytes(inBytes, length);
}
index = IX_RESERVED10_OFFSET;
offset = inIndexes[index];
length = inIndexes[index + 1] - offset;
ICUBinary.SkipBytes(inBytes, length);
index = IX_CE32S_OFFSET;
offset = inIndexes[index];
length = inIndexes[index + 1] - offset;
if (length >= 4)
{
if (data == null)
{
throw new ICUException("Tailored ce32s without tailored trie");
}
data.ce32s = ICUBinary.GetInts(inBytes, length / 4, length & 3);
}
else
{
ICUBinary.SkipBytes(inBytes, length);
}
int jamoCE32sStart = inIndexes[IX_JAMO_CE32S_START];
if (jamoCE32sStart >= 0)
{
if (data == null || data.ce32s == null)
{
throw new ICUException("JamoCE32sStart index into non-existent ce32s[]");
}
data.jamoCE32s = new int[CollationData.JAMO_CE32S_LENGTH];
// ICU4N specific - added extension method to IList<T> to handle "copy to"
data.ce32s.CopyTo(jamoCE32sStart, data.jamoCE32s, 0, CollationData.JAMO_CE32S_LENGTH);
}
else if (data == null)
{
// Nothing to do.
}
else if (baseData != null)
{
data.jamoCE32s = baseData.jamoCE32s;
}
else
{
throw new ICUException("Missing Jamo CE32s for Hangul processing");
}
index = IX_ROOT_ELEMENTS_OFFSET;
offset = inIndexes[index];
length = inIndexes[index + 1] - offset;
if (length >= 4)
{
int rootElementsLength = length / 4;
if (data == null)
{
throw new ICUException("Root elements but no mappings");
}
if (rootElementsLength <= CollationRootElements.IX_SEC_TER_BOUNDARIES)
{
throw new ICUException("Root elements array too short");
}
data.rootElements = new long[rootElementsLength];
for (int i = 0; i < rootElementsLength; ++i)
{
data.rootElements[i] = inBytes.GetInt32() & 0xffffffffL; // unsigned int -> long
}
long commonSecTer = data.rootElements[CollationRootElements.IX_COMMON_SEC_AND_TER_CE];
if (commonSecTer != Collation.COMMON_SEC_AND_TER_CE)
{
throw new ICUException("Common sec/ter weights in base data differ from the hardcoded value");
}
long secTerBoundaries = data.rootElements[CollationRootElements.IX_SEC_TER_BOUNDARIES];
if ((secTerBoundaries.TripleShift(24)) < CollationKeys.SEC_COMMON_HIGH)
{
// [fixed last secondary common byte] is too low,
// and secondary weights would collide with compressed common secondaries.
throw new ICUException("[fixed last secondary common byte] is too low");
}
length &= 3;
}
ICUBinary.SkipBytes(inBytes, length);
index = IX_CONTEXTS_OFFSET;
offset = inIndexes[index];
length = inIndexes[index + 1] - offset;
if (length >= 2)
{
if (data == null)
{
throw new ICUException("Tailored contexts without tailored trie");
}
data.contexts = ICUBinary.GetString(inBytes, length / 2, length & 1);
}
else
{
ICUBinary.SkipBytes(inBytes, length);
}
index = IX_UNSAFE_BWD_OFFSET;
offset = inIndexes[index];
length = inIndexes[index + 1] - offset;
if (length >= 2)
{
if (data == null)
{
throw new ICUException("Unsafe-backward-set but no mappings");
}
if (baseData == null)
{
// Create the unsafe-backward set for the root collator.
// Include all non-zero combining marks and trail surrogates.
// We do this at load time, rather than at build time,
// to simplify Unicode version bootstrapping:
// The root data builder only needs the new FractionalUCA.txt data,
// but it need not be built with a version of ICU already updated to
// the corresponding new Unicode Character Database.
//
// The following is an optimized version of
// new UnicodeSet("[[:^lccc=0:][\\udc00-\\udfff]]").
// It is faster and requires fewer code dependencies.
tailoring.UnsafeBackwardSet = new UnicodeSet(0xdc00, 0xdfff); // trail surrogates
data.nfcImpl.AddLcccChars(tailoring.UnsafeBackwardSet);
}
else
{
// Clone the root collator's set contents.
tailoring.UnsafeBackwardSet = baseData.unsafeBackwardSet.CloneAsThawed();
}
// Add the ranges from the data file to the unsafe-backward set.
USerializedSet sset = new USerializedSet();
char[] unsafeData = ICUBinary.GetChars(inBytes, length / 2, length & 1);
length = 0;
sset.GetSet(unsafeData, 0);
int count = sset.CountRanges();
int[] range = new int[2];
for (int i = 0; i < count; ++i)
{
sset.GetRange(i, range);
tailoring.UnsafeBackwardSet.Add(range[0], range[1]);
}
// Mark each lead surrogate as "unsafe"
// if any of its 1024 associated supplementary code points is "unsafe".
int c = 0x10000;
for (int lead = 0xd800; lead < 0xdc00; ++lead, c += 0x400)
{
if (!tailoring.UnsafeBackwardSet.ContainsNone(c, c + 0x3ff))
{
tailoring.UnsafeBackwardSet.Add(lead);
}
}
tailoring.UnsafeBackwardSet.Freeze();
data.unsafeBackwardSet = tailoring.UnsafeBackwardSet;
}
else if (data == null)
{
// Nothing to do.
}
else if (baseData != null)
{
// No tailoring-specific data: Alias the root collator's set.
data.unsafeBackwardSet = baseData.unsafeBackwardSet;
}
else
{
throw new ICUException("Missing unsafe-backward-set");
}
ICUBinary.SkipBytes(inBytes, length);
// If the fast Latin format version is different,
// or the version is set to 0 for "no fast Latin table",
// then just always use the normal string comparison path.
index = IX_FAST_LATIN_TABLE_OFFSET;
offset = inIndexes[index];
length = inIndexes[index + 1] - offset;
if (data != null)
{
data.fastLatinTable = null;
data.fastLatinTableHeader = null;
if (((inIndexes[IX_OPTIONS] >> 16) & 0xff) == CollationFastLatin.VERSION)
{
if (length >= 2)
{
char header0 = inBytes.GetChar();
int headerLength = header0 & 0xff;
data.fastLatinTableHeader = new char[headerLength];
data.fastLatinTableHeader[0] = header0;
for (int i = 1; i < headerLength; ++i)
{
data.fastLatinTableHeader[i] = inBytes.GetChar();
}
int tableLength = length / 2 - headerLength;
data.fastLatinTable = ICUBinary.GetChars(inBytes, tableLength, length & 1);
length = 0;
if ((header0 >> 8) != CollationFastLatin.VERSION)
{
throw new ICUException("Fast-Latin table version differs from version in data header");
}
}
else if (baseData != null)
{
data.fastLatinTable = baseData.fastLatinTable;
data.fastLatinTableHeader = baseData.fastLatinTableHeader;
}
}
}
ICUBinary.SkipBytes(inBytes, length);
index = IX_SCRIPTS_OFFSET;
offset = inIndexes[index];
length = inIndexes[index + 1] - offset;
if (length >= 2)
{
if (data == null)
{
throw new ICUException("Script order data but no mappings");
}
int scriptsLength = length / 2;
CharBuffer inChars = inBytes.AsCharBuffer();
data.numScripts = inChars.Get();
// There must be enough entries for both arrays, including more than two range starts.
int scriptStartsLength = scriptsLength - (1 + data.numScripts + 16);
if (scriptStartsLength <= 2)
{
throw new ICUException("Script order data too short");
}
inChars.Get(data.scriptsIndex = new char[data.numScripts + 16]);
inChars.Get(data.scriptStarts = new char[scriptStartsLength]);
if (!(data.scriptStarts[0] == 0 &&
data.scriptStarts[1] == ((Collation.MERGE_SEPARATOR_BYTE + 1) << 8) &&
data.scriptStarts[scriptStartsLength - 1] ==
(Collation.TRAIL_WEIGHT_BYTE << 8)))
{
throw new ICUException("Script order data not valid");
}
}
else if (data == null)
{
// Nothing to do.
}
else if (baseData != null)
{
data.numScripts = baseData.numScripts;
data.scriptsIndex = baseData.scriptsIndex;
data.scriptStarts = baseData.scriptStarts;
}
ICUBinary.SkipBytes(inBytes, length);
index = IX_COMPRESSIBLE_BYTES_OFFSET;
offset = inIndexes[index];
length = inIndexes[index + 1] - offset;
if (length >= 256)
{
if (data == null)
{
throw new ICUException("Data for compressible primary lead bytes but no mappings");
}
data.compressibleBytes = new bool[256];
for (int i = 0; i < 256; ++i)
{
data.compressibleBytes[i] = inBytes.Get() != 0;
}
length -= 256;
}
else if (data == null)
{
// Nothing to do.
}
else if (baseData != null)
{
data.compressibleBytes = baseData.compressibleBytes;
}
else
{
throw new ICUException("Missing data for compressible primary lead bytes");
}
ICUBinary.SkipBytes(inBytes, length);
index = IX_RESERVED18_OFFSET;
offset = inIndexes[index];
length = inIndexes[index + 1] - offset;
ICUBinary.SkipBytes(inBytes, length);
CollationSettings ts = tailoring.Settings.ReadOnly;
int options = inIndexes[IX_OPTIONS] & 0xffff;
char[] fastLatinPrimaries = new char[CollationFastLatin.LATIN_LIMIT];
int fastLatinOptions = CollationFastLatin.GetOptions(
tailoring.Data, ts, fastLatinPrimaries);
if (options == ts.Options && ts.VariableTop != 0 &&
Arrays.Equals(reorderCodes, ts.ReorderCodes) &&
fastLatinOptions == ts.FastLatinOptions &&
(fastLatinOptions < 0 ||
Arrays.Equals(fastLatinPrimaries, ts.FastLatinPrimaries)))
{
return;
}
CollationSettings settings = tailoring.Settings.CopyOnWrite();
settings.Options = options;
// Set variableTop from options and scripts data.
settings.VariableTop = tailoring.Data.GetLastPrimaryForGroup(
ReorderCodes.First + settings.MaxVariable);
if (settings.VariableTop == 0)
{
throw new ICUException("The maxVariable could not be mapped to a variableTop");
}
if (reorderCodesLength != 0)
{
settings.AliasReordering(baseData, reorderCodes, reorderCodesLength, reorderTable);
}
settings.FastLatinOptions = CollationFastLatin.GetOptions(
tailoring.Data, settings,
settings.FastLatinPrimaries);
}