internal RangeDescriptor(RBBISetBuilder.RangeDescriptor other)
{
fStartChar = other.fStartChar;
fEndChar = other.fEndChar;
fNum = other.fNum;
fIncludesSets = new ArrayList(other.fIncludesSets);
}
// -------------------------------------------------------------------------------------
//
// RangeDesriptor::split()
//
// -------------------------------------------------------------------------------------
internal void Split(int where)
{
IBM.ICU.Impl.Assert.Assrt(where > fStartChar && where <= fEndChar);
RBBISetBuilder.RangeDescriptor nr = new RBBISetBuilder.RangeDescriptor(this);
// RangeDescriptor copy constructor copies all fields.
// Only need to update those that are different after the split.
nr.fStartChar = where;
this.fEndChar = where - 1;
nr.fNext = this.fNext;
this.fNext = nr;
// TODO: fIncludesSets is not updated. Check it out.
// Probably because they haven't been populated yet,
// but still sloppy.
}
// ------------------------------------------------------------------------
//
// build Build the list of non-overlapping character ranges
// from the Unicode Sets.
//
// ------------------------------------------------------------------------
internal void Build()
{
RBBINode usetNode;
RBBISetBuilder.RangeDescriptor rlRange;
if (fRB.fDebugEnv != null && fRB.fDebugEnv.IndexOf("usets") >= 0)
{
PrintSets();
}
// Initialize the process by creating a single range encompassing all
// characters
// that is in no sets.
//
fRangeList = new RBBISetBuilder.RangeDescriptor();
fRangeList.fStartChar = 0;
fRangeList.fEndChar = 0x10ffff;
//
// Find the set of non-overlapping ranges of characters
//
IIterator ni = new ILOG.J2CsMapping.Collections.IteratorAdapter(fRB.fUSetNodes.GetEnumerator());
while (ni.HasNext())
{
usetNode = (RBBINode)ni.Next();
UnicodeSet inputSet = usetNode.fInputSet;
int inputSetRangeCount = inputSet.GetRangeCount();
int inputSetRangeIndex = 0;
rlRange = fRangeList;
for (;;)
{
if (inputSetRangeIndex >= inputSetRangeCount)
{
break;
}
int inputSetRangeBegin = inputSet
.GetRangeStart(inputSetRangeIndex);
int inputSetRangeEnd = inputSet.GetRangeEnd(inputSetRangeIndex);
// skip over ranges from the range list that are completely
// below the current range from the input unicode set.
while (rlRange.fEndChar < inputSetRangeBegin)
{
rlRange = rlRange.fNext;
}
// If the start of the range from the range list is before with
// the start of the range from the unicode set, split the range
// list range
// in two, with one part being before (wholly outside of) the
// unicode set
// and the other containing the rest.
// Then continue the loop; the post-split current range will
// then be skipped
// over
if (rlRange.fStartChar < inputSetRangeBegin)
{
rlRange.Split(inputSetRangeBegin);
continue;
}
// Same thing at the end of the ranges...
// If the end of the range from the range list doesn't coincide
// with
// the end of the range from the unicode set, split the range
// list
// range in two. The first part of the split range will be
// wholly inside the Unicode set.
if (rlRange.fEndChar > inputSetRangeEnd)
{
rlRange.Split(inputSetRangeEnd + 1);
}
// The current rlRange is now entirely within the UnicodeSet
// range.
// Add this unicode set to the list of sets for this rlRange
if (rlRange.fIncludesSets.IndexOf(usetNode) == -1)
{
ILOG.J2CsMapping.Collections.Generics.Collections.Add(rlRange.fIncludesSets, usetNode);
}
// Advance over ranges that we are finished with.
if (inputSetRangeEnd == rlRange.fEndChar)
{
inputSetRangeIndex++;
}
rlRange = rlRange.fNext;
}
}
if (fRB.fDebugEnv != null && fRB.fDebugEnv.IndexOf("range") >= 0)
{
PrintRanges();
}
//
// Group the above ranges, with each group consisting of one or more
// ranges that are in exactly the same set of original UnicodeSets.
// The groups are numbered, and these group numbers are the set of
// input symbols recognized by the run-time state machine.
//
// Numbering: # 0 (state table column 0) is unused.
// # 1 is reserved - table column 1 is for end-of-input
// # 2 is reserved - table column 2 is for beginning-in-input
// # 3 is the first range list.
//
RBBISetBuilder.RangeDescriptor rlSearchRange;
for (rlRange = fRangeList; rlRange != null; rlRange = rlRange.fNext)
{
for (rlSearchRange = fRangeList; rlSearchRange != rlRange; rlSearchRange = rlSearchRange.fNext)
{
if (rlRange.fIncludesSets.Equals(rlSearchRange.fIncludesSets))
{
rlRange.fNum = rlSearchRange.fNum;
break;
}
}
if (rlRange.fNum == 0)
{
fGroupCount++;
rlRange.fNum = fGroupCount + 2;
rlRange.SetDictionaryFlag();
AddValToSets(rlRange.fIncludesSets, fGroupCount + 2);
}
}
// Handle input sets that contain the special string {eof}.
// Column 1 of the state table is reserved for EOF on input.
// Column 2 is reserved for before-the-start-input.
// (This column can be optimized away later if there are no rule
// references to {bof}.)
// Add this column value (1 or 2) to the equivalent expression
// subtree for each UnicodeSet that contains the string {eof}
// Because {bof} and {eof} are not a characters in the normal sense,
// they doesn't affect the computation of ranges or TRIE.
String eofString = "eof";
String bofString = "bof";
ni = new ILOG.J2CsMapping.Collections.IteratorAdapter(fRB.fUSetNodes.GetEnumerator());
while (ni.HasNext())
{
usetNode = (RBBINode)ni.Next();
UnicodeSet inputSet_0 = usetNode.fInputSet;
if (inputSet_0.Contains(eofString))
{
AddValToSet(usetNode, 1);
}
if (inputSet_0.Contains(bofString))
{
AddValToSet(usetNode, 2);
fSawBOF = true;
}
}
if (fRB.fDebugEnv != null && fRB.fDebugEnv.IndexOf("rgroup") >= 0)
{
PrintRangeGroups();
}
if (fRB.fDebugEnv != null && fRB.fDebugEnv.IndexOf("esets") >= 0)
{
PrintSets();
}
// IntTrieBuilder(int aliasdata[], int maxdatalength,
// int initialvalue, int leadunitvalue,
// boolean latin1linear)
fTrie = new IntTrieBuilder(null, // Data array (utrie will allocate one)
100000, // Max Data Length
0, // Initial value for all code points
0, // Lead Surrogate unit value,
true); // Keep Latin 1 in separately.
for (rlRange = fRangeList; rlRange != null; rlRange = rlRange.fNext)
{
fTrie.SetRange(rlRange.fStartChar, rlRange.fEndChar + 1,
rlRange.fNum, true);
}
}
