public DirectIntersectTermsEnum(DirectPostingsFormat.DirectField outerInstance,
CompiledAutomaton compiled, BytesRef startTerm)
{
this.outerInstance = outerInstance;
runAutomaton = compiled.RunAutomaton;
compiledAutomaton = compiled;
termOrd = -1;
states = new State[1];
states[0] = new State(this);
states[0].changeOrd = outerInstance.terms.Length;
states[0].state = runAutomaton.InitialState;
states[0].transitions = compiledAutomaton.SortedTransitions[states[0].state];
states[0].transitionUpto = -1;
states[0].transitionMax = -1;
//System.out.println("IE.init startTerm=" + startTerm);
if (startTerm != null)
{
int skipUpto = 0;
if (startTerm.Length == 0)
{
if (outerInstance.terms.Length > 0 && outerInstance.termOffsets[1] == 0)
{
termOrd = 0;
}
}
else
{
termOrd++;
for (int i = 0; i < startTerm.Length; i++)
{
int label = startTerm.Bytes[startTerm.Offset + i] & 0xFF;
while (label > states[i].transitionMax)
{
states[i].transitionUpto++;
Debug.Assert(states[i].transitionUpto < states[i].transitions.Length);
states[i].transitionMin = states[i].transitions[states[i].transitionUpto].Min;
states[i].transitionMax = states[i].transitions[states[i].transitionUpto].Max;
Debug.Assert(states[i].transitionMin >= 0);
Debug.Assert(states[i].transitionMin <= 255);
Debug.Assert(states[i].transitionMax >= 0);
Debug.Assert(states[i].transitionMax <= 255);
}
// Skip forwards until we find a term matching
// the label at this position:
while (termOrd < outerInstance.terms.Length)
{
int skipOffset = outerInstance.skipOffsets[termOrd];
int numSkips = outerInstance.skipOffsets[termOrd + 1] - skipOffset;
int termOffset_i = outerInstance.termOffsets[termOrd];
int termLength = outerInstance.termOffsets[1 + termOrd] - termOffset_i;
// if (DEBUG) {
// System.out.println(" check termOrd=" + termOrd + " term=" + new BytesRef(termBytes, termOffset, termLength).utf8ToString() + " skips=" + Arrays.toString(skips) + " i=" + i);
// }
if (termOrd == states[stateUpto].changeOrd)
{
// if (DEBUG) {
// System.out.println(" end push return");
// }
stateUpto--;
termOrd--;
return;
}
if (termLength == i)
{
termOrd++;
skipUpto = 0;
// if (DEBUG) {
// System.out.println(" term too short; next term");
// }
}
else if (label < (outerInstance.termBytes[termOffset_i + i] & 0xFF))
{
termOrd--;
// if (DEBUG) {
// System.out.println(" no match; already beyond; return termOrd=" + termOrd);
// }
stateUpto -= skipUpto;
Debug.Assert(stateUpto >= 0);
return;
}
else if (label == (outerInstance.termBytes[termOffset_i + i] & 0xFF))
{
// if (DEBUG) {
// System.out.println(" label[" + i + "] matches");
// }
if (skipUpto < numSkips)
{
Grow();
int nextState = runAutomaton.Step(states[stateUpto].state, label);
// Automaton is required to accept startTerm:
Debug.Assert(nextState != -1);
stateUpto++;
states[stateUpto].changeOrd = outerInstance.skips[skipOffset + skipUpto++];
states[stateUpto].state = nextState;
states[stateUpto].transitions =
compiledAutomaton.SortedTransitions[nextState];
states[stateUpto].transitionUpto = -1;
states[stateUpto].transitionMax = -1;
//System.out.println(" push " + states[stateUpto].transitions.length + " trans");
// if (DEBUG) {
// System.out.println(" push skip; changeOrd=" + states[stateUpto].changeOrd);
// }
// Match next label at this same term:
goto nextLabelContinue;
}
else
{
// if (DEBUG) {
// System.out.println(" linear scan");
// }
// Index exhausted: just scan now (the
// number of scans required will be less
// than the minSkipCount):
int startTermOrd = termOrd;
while (termOrd < outerInstance.terms.Length &&
outerInstance.Compare(termOrd, startTerm) <= 0)
{
Debug.Assert(termOrd == startTermOrd ||
outerInstance.skipOffsets[termOrd] ==
outerInstance.skipOffsets[termOrd + 1]);
termOrd++;
}
Debug.Assert(termOrd - startTermOrd < outerInstance.minSkipCount);
termOrd--;
stateUpto -= skipUpto;
// if (DEBUG) {
// System.out.println(" end termOrd=" + termOrd);
// }
return;
}
}
else
{
if (skipUpto < numSkips)
{
termOrd = outerInstance.skips[skipOffset + skipUpto];
// if (DEBUG) {
// System.out.println(" no match; skip to termOrd=" + termOrd);
// }
}
else
{
// if (DEBUG) {
// System.out.println(" no match; next term");
// }
termOrd++;
}
skipUpto = 0;
}
}
// startTerm is >= last term so enum will not
// return any terms:
termOrd--;
// if (DEBUG) {
// System.out.println(" beyond end; no terms will match");
// }
return;
nextLabelContinue:
;
}
nextLabelBreak:
;
}
int termOffset = outerInstance.termOffsets[termOrd];
int termLen = outerInstance.termOffsets[1 + termOrd] - termOffset;
if (termOrd >= 0 &&
!startTerm.Equals(new BytesRef(outerInstance.termBytes, termOffset, termLen)))
{
stateUpto -= skipUpto;
termOrd--;
}
// if (DEBUG) {
// System.out.println(" loop end; return termOrd=" + termOrd + " stateUpto=" + stateUpto);
// }
}
}
// TODO: in some cases we can filter by length? eg
// regexp foo*bar must be at least length 6 bytes
public IntersectEnum(BlockTreeTermsReader.FieldReader outerInstance, CompiledAutomaton compiled, BytesRef startTerm)
{
this.OuterInstance = outerInstance;
// if (DEBUG) {
// System.out.println("\nintEnum.init seg=" + segment + " commonSuffix=" + brToString(compiled.commonSuffixRef));
// }
runAutomaton = compiled.RunAutomaton;
CompiledAutomaton = compiled;
@in = (IndexInput)[email protected]();
Stack = new Frame[5];
for (int idx = 0; idx < Stack.Length; idx++)
{
Stack[idx] = new Frame(this, idx);
}
for (int arcIdx = 0; arcIdx < Arcs.Length; arcIdx++)
{
Arcs[arcIdx] = new FST<BytesRef>.Arc<BytesRef>();
}
if (outerInstance.Index == null)
{
FstReader = null;
}
else
{
FstReader = outerInstance.Index.BytesReader;
}
// TODO: if the automaton is "smallish" we really
// should use the terms index to seek at least to
// the initial term and likely to subsequent terms
// (or, maybe just fallback to ATE for such cases).
// Else the seek cost of loading the frames will be
// too costly.
FST<BytesRef>.Arc<BytesRef> arc = outerInstance.Index.GetFirstArc(Arcs[0]);
// Empty string prefix must have an output in the index!
Debug.Assert(arc.Final);
// Special pushFrame since it's the first one:
Frame f = Stack[0];
f.Fp = f.FpOrig = outerInstance.RootBlockFP;
f.Prefix = 0;
f.State = runAutomaton.InitialState;
f.Arc = arc;
f.OutputPrefix = arc.Output;
f.Load(outerInstance.RootCode);
// for assert:
Debug.Assert(SetSavedStartTerm(startTerm));
CurrentFrame = f;
if (startTerm != null)
{
SeekToStartTerm(startTerm);
}
}
