public MockHoleInjectingTokenFilter(Random random, TokenStream @in)
: base(@in)
{
RandomSeed = random.Next();
PosIncAtt = AddAttribute<IPositionIncrementAttribute>();
PosLenAtt = AddAttribute<IPositionLengthAttribute>();
}
public GraphTokenizer(TextReader input)
: base(input)
{
TermAtt = AddAttribute<ICharTermAttribute>();
OffsetAtt = AddAttribute<IOffsetAttribute>();
PosIncrAtt = AddAttribute<IPositionIncrementAttribute>();
PosLengthAtt = AddAttribute<IPositionLengthAttribute>();
}
public CannedBinaryTokenStream(params BinaryToken[] tokens)
: base()
{
this.Tokens = tokens;
TermAtt = AddAttribute<IBinaryTermAttribute>();
PosIncrAtt = AddAttribute<IPositionIncrementAttribute>();
PosLengthAtt = AddAttribute<IPositionLengthAttribute>();
OffsetAtt = AddAttribute<IOffsetAttribute>();
}
/// <summary>
/// Construct a token stream filtering the given input using a Set of common
/// words to create bigrams. Outputs both unigrams with position increment and
/// bigrams with position increment 0 type=gram where one or both of the words
/// in a potential bigram are in the set of common words .
/// </summary>
/// <param name="input"> TokenStream input in filter chain </param>
/// <param name="commonWords"> The set of common words. </param>
public CommonGramsFilter(LuceneVersion matchVersion, TokenStream input, CharArraySet commonWords)
: base(input)
{
termAttribute = AddAttribute<ICharTermAttribute>();
offsetAttribute = AddAttribute<IOffsetAttribute>();
typeAttribute = AddAttribute<ITypeAttribute>();
posIncAttribute = AddAttribute<IPositionIncrementAttribute>();
posLenAttribute = AddAttribute<IPositionLengthAttribute>();
this.commonWords = commonWords;
}
/// <summary>
/// If inputText is non-null, and the TokenStream has
/// offsets, we include the surface form in each arc's
/// label.
/// </summary>
public TokenStreamToDot(string inputText, TokenStream @in, TextWriter @out)
{
this.@in = @in;
this.@out = @out;
this.InputText = inputText;
TermAtt = @in.AddAttribute<ICharTermAttribute>();
PosIncAtt = @in.AddAttribute<IPositionIncrementAttribute>();
PosLengthAtt = @in.AddAttribute<IPositionLengthAttribute>();
if (@in.HasAttribute<IOffsetAttribute>())
{
OffsetAtt = @in.AddAttribute<IOffsetAttribute>();
}
else
{
OffsetAtt = null;
}
}
public virtual void TestOutputHangsOffEnd()
{
b = new SynonymMap.Builder(true);
const bool keepOrig = false;
// b hangs off the end (no input token under it):
Add("a", "a b", keepOrig);
tokensIn = new MockTokenizer(new StringReader("a"), MockTokenizer.WHITESPACE, true);
tokensIn.Reset();
assertTrue(tokensIn.IncrementToken());
assertFalse(tokensIn.IncrementToken());
tokensIn.End();
tokensIn.Dispose();
tokensOut = new SynonymFilter(tokensIn, b.Build(), true);
termAtt = tokensOut.AddAttribute <ICharTermAttribute>();
posIncrAtt = tokensOut.AddAttribute <IPositionIncrementAttribute>();
offsetAtt = tokensOut.AddAttribute <IOffsetAttribute>();
posLenAtt = tokensOut.AddAttribute <IPositionLengthAttribute>();
// Make sure endOffset inherits from previous input token:
Verify("a", "a b:1");
}
public EdgeNGramTokenFilter(LuceneVersion version, TokenStream input, Side side, int minGram, int maxGram)
: base(input)
{
// LUCENENET specific - version cannot be null because it is a value type.
if (version.OnOrAfter(LuceneVersion.LUCENE_44) && side == Side.BACK)
{
throw new ArgumentException("Side.BACK is not supported anymore as of Lucene 4.4, use ReverseStringFilter up-front and afterward");
}
if (!side.IsDefined())
{
throw new ArgumentOutOfRangeException(nameof(side), "sideLabel must be either front or back"); // LUCENENET specific - changed from IllegalArgumentException to ArgumentOutOfRangeException (.NET convention)
}
if (minGram < 1)
{
throw new ArgumentOutOfRangeException(nameof(minGram), "minGram must be greater than zero"); // LUCENENET specific - changed from IllegalArgumentException to ArgumentOutOfRangeException (.NET convention)
}
if (minGram > maxGram)
{
throw new ArgumentException("minGram must not be greater than maxGram");
}
this.version = version;
this.charUtils = version.OnOrAfter(LuceneVersion.LUCENE_44) ? CharacterUtils.GetInstance(version) : CharacterUtils.GetJava4Instance(version);
this.minGram = minGram;
this.maxGram = maxGram;
this.side = side;
this.termAtt = AddAttribute <ICharTermAttribute>();
this.offsetAtt = AddAttribute <IOffsetAttribute>();
this.posIncrAtt = AddAttribute <IPositionIncrementAttribute>();
this.posLenAtt = AddAttribute <IPositionLengthAttribute>();
}
/// <param name="input"> input tokenstream </param>
/// <param name="synonyms"> synonym map </param>
/// <param name="ignoreCase"> case-folds input for matching with <seealso cref="Character#toLowerCase(int)"/>.
/// Note, if you set this to true, its your responsibility to lowercase
/// the input entries when you create the <seealso cref="SynonymMap"/> </param>
public SynonymFilter(TokenStream input, SynonymMap synonyms, bool ignoreCase)
: base(input)
{
termAtt = AddAttribute<ICharTermAttribute>();
posIncrAtt = AddAttribute<IPositionIncrementAttribute>();
posLenAtt = AddAttribute<IPositionLengthAttribute>();
typeAtt = AddAttribute<ITypeAttribute>();
offsetAtt = AddAttribute<IOffsetAttribute>();
this.synonyms = synonyms;
this.ignoreCase = ignoreCase;
this.fst = synonyms.fst;
if (fst == null)
{
throw new System.ArgumentException("fst must be non-null");
}
this.fstReader = fst.BytesReader;
// Must be 1+ so that when roll buffer is at full
// lookahead we can distinguish this full buffer from
// the empty buffer:
rollBufferSize = 1 + synonyms.maxHorizontalContext;
futureInputs = new PendingInput[rollBufferSize];
futureOutputs = new PendingOutputs[rollBufferSize];
for (int pos = 0; pos < rollBufferSize; pos++)
{
futureInputs[pos] = new PendingInput();
futureOutputs[pos] = new PendingOutputs();
}
//System.out.println("FSTFilt maxH=" + synonyms.maxHorizontalContext);
scratchArc = new FST.Arc<BytesRef>();
}
public virtual void TestBasic2()
{
b = new SynonymMap.Builder(true);
const bool keepOrig = false;
Add("aaa", "aaaa1 aaaa2 aaaa3", keepOrig);
Add("bbb", "bbbb1 bbbb2", keepOrig);
tokensIn = new MockTokenizer(new StringReader("a"), MockTokenizer.WHITESPACE, true);
tokensIn.Reset();
assertTrue(tokensIn.IncrementToken());
assertFalse(tokensIn.IncrementToken());
tokensIn.End();
tokensIn.Dispose();
tokensOut = new SynonymFilter(tokensIn, b.Build(), true);
termAtt = tokensOut.AddAttribute<ICharTermAttribute>();
posIncrAtt = tokensOut.AddAttribute<IPositionIncrementAttribute>();
posLenAtt = tokensOut.AddAttribute<IPositionLengthAttribute>();
offsetAtt = tokensOut.AddAttribute<IOffsetAttribute>();
#pragma warning disable 162
if (keepOrig)
{
Verify("xyzzy bbb pot of gold", "xyzzy bbb/bbbb1 pot/bbbb2 of gold");
Verify("xyzzy aaa pot of gold", "xyzzy aaa/aaaa1 pot/aaaa2 of/aaaa3 gold");
}
else
{
Verify("xyzzy bbb pot of gold", "xyzzy bbbb1 pot/bbbb2 of gold");
Verify("xyzzy aaa pot of gold", "xyzzy aaaa1 pot/aaaa2 of/aaaa3 gold");
}
#pragma warning restore 612, 618
}
private void InitParams()
{
TermAtt = AddAttribute<ICharTermAttribute>();
PosIncrAtt = AddAttribute<IPositionIncrementAttribute>();
PosLengthAtt = AddAttribute<IPositionLengthAttribute>();
OffsetAtt = AddAttribute<IOffsetAttribute>();
PayloadAtt = AddAttribute<IPayloadAttribute>();
}
public virtual void TestBasic()
{
b = new SynonymMap.Builder(true);
Add("a", "foo", true);
Add("a b", "bar fee", true);
Add("b c", "dog collar", true);
Add("c d", "dog harness holder extras", true);
Add("m c e", "dog barks loudly", false);
Add("i j k", "feep", true);
Add("e f", "foo bar", false);
Add("e f", "baz bee", false);
Add("z", "boo", false);
Add("y", "bee", true);
tokensIn = new MockTokenizer(new StringReader("a"), MockTokenizer.WHITESPACE, true);
tokensIn.Reset();
assertTrue(tokensIn.IncrementToken());
assertFalse(tokensIn.IncrementToken());
tokensIn.End();
tokensIn.Dispose();
tokensOut = new SynonymFilter(tokensIn, b.Build(), true);
termAtt = tokensOut.AddAttribute<ICharTermAttribute>();
posIncrAtt = tokensOut.AddAttribute<IPositionIncrementAttribute>();
posLenAtt = tokensOut.AddAttribute<IPositionLengthAttribute>();
offsetAtt = tokensOut.AddAttribute<IOffsetAttribute>();
Verify("a b c", "a/bar b/fee c");
// syn output extends beyond input tokens
Verify("x a b c d", "x a/bar b/fee c/dog d/harness holder extras");
Verify("a b a", "a/bar b/fee a/foo");
// outputs that add to one another:
Verify("c d c d", "c/dog d/harness c/holder/dog d/extras/harness holder extras");
// two outputs for same input
Verify("e f", "foo/baz bar/bee");
// verify multi-word / single-output offsets:
Verify("g i j k g", "g i/feep:7_3 j k g");
// mixed keepOrig true/false:
Verify("a m c e x", "a/foo dog barks loudly x");
Verify("c d m c e x", "c/dog d/harness holder/dog extras/barks loudly x");
assertTrue(tokensOut.CaptureCount > 0);
// no captureStates when no syns matched
Verify("p q r s t", "p q r s t");
assertEquals(0, tokensOut.CaptureCount);
// no captureStates when only single-input syns, w/ no
// lookahead needed, matched
Verify("p q z y t", "p q boo y/bee t");
assertEquals(0, tokensOut.CaptureCount);
}
public virtual void TestRandom()
{
int alphabetSize = TestUtil.NextInt(Random(), 2, 7);
int docLen = AtLeast(3000);
//final int docLen = 50;
string document = GetRandomString('a', alphabetSize, docLen);
if (VERBOSE)
{
Console.WriteLine("TEST: doc=" + document);
}
int numSyn = AtLeast(5);
//final int numSyn = 2;
IDictionary<string, OneSyn> synMap = new Dictionary<string, OneSyn>();
IList<OneSyn> syns = new List<OneSyn>();
bool dedup = Random().nextBoolean();
if (VERBOSE)
{
Console.WriteLine(" dedup=" + dedup);
}
b = new SynonymMap.Builder(dedup);
for (int synIDX = 0; synIDX < numSyn; synIDX++)
{
string synIn = GetRandomString('a', alphabetSize, TestUtil.NextInt(Random(), 1, 5)).Trim();
OneSyn s = synMap.ContainsKey(synIn) ? synMap[synIn] : null;
if (s == null)
{
s = new OneSyn();
s.@in = synIn;
syns.Add(s);
s.@out = new List<string>();
synMap[synIn] = s;
s.keepOrig = Random().nextBoolean();
}
string synOut = GetRandomString('0', 10, TestUtil.NextInt(Random(), 1, 5)).Trim();
[email protected](synOut);
Add(synIn, synOut, s.keepOrig);
if (VERBOSE)
{
Console.WriteLine(" syns[" + synIDX + "] = " + s.@in + " -> " + s.@out + " keepOrig=" + s.keepOrig);
}
}
tokensIn = new MockTokenizer(new StringReader("a"), MockTokenizer.WHITESPACE, true);
tokensIn.Reset();
assertTrue(tokensIn.IncrementToken());
assertFalse(tokensIn.IncrementToken());
tokensIn.End();
tokensIn.Dispose();
tokensOut = new SynonymFilter(tokensIn, b.Build(), true);
termAtt = tokensOut.AddAttribute<ICharTermAttribute>();
posIncrAtt = tokensOut.AddAttribute<IPositionIncrementAttribute>();
posLenAtt = tokensOut.AddAttribute<IPositionLengthAttribute>();
offsetAtt = tokensOut.AddAttribute<IOffsetAttribute>();
if (dedup)
{
PruneDups(syns);
}
string expected = SlowSynMatcher(document, syns, 5);
if (VERBOSE)
{
Console.WriteLine("TEST: expected=" + expected);
}
Verify(document, expected);
}
// offsetsAreCorrect also validates:
// - graph offsets are correct (all tokens leaving from
// pos X have the same startOffset; all tokens
// arriving to pos Y have the same endOffset)
// - offsets only move forwards (startOffset >=
// lastStartOffset)
public static void AssertTokenStreamContents(TokenStream ts, string[] output, int[] startOffsets, int[] endOffsets, string[] types, int[] posIncrements, int[] posLengths, int?finalOffset, int?finalPosInc, bool[] keywordAtts, bool offsetsAreCorrect)
{
Assert.IsNotNull(output);
var checkClearAtt = ts.AddAttribute <ICheckClearAttributesAttribute>();
ICharTermAttribute termAtt = null;
if (output.Length > 0)
{
Assert.IsTrue(ts.HasAttribute <ICharTermAttribute>(), "has no CharTermAttribute");
termAtt = ts.GetAttribute <ICharTermAttribute>();
}
IOffsetAttribute offsetAtt = null;
if (startOffsets != null || endOffsets != null || finalOffset != null)
{
Assert.IsTrue(ts.HasAttribute <IOffsetAttribute>(), "has no OffsetAttribute");
offsetAtt = ts.GetAttribute <IOffsetAttribute>();
}
ITypeAttribute typeAtt = null;
if (types != null)
{
Assert.IsTrue(ts.HasAttribute <ITypeAttribute>(), "has no TypeAttribute");
typeAtt = ts.GetAttribute <ITypeAttribute>();
}
IPositionIncrementAttribute posIncrAtt = null;
if (posIncrements != null || finalPosInc != null)
{
Assert.IsTrue(ts.HasAttribute <IPositionIncrementAttribute>(), "has no PositionIncrementAttribute");
posIncrAtt = ts.GetAttribute <IPositionIncrementAttribute>();
}
IPositionLengthAttribute posLengthAtt = null;
if (posLengths != null)
{
Assert.IsTrue(ts.HasAttribute <IPositionLengthAttribute>(), "has no PositionLengthAttribute");
posLengthAtt = ts.GetAttribute <IPositionLengthAttribute>();
}
IKeywordAttribute keywordAtt = null;
if (keywordAtts != null)
{
Assert.IsTrue(ts.HasAttribute <IKeywordAttribute>(), "has no KeywordAttribute");
keywordAtt = ts.GetAttribute <IKeywordAttribute>();
}
// Maps position to the start/end offset:
IDictionary <int?, int?> posToStartOffset = new Dictionary <int?, int?>();
IDictionary <int?, int?> posToEndOffset = new Dictionary <int?, int?>();
ts.Reset();
int pos = -1;
int lastStartOffset = 0;
for (int i = 0; i < output.Length; i++)
{
// extra safety to enforce, that the state is not preserved and also assign bogus values
ts.ClearAttributes();
termAtt.SetEmpty().Append("bogusTerm");
if (offsetAtt != null)
{
offsetAtt.SetOffset(14584724, 24683243);
}
if (typeAtt != null)
{
typeAtt.Type = "bogusType";
}
if (posIncrAtt != null)
{
posIncrAtt.PositionIncrement = 45987657;
}
if (posLengthAtt != null)
{
posLengthAtt.PositionLength = 45987653;
}
if (keywordAtt != null)
{
keywordAtt.Keyword = (i & 1) == 0;
}
bool reset = checkClearAtt.AndResetClearCalled; // reset it, because we called clearAttribute() before
Assert.IsTrue(ts.IncrementToken(), "token " + i + " does not exist");
Assert.IsTrue(reset, "ClearAttributes() was not called correctly in TokenStream chain");
Assert.AreEqual(output[i], termAtt.ToString(), "term " + i + ", output[i] = " + output[i] + ", termAtt = " + termAtt.ToString());
if (startOffsets != null)
{
Assert.AreEqual(startOffsets[i], offsetAtt.StartOffset(), "startOffset " + i);
}
if (endOffsets != null)
{
Assert.AreEqual(endOffsets[i], offsetAtt.EndOffset(), "endOffset " + i);
}
if (types != null)
{
Assert.AreEqual(types[i], typeAtt.Type, "type " + i);
}
if (posIncrements != null)
{
Assert.AreEqual(posIncrements[i], posIncrAtt.PositionIncrement, "posIncrement " + i);
}
if (posLengths != null)
{
Assert.AreEqual(posLengths[i], posLengthAtt.PositionLength, "posLength " + i);
}
if (keywordAtts != null)
{
Assert.AreEqual(keywordAtts[i], keywordAtt.Keyword, "keywordAtt " + i);
}
// we can enforce some basic things about a few attributes even if the caller doesn't check:
if (offsetAtt != null)
{
int startOffset = offsetAtt.StartOffset();
int endOffset = offsetAtt.EndOffset();
if (finalOffset != null)
{
Assert.IsTrue(startOffset <= (int)finalOffset, "startOffset must be <= finalOffset");
Assert.IsTrue(endOffset <= (int)finalOffset, "endOffset must be <= finalOffset: got endOffset=" + endOffset + " vs finalOffset=" + (int)finalOffset);
}
if (offsetsAreCorrect)
{
Assert.IsTrue(offsetAtt.StartOffset() >= lastStartOffset, "offsets must not go backwards startOffset=" + startOffset + " is < lastStartOffset=" + lastStartOffset);
lastStartOffset = offsetAtt.StartOffset();
}
if (offsetsAreCorrect && posLengthAtt != null && posIncrAtt != null)
{
// Validate offset consistency in the graph, ie
// all tokens leaving from a certain pos have the
// same startOffset, and all tokens arriving to a
// certain pos have the same endOffset:
int posInc = posIncrAtt.PositionIncrement;
pos += posInc;
int posLength = posLengthAtt.PositionLength;
if (!posToStartOffset.ContainsKey(pos))
{
// First time we've seen a token leaving from this position:
posToStartOffset[pos] = startOffset;
//System.out.println(" + s " + pos + " -> " + startOffset);
}
else
{
// We've seen a token leaving from this position
// before; verify the startOffset is the same:
//System.out.println(" + vs " + pos + " -> " + startOffset);
Assert.AreEqual((int)posToStartOffset[pos], startOffset, "pos=" + pos + " posLen=" + posLength + " token=" + termAtt);
}
int endPos = pos + posLength;
if (!posToEndOffset.ContainsKey(endPos))
{
// First time we've seen a token arriving to this position:
posToEndOffset[endPos] = endOffset;
//System.out.println(" + e " + endPos + " -> " + endOffset);
}
else
{
// We've seen a token arriving to this position
// before; verify the endOffset is the same:
//System.out.println(" + ve " + endPos + " -> " + endOffset);
Assert.AreEqual((int)posToEndOffset[endPos], endOffset, "pos=" + pos + " posLen=" + posLength + " token=" + termAtt);
}
}
}
if (posIncrAtt != null)
{
if (i == 0)
{
Assert.IsTrue(posIncrAtt.PositionIncrement >= 1, "first posIncrement must be >= 1");
}
else
{
Assert.IsTrue(posIncrAtt.PositionIncrement >= 0, "posIncrement must be >= 0");
}
}
if (posLengthAtt != null)
{
Assert.IsTrue(posLengthAtt.PositionLength >= 1, "posLength must be >= 1");
}
}
if (ts.IncrementToken())
{
Assert.Fail("TokenStream has more tokens than expected (expected count=" + output.Length + "); extra token=" + termAtt);
}
// repeat our extra safety checks for End()
ts.ClearAttributes();
if (termAtt != null)
{
termAtt.SetEmpty().Append("bogusTerm");
}
if (offsetAtt != null)
{
offsetAtt.SetOffset(14584724, 24683243);
}
if (typeAtt != null)
{
typeAtt.Type = "bogusType";
}
if (posIncrAtt != null)
{
posIncrAtt.PositionIncrement = 45987657;
}
if (posLengthAtt != null)
{
posLengthAtt.PositionLength = 45987653;
}
var reset_ = checkClearAtt.AndResetClearCalled; // reset it, because we called clearAttribute() before
ts.End();
Assert.IsTrue(checkClearAtt.AndResetClearCalled, "super.End()/ClearAttributes() was not called correctly in End()");
if (finalOffset != null)
{
Assert.AreEqual((int)finalOffset, offsetAtt.EndOffset(), "finalOffset");
}
if (offsetAtt != null)
{
Assert.IsTrue(offsetAtt.EndOffset() >= 0, "finalOffset must be >= 0");
}
if (finalPosInc != null)
{
Assert.AreEqual((int)finalPosInc, posIncrAtt.PositionIncrement, "finalPosInc");
}
ts.Dispose();
}
public virtual void TestOutputHangsOffEnd()
{
b = new SynonymMap.Builder(true);
const bool keepOrig = false;
// b hangs off the end (no input token under it):
Add("a", "a b", keepOrig);
tokensIn = new MockTokenizer(new StringReader("a"), MockTokenizer.WHITESPACE, true);
tokensIn.Reset();
assertTrue(tokensIn.IncrementToken());
assertFalse(tokensIn.IncrementToken());
tokensIn.End();
tokensIn.Dispose();
tokensOut = new SynonymFilter(tokensIn, b.Build(), true);
termAtt = tokensOut.AddAttribute<ICharTermAttribute>();
posIncrAtt = tokensOut.AddAttribute<IPositionIncrementAttribute>();
offsetAtt = tokensOut.AddAttribute<IOffsetAttribute>();
posLenAtt = tokensOut.AddAttribute<IPositionLengthAttribute>();
// Make sure endOffset inherits from previous input token:
Verify("a", "a b:1");
}
public EdgeNGramTokenFilter(LuceneVersion version, TokenStream input, Side side, int minGram, int maxGram)
: base(input)
{
//if (version == null)
//{
// throw new System.ArgumentException("version must not be null");
//}
if (version.OnOrAfter(LuceneVersion.LUCENE_44) && side == Side.BACK)
{
throw new System.ArgumentException("Side.BACK is not supported anymore as of Lucene 4.4, use ReverseStringFilter up-front and afterward");
}
if (!Enum.IsDefined(typeof(Side), side))
{
throw new System.ArgumentException("sideLabel must be either front or back");
}
if (minGram < 1)
{
throw new System.ArgumentException("minGram must be greater than zero");
}
if (minGram > maxGram)
{
throw new System.ArgumentException("minGram must not be greater than maxGram");
}
this.version = version;
this.charUtils = version.OnOrAfter(LuceneVersion.LUCENE_44) ? CharacterUtils.GetInstance(version) : CharacterUtils.Java4Instance;
this.minGram = minGram;
this.maxGram = maxGram;
this.side = side;
this.termAtt = AddAttribute<ICharTermAttribute>();
this.offsetAtt = AddAttribute<IOffsetAttribute>();
this.posIncrAtt = AddAttribute<IPositionIncrementAttribute>();
this.posLenAtt = AddAttribute<IPositionLengthAttribute>();
}
/// <summary>
/// Create a new CJKBigramFilter, specifying which writing systems should be bigrammed,
/// and whether or not unigrams should also be output. </summary>
/// <param name="flags"> OR'ed set from <seealso cref="CJKBigramFilter#HAN"/>, <seealso cref="CJKBigramFilter#HIRAGANA"/>,
/// <seealso cref="CJKBigramFilter#KATAKANA"/>, <seealso cref="CJKBigramFilter#HANGUL"/> </param>
/// <param name="outputUnigrams"> true if unigrams for the selected writing systems should also be output.
/// when this is false, this is only done when there are no adjacent characters to form
/// a bigram. </param>
public CJKBigramFilter(TokenStream @in, int flags, bool outputUnigrams)
: base(@in)
{
doHan = (flags & HAN) == 0 ? NO : HAN_TYPE;
doHiragana = (flags & HIRAGANA) == 0 ? NO : HIRAGANA_TYPE;
doKatakana = (flags & KATAKANA) == 0 ? NO : KATAKANA_TYPE;
doHangul = (flags & HANGUL) == 0 ? NO : HANGUL_TYPE;
this.outputUnigrams = outputUnigrams;
this.termAtt = AddAttribute<ICharTermAttribute>();
this.typeAtt = AddAttribute<ITypeAttribute>();
this.offsetAtt = AddAttribute<IOffsetAttribute>();
this.posIncAtt = AddAttribute<IPositionIncrementAttribute>();
this.posLengthAtt = AddAttribute<IPositionLengthAttribute>();
}
/// <summary>
/// Retrieve suggestions.
/// </summary>
public virtual IList <LookupResult> DoLookup(string key, IEnumerable <BytesRef> contexts, int num)
{
if (contexts != null)
{
throw new System.ArgumentException("this suggester doesn't support contexts");
}
TokenStream ts = queryAnalyzer.GetTokenStream("", key.ToString());
try
{
ITermToBytesRefAttribute termBytesAtt = ts.AddAttribute <ITermToBytesRefAttribute>();
IOffsetAttribute offsetAtt = ts.AddAttribute <IOffsetAttribute>();
IPositionLengthAttribute posLenAtt = ts.AddAttribute <IPositionLengthAttribute>();
IPositionIncrementAttribute posIncAtt = ts.AddAttribute <IPositionIncrementAttribute>();
ts.Reset();
var lastTokens = new BytesRef[grams];
//System.out.println("lookup: key='" + key + "'");
// Run full analysis, but save only the
// last 1gram, last 2gram, etc.:
BytesRef tokenBytes = termBytesAtt.BytesRef;
int maxEndOffset = -1;
bool sawRealToken = false;
while (ts.IncrementToken())
{
termBytesAtt.FillBytesRef();
sawRealToken |= tokenBytes.Length > 0;
// TODO: this is somewhat iffy; today, ShingleFilter
// sets posLen to the gram count; maybe we should make
// a separate dedicated att for this?
int gramCount = posLenAtt.PositionLength;
Debug.Assert(gramCount <= grams);
// Safety: make sure the recalculated count "agrees":
if (CountGrams(tokenBytes) != gramCount)
{
throw new System.ArgumentException("tokens must not contain separator byte; got token=" + tokenBytes + " but gramCount=" + gramCount + " does not match recalculated count=" + CountGrams(tokenBytes));
}
maxEndOffset = Math.Max(maxEndOffset, offsetAtt.EndOffset);
lastTokens[gramCount - 1] = BytesRef.DeepCopyOf(tokenBytes);
}
ts.End();
if (!sawRealToken)
{
throw new System.ArgumentException("no tokens produced by analyzer, or the only tokens were empty strings");
}
// Carefully fill last tokens with _ tokens;
// ShingleFilter appraently won't emit "only hole"
// tokens:
int endPosInc = posIncAtt.PositionIncrement;
// Note this will also be true if input is the empty
// string (in which case we saw no tokens and
// maxEndOffset is still -1), which in fact works out OK
// because we fill the unigram with an empty BytesRef
// below:
bool lastTokenEnded = offsetAtt.EndOffset > maxEndOffset || endPosInc > 0;
//System.out.println("maxEndOffset=" + maxEndOffset + " vs " + offsetAtt.EndOffset);
if (lastTokenEnded)
{
//System.out.println(" lastTokenEnded");
// If user hit space after the last token, then
// "upgrade" all tokens. This way "foo " will suggest
// all bigrams starting w/ foo, and not any unigrams
// starting with "foo":
for (int i = grams - 1; i > 0; i--)
{
BytesRef token = lastTokens[i - 1];
if (token == null)
{
continue;
}
token.Grow(token.Length + 1);
token.Bytes[token.Length] = separator;
token.Length++;
lastTokens[i] = token;
}
lastTokens[0] = new BytesRef();
}
var arc = new FST.Arc <long?>();
var bytesReader = fst.GetBytesReader();
// Try highest order models first, and if they return
// results, return that; else, fallback:
double backoff = 1.0;
List <LookupResult> results = new List <LookupResult>(num);
// We only add a given suffix once, from the highest
// order model that saw it; for subsequent lower order
// models we skip it:
var seen = new HashSet <BytesRef>();
for (int gram = grams - 1; gram >= 0; gram--)
{
BytesRef token = lastTokens[gram];
// Don't make unigram predictions from empty string:
if (token == null || (token.Length == 0 && key.Length > 0))
{
// Input didn't have enough tokens:
//System.out.println(" gram=" + gram + ": skip: not enough input");
continue;
}
if (endPosInc > 0 && gram <= endPosInc)
{
// Skip hole-only predictions; in theory we
// shouldn't have to do this, but we'd need to fix
// ShingleFilter to produce only-hole tokens:
//System.out.println(" break: only holes now");
break;
}
//System.out.println("try " + (gram+1) + " gram token=" + token.utf8ToString());
// TODO: we could add fuzziness here
// match the prefix portion exactly
//Pair<Long,BytesRef> prefixOutput = null;
long?prefixOutput = null;
try
{
prefixOutput = LookupPrefix(fst, bytesReader, token, arc);
}
catch (IOException bogus)
{
throw new Exception(bogus.ToString(), bogus);
}
//System.out.println(" prefixOutput=" + prefixOutput);
if (prefixOutput == null)
{
// This model never saw this prefix, e.g. the
// trigram model never saw context "purple mushroom"
backoff *= ALPHA;
continue;
}
// TODO: we could do this division at build time, and
// bake it into the FST?
// Denominator for computing scores from current
// model's predictions:
long contextCount = totTokens;
BytesRef lastTokenFragment = null;
for (int i = token.Length - 1; i >= 0; i--)
{
if (token.Bytes[token.Offset + i] == separator)
{
BytesRef context = new BytesRef(token.Bytes, token.Offset, i);
long? output = Lucene.Net.Util.Fst.Util.Get(fst, Lucene.Net.Util.Fst.Util.ToInt32sRef(context, new Int32sRef()));
Debug.Assert(output != null);
contextCount = DecodeWeight(output);
lastTokenFragment = new BytesRef(token.Bytes, token.Offset + i + 1, token.Length - i - 1);
break;
}
}
BytesRef finalLastToken;
if (lastTokenFragment == null)
{
finalLastToken = BytesRef.DeepCopyOf(token);
}
else
{
finalLastToken = BytesRef.DeepCopyOf(lastTokenFragment);
}
Debug.Assert(finalLastToken.Offset == 0);
CharsRef spare = new CharsRef();
// complete top-N
Util.Fst.Util.TopResults <long?> completions = null;
try
{
// Because we store multiple models in one FST
// (1gram, 2gram, 3gram), we must restrict the
// search so that it only considers the current
// model. For highest order model, this is not
// necessary since all completions in the FST
// must be from this model, but for lower order
// models we have to filter out the higher order
// ones:
// Must do num+seen.size() for queue depth because we may
// reject up to seen.size() paths in acceptResult():
Util.Fst.Util.TopNSearcher <long?> searcher = new TopNSearcherAnonymousInnerClassHelper(this, fst, num, num + seen.Count, weightComparer, seen, finalLastToken);
// since this search is initialized with a single start node
// it is okay to start with an empty input path here
searcher.AddStartPaths(arc, prefixOutput, true, new Int32sRef());
completions = searcher.Search();
Debug.Assert(completions.IsComplete);
}
catch (IOException bogus)
{
throw new Exception(bogus.ToString(), bogus);
}
int prefixLength = token.Length;
BytesRef suffix = new BytesRef(8);
//System.out.println(" " + completions.length + " completions");
foreach (Util.Fst.Util.Result <long?> completion in completions)
{
token.Length = prefixLength;
// append suffix
Util.Fst.Util.ToBytesRef(completion.Input, suffix);
token.Append(suffix);
//System.out.println(" completion " + token.utf8ToString());
// Skip this path if a higher-order model already
// saw/predicted its last token:
BytesRef lastToken = token;
for (int i = token.Length - 1; i >= 0; i--)
{
if (token.Bytes[token.Offset + i] == separator)
{
Debug.Assert(token.Length - i - 1 > 0);
lastToken = new BytesRef(token.Bytes, token.Offset + i + 1, token.Length - i - 1);
break;
}
}
if (seen.Contains(lastToken))
{
//System.out.println(" skip dup " + lastToken.utf8ToString());
goto nextCompletionContinue;
}
seen.Add(BytesRef.DeepCopyOf(lastToken));
spare.Grow(token.Length);
UnicodeUtil.UTF8toUTF16(token, spare);
LookupResult result = new LookupResult(spare.ToString(),
// LUCENENET NOTE: We need to calculate this as decimal because when using double it can sometimes
// return numbers that are greater than long.MaxValue, which results in a negative long number.
(long)(long.MaxValue * (decimal)backoff * ((decimal)DecodeWeight(completion.Output)) / contextCount));
results.Add(result);
Debug.Assert(results.Count == seen.Count);
//System.out.println(" add result=" + result);
nextCompletionContinue :;
}
backoff *= ALPHA;
}
results.Sort(new ComparerAnonymousInnerClassHelper(this));
if (results.Count > num)
{
results.SubList(num, results.Count).Clear();
}
return(results);
}
finally
{
IOUtils.DisposeWhileHandlingException(ts);
}
}
/// <summary>
/// Constructs a ShingleFilter with the specified shingle size from the
/// <seealso cref="TokenStream"/> <code>input</code>
/// </summary>
/// <param name="input"> input stream </param>
/// <param name="minShingleSize"> minimum shingle size produced by the filter. </param>
/// <param name="maxShingleSize"> maximum shingle size produced by the filter. </param>
public ShingleFilter(TokenStream input, int minShingleSize, int maxShingleSize)
: base(input)
{
MaxShingleSize = maxShingleSize;
MinShingleSize = minShingleSize;
termAtt = AddAttribute<ICharTermAttribute>();
offsetAtt = AddAttribute<IOffsetAttribute>();
posIncrAtt = AddAttribute<IPositionIncrementAttribute>();
posLenAtt = AddAttribute<IPositionLengthAttribute>();
typeAtt = AddAttribute<ITypeAttribute>();
}
/// <summary>
/// Creates NGramTokenFilter with given min and max n-grams. </summary>
/// <param name="version"> Lucene version to enable correct position increments.
/// See <a href="#version">above</a> for details. </param>
/// <param name="input"> <seealso cref="TokenStream"/> holding the input to be tokenized </param>
/// <param name="minGram"> the smallest n-gram to generate </param>
/// <param name="maxGram"> the largest n-gram to generate </param>
public NGramTokenFilter(LuceneVersion version, TokenStream input, int minGram, int maxGram)
: base(new CodepointCountFilter(version, input, minGram, int.MaxValue))
{
this.version = version;
this.charUtils = version.OnOrAfter(
#pragma warning disable 612, 618
LuceneVersion.LUCENE_44) ?
#pragma warning restore 612, 618
CharacterUtils.GetInstance(version) : CharacterUtils.Java4Instance;
if (minGram < 1)
{
throw new System.ArgumentException("minGram must be greater than zero");
}
if (minGram > maxGram)
{
throw new System.ArgumentException("minGram must not be greater than maxGram");
}
this.minGram = minGram;
this.maxGram = maxGram;
#pragma warning disable 612, 618
if (version.OnOrAfter(LuceneVersion.LUCENE_44))
#pragma warning restore 612, 618
{
posIncAtt = AddAttribute<IPositionIncrementAttribute>();
posLenAtt = AddAttribute<IPositionLengthAttribute>();
}
else
{
posIncAtt = new PositionIncrementAttributeAnonymousInnerClassHelper(this);
posLenAtt = new PositionLengthAttributeAnonymousInnerClassHelper(this);
}
termAtt = AddAttribute<ICharTermAttribute>();
offsetAtt = AddAttribute<IOffsetAttribute>();
}
public virtual void TestRandom()
{
int alphabetSize = TestUtil.NextInt32(Random, 2, 7);
int docLen = AtLeast(3000);
//final int docLen = 50;
string document = GetRandomString('a', alphabetSize, docLen);
if (Verbose)
{
Console.WriteLine("TEST: doc=" + document);
}
int numSyn = AtLeast(5);
//final int numSyn = 2;
IDictionary <string, OneSyn> synMap = new Dictionary <string, OneSyn>();
IList <OneSyn> syns = new JCG.List <OneSyn>();
bool dedup = Random.nextBoolean();
if (Verbose)
{
Console.WriteLine(" dedup=" + dedup);
}
b = new SynonymMap.Builder(dedup);
for (int synIDX = 0; synIDX < numSyn; synIDX++)
{
string synIn = GetRandomString('a', alphabetSize, TestUtil.NextInt32(Random, 1, 5)).Trim();
if (!synMap.TryGetValue(synIn, out OneSyn s) || s is null)
{
s = new OneSyn();
s.@in = synIn;
syns.Add(s);
s.@out = new JCG.List <string>();
synMap[synIn] = s;
s.keepOrig = Random.nextBoolean();
}
string synOut = GetRandomString('0', 10, TestUtil.NextInt32(Random, 1, 5)).Trim();
[email protected](synOut);
Add(synIn, synOut, s.keepOrig);
if (Verbose)
{
Console.WriteLine(" syns[" + synIDX + "] = " + s.@in + " -> " + s.@out + " keepOrig=" + s.keepOrig);
}
}
tokensIn = new MockTokenizer(new StringReader("a"), MockTokenizer.WHITESPACE, true);
tokensIn.Reset();
assertTrue(tokensIn.IncrementToken());
assertFalse(tokensIn.IncrementToken());
tokensIn.End();
tokensIn.Dispose();
tokensOut = new SynonymFilter(tokensIn, b.Build(), true);
termAtt = tokensOut.AddAttribute <ICharTermAttribute>();
posIncrAtt = tokensOut.AddAttribute <IPositionIncrementAttribute>();
posLenAtt = tokensOut.AddAttribute <IPositionLengthAttribute>();
offsetAtt = tokensOut.AddAttribute <IOffsetAttribute>();
if (dedup)
{
PruneDups(syns);
}
string expected = SlowSynMatcher(document, syns, 5);
if (Verbose)
{
Console.WriteLine("TEST: expected=" + expected);
}
Verify(document, expected);
}
private void Init(LuceneVersion version, int minGram, int maxGram, bool edgesOnly)
{
#pragma warning disable 612, 618
if (!version.OnOrAfter(LuceneVersion.LUCENE_44))
#pragma warning restore 612, 618
{
throw new System.ArgumentException("This class only works with Lucene 4.4+. To emulate the old (broken) behavior of NGramTokenizer, use Lucene43NGramTokenizer/Lucene43EdgeNGramTokenizer");
}
#pragma warning disable 612, 618
charUtils = version.OnOrAfter(LuceneVersion.LUCENE_44) ?
#pragma warning restore 612, 618
CharacterUtils.GetInstance(version) : CharacterUtils.Java4Instance;
if (minGram < 1)
{
throw new System.ArgumentException("minGram must be greater than zero");
}
if (minGram > maxGram)
{
throw new System.ArgumentException("minGram must not be greater than maxGram");
}
termAtt = AddAttribute<ICharTermAttribute>();
posIncAtt = AddAttribute<IPositionIncrementAttribute>();
posLenAtt = AddAttribute<IPositionLengthAttribute>();
offsetAtt = AddAttribute<IOffsetAttribute>();
this.minGram = minGram;
this.maxGram = maxGram;
this.edgesOnly = edgesOnly;
charBuffer = CharacterUtils.NewCharacterBuffer(2 * maxGram + 1024); // 2 * maxGram in case all code points require 2 chars and + 1024 for buffering to not keep polling the Reader
buffer = new int[charBuffer.Buffer.Length];
// Make the term att large enough
termAtt.ResizeBuffer(2 * maxGram);
}
