+ * By default, the following processing is performed: + *
+ * Default caseless matching, or case-folding is more than just conversion to + * lowercase. For example, it handles cases such as the Greek sigma, so that + * "Μάϊος" and "ΜΆΪΟΣ" will match correctly. + *
+ *+ * Case-folding is still only an approximation of the language-specific rules + * governing case. If the specific language is known, consider using + * ICUCollationKeyFilter and indexing collation keys instead. This filter + * performs the "full" case-folding specified in the Unicode standard, and this + * may change the length of the term. For example, the German ß is case-folded + * to the string 'ss'. + *
+ *+ * This filter respects both canonical and compatibility equivalence. The + * default caseless matching in the Unicode standard respects canonical + * equivalence. Additional logic has been added to this filter to respect + * compatibility equivalence, avoiding an extra normalization. This logic also + * ensures the output is closed under all compatibility forms. You do not + * need to normalize before folding, or fold and normalize twice. Closure means + * all of the below conditions are true: + *
+ * This filter does not preserve normalization forms. Instead, suggested usage + * is to first case-fold, then normalize. + *
+ * For more details, see Unicode Standard sections 5.18: Caseless + * Matching and 3.13: Default + * Caseless Matching + */ + +public final class ICUCaseFoldingFilter extends TokenFilter { + // this is true for NFKC or NFKD. When set, the FC_NFKC_Closure mappings from + // DerivedNormalizationProps.txt will be applied. + private final boolean nfkcClosure; + + // case-folding output buffer, will be resized if necessary. + private char buffer[] = new char[4096]; + + // new api term attribute, will be updated with folded text. + private TermAttribute termAtt; + + /** + * Create a new ICUCaseFoldingFilter, operating on the provided input stream. + * Output will be case-folded and closed under the supplied normalization + * mode. + * + * @param input {@link TokenStream} to filter + * @param mode Normalization mode hint, will ensure output is closed under + * that mode. + */ + public ICUCaseFoldingFilter(TokenStream input, Normalizer.Mode mode) { + super(input); + nfkcClosure = (mode == Normalizer.NFKC || mode == Normalizer.NFKD); + termAtt = (TermAttribute) addAttribute(TermAttribute.class); + } + + /* + * Algorithm: Up-front, the set of BMP codepoints that are case-insensitive + * are calculated. For these, and for basic latin, perform simple lower-case, + * as its equivalent to case-folding. + * + * If a surrogate or case-sensitive (outside of basic latin) character is + * encountered, bail and proceed down the slower path. + */ + + public boolean incrementToken() throws IOException { + + if (input.incrementToken()) { + final char src[] = termAtt.termBuffer(); + final int length = termAtt.termLength(); + + for (int i = 0; i < length; i++) { + final char ch = src[i]; + + if (ch <= 0x7F /* basic latin */ + || (!UTF16.isSurrogate(ch) && insensitive.contains(ch))) { + src[i] = Character.toLowerCase(ch); // either basic latin or a + // BMP-case-insensitive codepoint + } else { + + /* + * a case-sensitive codepoint outside of basic latin, or a surrogate + * has been encountered. bail out completely and invoke the 'slow' + * case folding algorithm. + */ + + final int requiredLength = length << 2; // Max expansion factor: 3x + // for case folding. 4x for + // case folding + NFKC closure + + if (buffer.length < requiredLength) + buffer = new char[ArrayUtil.getNextSize(requiredLength)]; + + final int newLength = fold(src, length, buffer, nfkcClosure); + termAtt.setTermBuffer(buffer, 0, newLength); + return true; + } + } + + return true; + } else { + return false; + } + } + + /* + * What ICU should have, some way to do full case folding on char[] Behind the + * scenes this is still bad for the single-many case (look at UCaseProps.java) + * For really good performance, its not hard to create IntTrie data files from + * CaseFolding.txt and DerivedNormalizationProps.txt But this would introduce + * a maintenance hassle: would have to update data files whenever the Unicode + * standard is updated. Finally, there is a ticket to improve this in ICU: + * http://bugs.icu-project.org/trac/ticket/5072 + * + * Add a function that performs string normalization and case folding + * according to the Unicode Standard, chapter 5 "Implementation Guidelines", + * section "Case Mappings". The availability of such a function would help + * avoid errors by users who are unaware of the complications involved. There + * should be a parameter for the output normalization form. For "K" forms, the + * FC_NFKC_Closure should be applied as well. + * + * Until this functionality is implemented, this looks to be the best overall + * tradeoff. When it is, the logic below should be removed! + */ + + // Low-level unicode case properties functionality + private static final UCaseProps caseProps; + + /* + * Output buffer for when a single codepoint folds to multiple codepoints, + * required by UCaseProps. Fortunately, this is only used for the case where + * the length of a string is increased. Unfortunately, behind the scenes + * UCaseProps appends to this buffer with 'new String(xxx)'... + */ + private final StringBuffer foldOut = new StringBuffer(4); + + // Output buffer for when a codepoint has an FC_NFKC_Closure mapping + private final char replacement[] = new char[4]; + + // set of case-insensitive BMP characters. + private static final UnicodeSet insensitive; + + static { + + /* + * In the ICU UCharacter implementation there is some logic here involving + * getDummy(). This is supposedly for the case in which the case-properties + * data cannot be loaded into memory. + * + * In this case, there are bigger problems if there isn't enough memory for + * the JVM to even lowercase! + */ + + try { + caseProps = UCaseProps.getSingleton(); + } catch (IOException e) { + throw new RuntimeException(e); + } + + /* + * Perhaps too conservative, but still over 60,000 BMP code points. This is + * nice because it is correct even if FC_NFKC_Closure mapping is being + * applied (NFKD/NFKC case) If the codepoint has NFKC_Quick_Check=Yes, then + * it cannot have an FC_NFKC_Closure mapping. Substract from this set any of + * these that are case-sensitive. + */ + + insensitive = new UnicodeSet( + "[[:NFKC_Quick_Check=Yes:]-[:Case_Sensitive=True:]]"); + insensitive.compact(); + insensitive.freeze(); + } + + /** + * Folds characters from src to dst, optionally applying FC_NFKC_Closure + * mappings. The output buffer must have enough storage to contain the folded + * text. + * + * There is no such maximum limit specified in the standard, but a maximum + * expansion factor of 3x is specified here: + * http://unicode.org/reports/tr36/tr36-6.html#Buffer_Overflows + * + * When applying FC_NFKC_Closure mappings, this expansion factor increases to + * 4x. + * + * @param src input buffer + * @param length input length + * @param dst output buffer + * @param closure true if FC_NFKC_Closure mappings should be applied + * @return length of folded output. + */ + private int fold(char src[], int length, char dst[], boolean closure) { + int codepoint; + int folded; + int closed; + int newlen = 0; + + for (int i = 0; i < length; i += UTF16.getCharCount(codepoint)) { + codepoint = UTF16.charAt(src, 0, length, i); + + /* + * If closure mappings are requested, and a mapping is applied, then + * case-folding is complete for this codepoint. + */ + + if (closure) { + closed = Normalizer.getFC_NFKC_Closure(codepoint, replacement); + if (closed > 0) { // apply the closure mapping, no need to apply + // toFullFolding. + System.arraycopy(replacement, 0, dst, newlen, closed); + newlen += closed; + continue; + } + } + + /* + * The UCaseProps toFullFolding has several possible return values for + * input codepoint x: ~x: This means x is already case-folded. x <= + * MAX_STRING_LENGTH (31): This means x maps to multiple codepoints, which + * are placed in the supplied StringBuffer. In this case x represents the + * length of that sequence. x > MAX_STRING_LENGTH: x is the case-folded + * single-codepoint result. + */ + + folded = caseProps.toFullFolding(codepoint, foldOut, + UCharacter.FOLD_CASE_DEFAULT); + + if (folded < 0) { // codepoint is already folded + newlen += UCharacter.toChars(codepoint, dst, newlen); + } else if (folded <= UCaseProps.MAX_STRING_LENGTH) { // codepoint folds to + // multiple + // codepoints + foldOut.getChars(0, folded, dst, newlen); + foldOut.setLength(0); + newlen += folded; + } else if (folded < UTF16.SUPPLEMENTARY_MIN_VALUE) { // an attempt to + // optimize the BMP + // case. + dst[newlen++] = (char) folded; + } else { + newlen += UCharacter.toChars(folded, dst, newlen); + } + } + + return newlen; + } + +} Property changes on: contrib\icu\src\java\org\apache\lucene\icu\ICUCaseFoldingFilter.java ___________________________________________________________________ Added: svn:eol-style + native Index: contrib/icu/src/java/org/apache/lucene/icu/ICUDigitFoldingFilter.java =================================================================== --- contrib/icu/src/java/org/apache/lucene/icu/ICUDigitFoldingFilter.java (revision 0) +++ contrib/icu/src/java/org/apache/lucene/icu/ICUDigitFoldingFilter.java (revision 0) @@ -0,0 +1,113 @@ +package org.apache.lucene.icu; + +/** + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +import java.io.IOException; + +import org.apache.lucene.analysis.TokenFilter; +import org.apache.lucene.analysis.TokenStream; +import org.apache.lucene.analysis.tokenattributes.TermAttribute; +import org.apache.lucene.util.ArrayUtil; + +import com.ibm.icu.lang.UCharacter; +import com.ibm.icu.text.UTF16; + +/** + * A {@link TokenFilter} that folds numeric digits to their ASCII form. + *+ * There are many different representations of numeric digits for different + * scripts in Unicode. This filter folds numeric digits to their ASCII form. + *
+ * For example, '৭০৬' will be folded to '706' + * + */ + +public final class ICUDigitFoldingFilter extends TokenFilter { + + private char buffer[] = new char[4096]; + + private TermAttribute termAtt; + + public ICUDigitFoldingFilter(TokenStream input) { + super(input); + termAtt = (TermAttribute) addAttribute(TermAttribute.class); + } + + /* + * Algorithm: Native digits are typically rare, as are surrogates... Fast path + * is to simply verify the text has no surrogates or digits outside of basic + * latin. + * + * If a surrogate or native digit (outside of basic latin) character is + * encountered, bail and proceed down the slower path. + */ + + public boolean incrementToken() throws IOException { + if (input.incrementToken()) { + final char src[] = termAtt.termBuffer(); + final int length = termAtt.termLength(); + + for (int i = 0; i < length; i++) { + final char ch = src[i]; + + if (ch > 0x7F && (UTF16.isSurrogate(ch) || UCharacter.isDigit(ch))) { + + /* + * char is a surrogate or digit outside of basic latin bail out + * completely and invoke the slow folding algorithm + */ + + if (buffer.length < length) + buffer = new char[ArrayUtil.getNextSize(length)]; + + final int newLength = foldNumerics(src, length, buffer); + + termAtt.setTermBuffer(buffer, 0, newLength); + return true; + } + } + + return true; + } else { + return false; + } + } + + /* + * There are no closure issues here. For all codepoints, NFKC(foldNumeric(x)) + * = NFKC(foldNumeric(NFKC(foldNumeric(x)))) + */ + + private int foldNumerics(char src[], int length, char dst[]) { + int codepoint; + int folded; + int newlen = 0; + + for (int i = 0; i < length; i += UTF16.getCharCount(codepoint)) { + codepoint = UTF16.charAt(src, 0, length, i); + + if (UCharacter.isDigit(codepoint)) + folded = UCharacter.forDigit(UCharacter.getNumericValue(codepoint), 10); + else + folded = codepoint; + + newlen += UCharacter.toChars(folded, dst, newlen); + } + return newlen; + } +} Property changes on: contrib\icu\src\java\org\apache\lucene\icu\ICUDigitFoldingFilter.java ___________________________________________________________________ Added: svn:eol-style + native Index: contrib/icu/src/java/org/apache/lucene/icu/ICUFormatFilter.java =================================================================== --- contrib/icu/src/java/org/apache/lucene/icu/ICUFormatFilter.java (revision 0) +++ contrib/icu/src/java/org/apache/lucene/icu/ICUFormatFilter.java (revision 0) @@ -0,0 +1,119 @@ +package org.apache.lucene.icu; + +/** + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +import java.io.IOException; + +import org.apache.lucene.analysis.TokenFilter; +import org.apache.lucene.analysis.TokenStream; +import org.apache.lucene.analysis.tokenattributes.TermAttribute; +import org.apache.lucene.util.ArrayUtil; + +import com.ibm.icu.lang.UCharacter; +import com.ibm.icu.text.UTF16; +import com.ibm.icu.text.UnicodeSet; + +/** + * A {@link TokenFilter} that removes codepoints that affect the formatting and + * display of text. + *+ * Some codepoints in Unicode exist only to alter the formatting or display of + * text. This filter removes Format and Variation Selector codepoints. + *
+ * TODO: more docs + */ +public final class ICUFormatFilter extends TokenFilter { + + private char buffer[] = new char[4096]; + + private TermAttribute termAtt; + + public ICUFormatFilter(TokenStream input) { + super(input); + termAtt = (TermAttribute) addAttribute(TermAttribute.class); + } + + /* + * Algorithm: Format/Variation Selectors are typically rare, as are + * surrogates... Fast path is to simply verify the text has no surrogates or + * Format/Variation Selectors + * + * If a surrogate or Format/Variation character is encountered, bail and + * proceed down the slower path. + * + * The basic latin range is excluded because it contains no format codepoints, + * and because its much faster to exclude it than to execute the + * UnicodeSet.contains() binary search. + */ + + public boolean incrementToken() throws IOException { + if (input.incrementToken()) { + final char src[] = termAtt.termBuffer(); + final int length = termAtt.termLength(); + + for (int i = 0; i < length; i++) { + final char ch = src[i]; + + if (ch > 0x7F && (UTF16.isSurrogate(ch) || formatSet.contains(ch))) { + + /* + * char is a surrogate or format codepoint bail out completely and + * invoke the slow removal algorithm + */ + + if (buffer.length < length) + buffer = new char[ArrayUtil.getNextSize(length)]; + + final int newLength = removeFormat(src, length, buffer); + + termAtt.setTermBuffer(buffer, 0, newLength); + return true; + } + } + + return true; + } else { + return false; + } + } + + private static final UnicodeSet formatSet; + + static { + formatSet = new UnicodeSet( + "[[:General_Category=Format:][:Variation_Selector=True:]]"); + formatSet.compact(); + formatSet.freeze(); + } + + /* + * TODO: validate closure for NFKC/NFKD + */ + private static int removeFormat(char src[], int length, char dst[]) { + int codepoint; + int newlen = 0; + + for (int i = 0; i < length; i += UTF16.getCharCount(codepoint)) { + codepoint = UTF16.charAt(src, 0, length, i); + + if (!formatSet.contains(codepoint)) + newlen += UCharacter.toChars(codepoint, dst, newlen); + } + return newlen; + } +} Property changes on: contrib\icu\src\java\org\apache\lucene\icu\ICUFormatFilter.java ___________________________________________________________________ Added: svn:eol-style + native Index: contrib/icu/src/java/org/apache/lucene/icu/ICUNormalizationFilter.java =================================================================== --- contrib/icu/src/java/org/apache/lucene/icu/ICUNormalizationFilter.java (revision 0) +++ contrib/icu/src/java/org/apache/lucene/icu/ICUNormalizationFilter.java (revision 0) @@ -0,0 +1,249 @@ +package org.apache.lucene.icu; + +/** + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +import java.io.IOException; + +import org.apache.lucene.analysis.TokenFilter; +import org.apache.lucene.analysis.TokenStream; +import org.apache.lucene.analysis.tokenattributes.TermAttribute; +import org.apache.lucene.util.ArrayUtil; + +import org.apache.lucene.analysis.ASCIIFoldingFilter; // for javadoc only, example of a filter that does not respect canonical equivalence. + +import com.ibm.icu.text.Normalizer; + +/** + * A {@link TokenFilter} that performs Unicode text normalization. + *+ * Normalization standardizes different forms of the same character in Unicode. + * For any search application, it is essential to apply normalization to at + * least ensure canonical equivalence. For example, a Vietnamese input method on + * one operating system might represent the character ằ as one codepoint: LATIN + * SMALL LETTER A WITH BREVE AND GRAVE, whereas the input method on another + * operating system might represent the same character as two: LATIN SMALL + * LETTER A WITH BREVE followed by COMBINING GRAVE ACCENT. Unless text is + * normalized to a standard form, queries and documents from these different + * systems will not be interpreted as the same character! + *
+ *+ * There are four modes that text can be normalized to: + *
+ * For most search tasks, it makes sense to normalize to NFC or NFKC, as the + * composed form will generally be shorter than the decomposed form. The + * decomposed forms can still be useful for some tasks, for example providing + * decomposed Korean text to a downstream {@link TokenFilter} would allow that + * filter to work with individual Jamo instead of composed Hangul syllables. + *
+ *+ * For a typical search application, the way in which the text will be + * standardized is the most important, and the two types of standardization are + * described below. + *
+ *+ * Canonical equivalence is where there are multiple ways to encode the same + * character or sequence of characters in Unicode. These differences display the + * same to a user, but are different to the computer. + *
+ *+ * For example, é can be encoded in Unicode in at least two different ways: + *
+ * Compatibility equivalence is a weaker form of equivalence than canonical + * equivalence. Similar to canonical equivalents, compatibility equivalents are + * different ways to represent the same character. The difference is that unlike + * canonical equivalents, compatibility equivalents may have different visual + * appearance or format. + *
+ *+ * For example, the letter Ａ appears different than A, because of its width. The + * below two forms are not canonical equivalents, but are compatibility + * equivalents: + *
+ * Normalization is computationally expensive and can both reorder characters + * and change the length of text. In practice, typically the majority of text is + * already normalized. This filter first performs a quick-check, and performs + * normalization only when this quick-check fails or is uncertain. + *
+ *
+ * When designing an analysis pipeline, it is important to minimize the number
+ * of times you invoke ICUNormalizationFilter. At the same time, it is equally
+ * important that the analysis process behaves in such a way that all equivalent
+ * text is treated the same. The naïve solution to this problem is to invoke
+ * ICUNormalizationFilter both before and after every {@link TokenFilter} in the
+ * pipeline. This ensures that all equivalent text is treated the same and
+ * remains normalized, but is very inefficient.
+ * Instead the two simple rules below can be followed to minimize the number of
+ * invocations:
+ *
+ * If a {@link TokenFilter} does not respect the equivalence defined
+ * for the normalization form, ICUNormalizationFilter must be called before that
+ * {@link TokenFilter}.
+ * This way, text is provided to that TokenFilter in a form that it understands,
+ * and will be processed correctly. For example, the {@link ASCIIFoldingFilter}
+ * does not respect canonical equivalence: it only folds precomposed
+ * character+accent combinations to an accent-free form. Because of this, the
+ * two forms of é listed in the example above will be treated differently; only
+ * one will have its accent mark removed! By invoking ICUNormalizationFilter
+ * with NFC first, you can ensure that both are treated the same; both will have
+ * their accent marks removed.
+ *
+ * If a {@link TokenFilter} does not preserve the normalization form,
+ * ICUNormalizationFilter must be called at some point after that
+ * {@link TokenFilter} before indexing.
+ * When a {@link TokenFilter} modifies text, it might cause text to become
+ * denormalized. There are a number of ways this can happen, even concatenation
+ * of two normalized chunks of text can produce a denormalized result. For
+ * example, although the {@link ICUCaseFoldingFilter} respects both canonical
+ * and compatibility equivalence, it does not preserve normalization forms. By
+ * the first rule above, because it respects canonical equivalence, the
+ * ICUNormalizationFilter need not be invoked before it for normalization form
+ * NFC. But, because it does does not preserve normalization form NFC, the
+ * ICUNormalizationFilter must be invoked before indexing, or before any
+ * downstream TokenFilter that does not respect canonical equivalence, whichever
+ * comes first.
+ *
+ * It is generally more difficult to preserve normalization forms than it is to + * respect equivalence. Respecting equivalence is usually a simple matter of + * adding additional mappings. When designing an analysis pipeline, it is + * recommended that every {@link TokenFilter} respect equivalence, and at the + * end of the pipeline ICUNormalizationFilter can be invoked a single time. + *
+ * For more details, see UAX #15: Unicode Normalization Forms + */ + +public final class ICUNormalizationFilter extends TokenFilter { + // the mode this normalizer uses + private final Normalizer.Mode mode; + + // normalization output buffer, will be resized if needed. + private char buffer[] = new char[4096]; + + // new api term attribute, will be updated with normalized text if necessary. + private TermAttribute termAtt; + + /** + * Create an ICUNormalizationFilter that normalizes text to the specified + * mode. + * + * @param input {@link TokenStream} to filter + * @param mode Normalization mode to apply + */ + public ICUNormalizationFilter(TokenStream input, Normalizer.Mode mode) { + super(input); + this.mode = mode; + termAtt = (TermAttribute) addAttribute(TermAttribute.class); + } + + public boolean incrementToken() throws IOException { + + /* + * First do a quick-check (this will be the significant majority of text). + * If the text is already normalized, simply return it. Otherwise, normalize + * the text. + */ + + if (input.incrementToken()) { + final char src[] = termAtt.termBuffer(); + final int length = termAtt.termLength(); + + /* + * This quick-check returns three possible values: YES, NO, or MAYBE. When + * it returns YES, the text is already normalized. When it returns NO, the + * text is definitely not normalized. When it returns MAYBE, the only way + * to determine if the text is normalized is to actually normalize it. See + * http://www.unicode.org/unicode/reports/tr15/tr15-23.html#Annex8 + */ + + if (Normalizer.quickCheck(src, 0, length, mode, 0) == Normalizer.YES) + return true; + + /* + * There are known maximum expansions for the different forms that could + * remove the loop/exception handling below. These may change in new + * versions of the Unicode standard, and are sometimes large. The loop is + * for simplicity and ease of maintenance; with a large default buffer + * size it should rarely execute more than once. + * + * From http://unicode.org/reports/tr36/tr36-6.html#Buffer_Overflows: The + * very large factors in the case of NFKC/D are due to some extremely rare + * characters. Thus algorithms can use much smaller expansion factors for + * the typical cases as long as they have a fallback process that accounts + * for the possibility of these characters in data. + * + * For example, under normalization forms NFKC or NFKD, ﷺ (FDFA, ARABIC + * LIGATURE SALLALLAHOU ALAYHE WASALLAM) will be expanded to صلى الله عليه + * وسلم + */ + + do { + try { + + /* + * This method is documented in the public API to throw + * IndexOutOfBoundsException if there is not enough space. Its an + * unfortunate mechanism, it would be a lot nicer if instead it + * behaved like the ArabicShaping API, whereas instead it would return + * the necessary length, possibly more than the buffer supplied. This + * would simplify things, instead a call with a 0-length output buffer + * would return the necessary length. + */ + + final int newLength = Normalizer.normalize(src, 0, length, buffer, 0, + buffer.length, mode, 0); + termAtt.setTermBuffer(buffer, 0, newLength); + return true; + } catch (IndexOutOfBoundsException e) { + // technically, ICU encodes the necessary size as a String in the + // exception, but don't depend on that... + buffer = new char[ArrayUtil.getNextSize(buffer.length << 1)]; + } + } while (true); + } else { + return false; + } + } +} Property changes on: contrib\icu\src\java\org\apache\lucene\icu\ICUNormalizationFilter.java ___________________________________________________________________ Added: svn:eol-style + native Index: contrib/icu/src/java/org/apache/lucene/icu/ICUTransformFilter.java =================================================================== --- contrib/icu/src/java/org/apache/lucene/icu/ICUTransformFilter.java (revision 0) +++ contrib/icu/src/java/org/apache/lucene/icu/ICUTransformFilter.java (revision 0) @@ -0,0 +1,235 @@ +package org.apache.lucene.icu; + +/** + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +import java.io.IOException; + +import org.apache.lucene.analysis.TokenFilter; +import org.apache.lucene.analysis.TokenStream; +import org.apache.lucene.analysis.tokenattributes.TermAttribute; + +import org.apache.lucene.icu.ICUNormalizationFilter; // only used for javadoc. +import org.apache.lucene.util.ArrayUtil; + +import com.ibm.icu.text.Replaceable; +import com.ibm.icu.text.RuleBasedTransliterator; // only used for optimizing the transform, see below +import com.ibm.icu.text.Transliterator; +import com.ibm.icu.text.UTF16; +import com.ibm.icu.text.UnicodeSet; + +/** + * A {@link TokenFilter} that transforms text with ICU. + *+ * ICU provides text-transformation functionality via its Transliteration API. + * Although script conversion is its most common use, a transliterator can + * actually perform a more general class of tasks. In fact, Transliterator + * defines a very general API which specifies only that a segment of the input + * text is replaced by new text. The particulars of this conversion are + * determined entirely by subclasses of Transliterator. + *
+ *+ * Some useful transformations for search are built-in: + *
+ * Example usage:
stream = new ICUTransformFilter(stream, + * Transliterator.getInstance("Traditional-Simplified"));+ * + *
+ * Whether or not this filter respects equivalence or preserves normalization + * forms depends entirely upon the ruleset being applied. + *
+ *+ * For good performance, it is helpful to declare a filter in any custom + * transform you build. This allows the transform to efficiently skip over + * unaffected text. It is also useful to consider if there are simpler + * solutions. For example, if you want to standardize Fullwidth and Halfwidth + * forms, use of {@link ICUNormalizationFilter} with compatibility decomposition + * will erase width differences, with better performance. + *
+ * For more details, see the ICU User + * Guide. + */ + +public final class ICUTransformFilter extends TokenFilter { + // Transliterator to transform the text + private final Transliterator transform; + + // Reusable position object + private final Transliterator.Position position = new Transliterator.Position(); + + // Wraps a termAttribute around the replaceable interface. + private final ReplaceableTermAttribute replaceableAttribute = new ReplaceableTermAttribute(); + + // new api term attribute, will be updated with transformed text. + private TermAttribute termAtt; + + /** + * Create a new ICUTransformFilter that transforms text on the given stream. + * + * @param input {@link TokenStream} to filter. + * @param transform Transliterator to transform the text. + */ + public ICUTransformFilter(TokenStream input, Transliterator transform) { + super(input); + this.transform = transform; + termAtt = (TermAttribute) addAttribute(TermAttribute.class); + + /* + * A good UnicodeFilter is vital for performance. Unfortunately, sometimes + * people omit filters in their rulesets. However, in the special case that + * the transform is a RuleBasedTransliterator, this situation can be + * corrected. It can only be applied to a pure RuleBasedTransliterator, and + * it is only applied when there is no supplied filter. + * + * For a great example of a ruleset like this, see the built-in + * Simplified/Traditional ruleset from CLDR. This is a massive performance + * optimization for that case! + * + * If CompoundTransliterator and its children were exposed (its + * package-private and children are inaccessible), then more cases could be + * optimized. + * + * Regardless of who wrote the rules, you can ALWAYS apply a filter own your + * own:Transliterator.getInstance( + * "[:Arabic:] UnfilteredTransformThatOnlyProcessesTheArabicBlock"); + * + * Just be careful to ensure you don't filter characters that should be + * converted! This can be tricky if, for example, the transliterator + * internally invokes ::NFKC(). + */ + + if (transform.getFilter() == null + && (transform instanceof RuleBasedTransliterator)) { + final UnicodeSet sourceSet = transform.getSourceSet(); + if (sourceSet != null && !sourceSet.isEmpty()) + transform.setFilter(sourceSet); + } + } + + public boolean incrementToken() throws IOException { + + /* + * Wrap the TermAttribute around the replaceable interface, clear the + * positions, and transliterate. Finally, update the TermAttribute with the + * [potentially different] length. + */ + + if (input.incrementToken()) { + final int length = termAtt.termLength(); + replaceableAttribute.setText(termAtt); + + position.start = 0; + position.limit = length; + position.contextStart = 0; + position.contextLimit = length; + + transform.filteredTransliterate(replaceableAttribute, position, false); + termAtt.setTermLength(replaceableAttribute.length()); + return true; + } else { + return false; + } + } + + /** + * Wrap a {@link TermAttribute} with the Replaceable API. + * + * This allows for ICU transforms to run without unnecessary object creation. + * + * This wrapper does not keep the TermAttribute's length up to date at all + * times, when you are done you must finalize the replacement process by setting + * the TermAttribute's length. + */ + + final class ReplaceableTermAttribute implements Replaceable { + private char buffer[]; + + private int length; + + private TermAttribute token; + + ReplaceableTermAttribute() { + } + + void setText(final TermAttribute reusableToken) { + this.token = reusableToken; + this.buffer = reusableToken.termBuffer(); + this.length = reusableToken.termLength(); + } + + public int char32At(int pos) { + return UTF16.charAt(buffer, 0, length, pos); + } + + public char charAt(int pos) { + return buffer[pos]; + } + + public void copy(int start, int limit, int dest) { + char text[] = new char[limit - start]; + getChars(start, limit, text, 0); + replace(dest, dest, text, 0, limit - start); + } + + public void getChars(int srcStart, int srcLimit, char[] dst, int dstStart) { + System.arraycopy(buffer, srcStart, dst, dstStart, srcLimit - srcStart); + } + + public boolean hasMetaData() { + return false; + } + + public int length() { + return length; + } + + public void replace(int start, int limit, String text) { + replace(start, limit, text.toCharArray(), 0, text.length()); + } + + public void replace(int start, int limit, char[] text, int charsStart, + int charsLen) { + final int replacementLength = limit - start; + final int newLength = length - replacementLength + charsLen; + // resize if necessary + if (newLength > length) + buffer = token.resizeTermBuffer(ArrayUtil.getNextSize(newLength)); + // if the substring being replaced is longer or shorter than the + // replacement, need to shift things around + if (replacementLength != charsLen && limit < length) + System.arraycopy(buffer, limit, buffer, start + charsLen, length - limit); + // insert the replacement text + System.arraycopy(text, charsStart, buffer, start, charsLen); + length = newLength; + } + } +} Property changes on: contrib\icu\src\java\org\apache\lucene\icu\ICUTransformFilter.java ___________________________________________________________________ Added: svn:eol-style + native Index: contrib/icu/src/java/org/apache/lucene/icu/package.html =================================================================== --- contrib/icu/src/java/org/apache/lucene/icu/package.html (revision 0) +++ contrib/icu/src/java/org/apache/lucene/icu/package.html (revision 0) @@ -0,0 +1,22 @@ + + + + +An analyzer and set of analysis components for Unicode text, based on ICU. + + \ No newline at end of file Property changes on: contrib\icu\src\java\org\apache\lucene\icu\package.html ___________________________________________________________________ Added: svn:eol-style + native Index: contrib/icu/src/java/org/apache/lucene/icu/tokenizer/BreakIteratorWrapper.java =================================================================== --- contrib/icu/src/java/org/apache/lucene/icu/tokenizer/BreakIteratorWrapper.java (revision 0) +++ contrib/icu/src/java/org/apache/lucene/icu/tokenizer/BreakIteratorWrapper.java (revision 0) @@ -0,0 +1,181 @@ +package org.apache.lucene.icu.tokenizer; + +/** + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +import java.text.CharacterIterator; + +import com.ibm.icu.lang.UCharacter; +import com.ibm.icu.text.BreakIterator; +import com.ibm.icu.text.DictionaryBasedBreakIterator; +import com.ibm.icu.text.RuleBasedBreakIterator; +import com.ibm.icu.text.UTF16; + +/** + * Contain all the issues surrounding BreakIterators in ICU in one place. + * Basically this boils down to the fact that they aren't very friendly to any + * sort of OO design. + * + * http://bugs.icu-project.org/trac/ticket/5901: RBBI.getRuleStatus(), hoist to + * BreakIterator from RuleBasedBreakIterator + * + * DictionaryBasedBreakIterator is a subclass of RuleBasedBreakIterator, but + * doesn't actually behave as a subclass: it always returns 0 for + * getRuleStatus() + * + * http://bugs.icu-project.org/trac/ticket/4730: Thai RBBI, no boundary type + * tags + * + */ +abstract class BreakIteratorWrapper { + protected final CharArrayIterator textIterator = new CharArrayIterator(); + + protected char text[]; + + protected int start; + + protected int length; + + abstract int next(); + + abstract int current(); + + abstract int getRuleStatus(); + + abstract void setText(CharacterIterator text); + + void setText(char text[], int start, int length) { + this.text = text; + this.start = start; + this.length = length; + textIterator.setText(text, start, length); + setText(textIterator); + } + + /* + * If its a DictionaryBasedBreakIterator, it doesn't return rulestatus, so + * treat it like a generic BreakIterator If its any other + * RuleBasedBreakIterator, the rule status can be used for token type. If its + * any other BreakIterator, the rulestatus method is not available, so treat + * it like a generic BreakIterator. + */ + static BreakIteratorWrapper wrap(BreakIterator breakIterator) { + if (breakIterator instanceof RuleBasedBreakIterator + && !(breakIterator instanceof DictionaryBasedBreakIterator)) + return new RBBIWrapper((RuleBasedBreakIterator) breakIterator); + else + return new BIWrapper(breakIterator); + } + + /* + * RuleBasedBreakIterator wrapper: RuleBasedBreakIterator (as long as its not + * a DictionaryBasedBreakIterator) behaves correctly. + */ + static final class RBBIWrapper extends BreakIteratorWrapper { + private final RuleBasedBreakIterator rbbi; + + RBBIWrapper(RuleBasedBreakIterator rbbi) { + this.rbbi = rbbi; + } + + int current() { + return rbbi.current(); + } + + int getRuleStatus() { + return rbbi.getRuleStatus(); + } + + int next() { + return rbbi.next(); + } + + void setText(CharacterIterator text) { + rbbi.setText(text); + } + } + + /* + * Generic BreakIterator wrapper: Either the rulestatus method is not + * available or always returns 0. Calculate a rulestatus here so it behaves + * like RuleBasedBreakIterator. + * + * Note: This is slower than RuleBasedBreakIterator. + */ + static final class BIWrapper extends BreakIteratorWrapper { + private final BreakIterator bi; + + private int status; + + BIWrapper(BreakIterator bi) { + this.bi = bi; + } + + int current() { + return bi.current(); + } + + int getRuleStatus() { + return status; + } + + int next() { + int current = bi.current(); + int next = bi.next(); + status = calcStatus(current, next); + return next; + } + + private int calcStatus(int current, int next) { + if (current == BreakIterator.DONE || next == BreakIterator.DONE) + return RuleBasedBreakIterator.WORD_NONE; + + int begin = start + current; + int end = start + next; + + /* + * TODO: Consider optimizing BMP with this idiom: + * http://www.icu-project.org/docs/papers/supplementaries_iuc21.ppt + * + * for (int i = 0; i < s.length(); ++i) { int c = s.charAt(i); if (0xD800 + * <= c && c <= 0xDBFF) { c = UTF16.charAt(s, i); i += + * UTF16.getCharCount(c) - 1; } if (UCharacter.isLetter(c)) { + * doSomething(c); } } + */ + + int codepoint; + for (int i = begin; i < end; i += UTF16.getCharCount(codepoint)) { + codepoint = UTF16.charAt(text, 0, end, begin); + + if (UCharacter.isDigit(codepoint)) + return RuleBasedBreakIterator.WORD_NUMBER; + else if (UCharacter.isLetter(codepoint)) { + // TODO try to separately specify ideographic, kana? [currently all + // bundled as letter for this case] + return RuleBasedBreakIterator.WORD_LETTER; + } + } + + return RuleBasedBreakIterator.WORD_NONE; + } + + void setText(CharacterIterator text) { + bi.setText(text); + status = RuleBasedBreakIterator.WORD_NONE; + } + } +} Property changes on: contrib\icu\src\java\org\apache\lucene\icu\tokenizer\BreakIteratorWrapper.java ___________________________________________________________________ Added: svn:eol-style + native Index: contrib/icu/src/java/org/apache/lucene/icu/tokenizer/CharArrayIterator.java =================================================================== --- contrib/icu/src/java/org/apache/lucene/icu/tokenizer/CharArrayIterator.java (revision 0) +++ contrib/icu/src/java/org/apache/lucene/icu/tokenizer/CharArrayIterator.java (revision 0) @@ -0,0 +1,117 @@ +package org.apache.lucene.icu.tokenizer; + +/** + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +import java.text.CharacterIterator; + +/** + * Wraps a char[] as CharacterIterator for processing with a BreakIterator + * + */ +final class CharArrayIterator implements CharacterIterator { + private char array[]; + private int start; + private int index; + private int length; + private int limit; + + public char [] getText() { + return array; + } + + public int getStart() { + return start; + } + + public int getLength() { + return length; + } + + /** + * Set a new region of text to be examined by this iterator + * + * @param array text buffer to examine + * @param start offset into buffer + * @param length maximum length to examine + */ + void setText(final char array[], int start, int length) { + this.array = array; + this.start = start; + this.index = start; + this.length = length; + this.limit = start + length; + } + + public char current() { + return (index == limit) ? DONE : array[index]; + } + + public char first() { + index = start; + return current(); + } + + public int getBeginIndex() { + return 0; + } + + public int getEndIndex() { + return length; + } + + public int getIndex() { + return index - start; + } + + public char last() { + index = (limit == start) ? limit : limit - 1; + return current(); + } + + public char next() { + if (++index >= limit) { + index = limit; + return DONE; + } else { + return current(); + } + } + + public char previous() { + if (--index < start) { + index = start; + return DONE; + } else { + return current(); + } + } + + public char setIndex(int position) { + if (position < getBeginIndex() || position > getEndIndex()) + throw new IllegalArgumentException("Illegal Position: " + position); + index = start + position; + return current(); + } + + public Object clone() { + CharArrayIterator clone = new CharArrayIterator(); + clone.setText(array, start, length); + clone.index = index; + return clone; + } +} Property changes on: contrib\icu\src\java\org\apache\lucene\icu\tokenizer\CharArrayIterator.java ___________________________________________________________________ Added: svn:eol-style + native Index: contrib/icu/src/java/org/apache/lucene/icu/tokenizer/CompositeBreakIterator.java =================================================================== --- contrib/icu/src/java/org/apache/lucene/icu/tokenizer/CompositeBreakIterator.java (revision 0) +++ contrib/icu/src/java/org/apache/lucene/icu/tokenizer/CompositeBreakIterator.java (revision 0) @@ -0,0 +1,144 @@ +package org.apache.lucene.icu.tokenizer; + +/** + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +import java.util.Iterator; +import java.util.Map; + +import com.ibm.icu.lang.UScript; +import com.ibm.icu.text.BreakIterator; + +/** + * An internal BreakIterator for multilingual text, following recommendations + * from: UAX #29: Unicode Text Segmentation. (http://unicode.org/reports/tr29/) + * + * This Break Iterator can be customized in up to two ways: 1. Providing a + * custom default BreakIterator. 2. Providing a mapping of BreakIterators that + * work with specific scripts. + * + * See http://unicode.org/reports/tr29/#Tailoring for the motivation of this + * design + * + * Text is first divided into script boundaries. The processing is then + * delegated to the appropriate break iterator for that specific script. If no + * script-specific break iterator is available, the Unicode default bounds + * properties are used. + * + * This break iterator also allows you to retrieve the ISO 15924 script code + * associated with a piece of text. + * + * See also UAX #29, UTR #24 + * + */ +final class CompositeBreakIterator { + private final BreakIteratorWrapper wordBreakers[] = new BreakIteratorWrapper[UScript.CODE_LIMIT]; + + private BreakIteratorWrapper rbbi; + + private final ScriptIterator scriptIterator = new ScriptIterator(); + + private char text[]; + + CompositeBreakIterator(BreakIterator rbbi, Map scriptHandlers) { + // the default word breaker + BreakIteratorWrapper defaultBreaker = BreakIteratorWrapper.wrap(rbbi); + for (int i = 0; i < wordBreakers.length; i++) + wordBreakers[i] = defaultBreaker; + + // register any custom script-specific word-breakers + for (Iterator iterator = scriptHandlers.entrySet().iterator(); iterator + .hasNext();) { + Map.Entry handler = (Map.Entry) iterator.next(); + Integer scriptCode = (Integer) handler.getKey(); + BreakIterator scriptHandler = (BreakIterator) handler.getValue(); + wordBreakers[scriptCode.intValue()] = BreakIteratorWrapper + .wrap(scriptHandler); + } + } + + /** + * Retrieve the next break position. If the RBBI range is exhausted within the + * script boundary, examine the next script boundary. + * + * @return the next break position or BreakIterator.DONE + */ + int next() { + int next = rbbi.next(); + while (next == BreakIterator.DONE && scriptIterator.next()) { + rbbi = wordBreakers[scriptIterator.getScriptCode()]; + rbbi.setText(text, scriptIterator.getScriptStart(), scriptIterator + .getScriptLimit() + - scriptIterator.getScriptStart()); + next = rbbi.next(); + } + return (next == BreakIterator.DONE) ? BreakIterator.DONE : next + + scriptIterator.getScriptStart(); + } + + /** + * Retrieve the current break position. + * + * @return the current break position or BreakIterator.DONE + */ + int current() { + final int current = rbbi.current(); + return (current == BreakIterator.DONE) ? BreakIterator.DONE : current + + scriptIterator.getScriptStart(); + } + + /** + * Retrieve the rule status code (token type) from the underlying break + * iterator + * + * @return rule status code (see RuleBasedBreakIterator constants) + */ + int getRuleStatus() { + return rbbi.getRuleStatus(); + } + + /** + * Retrieve the UScript script code for the current token. This code can be + * decoded with UScript into a name or ISO 15924 code. + * + * @return UScript script code for the current token. + */ + int getScriptCode() { + return scriptIterator.getScriptCode(); + } + + /** + * Set a new region of text to be examined by this iterator + * + * @param text buffer of text + * @param start offset into buffer + * @param length maximum length to examine + */ + void setText(final char text[], int start, int length) { + this.text = text; + scriptIterator.setText(text, start, length); + if (scriptIterator.next()) { + rbbi = wordBreakers[scriptIterator.getScriptCode()]; + rbbi.setText(text, scriptIterator.getScriptStart(), scriptIterator + .getScriptLimit() + - scriptIterator.getScriptStart()); + } else { + rbbi = wordBreakers[UScript.COMMON]; + rbbi.setText(text, 0, 0); + } + } +} Property changes on: contrib\icu\src\java\org\apache\lucene\icu\tokenizer\CompositeBreakIterator.java ___________________________________________________________________ Added: svn:eol-style + native Index: contrib/icu/src/java/org/apache/lucene/icu/tokenizer/ICUTokenizer.java =================================================================== --- contrib/icu/src/java/org/apache/lucene/icu/tokenizer/ICUTokenizer.java (revision 0) +++ contrib/icu/src/java/org/apache/lucene/icu/tokenizer/ICUTokenizer.java (revision 0) @@ -0,0 +1,331 @@ +package org.apache.lucene.icu.tokenizer; + +/** + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +import java.io.IOException; +import java.io.InputStream; +import java.io.Reader; +import java.util.HashMap; +import java.util.Map; + +import org.apache.lucene.analysis.Tokenizer; +import org.apache.lucene.analysis.tokenattributes.FlagsAttribute; +import org.apache.lucene.analysis.tokenattributes.OffsetAttribute; +import org.apache.lucene.analysis.tokenattributes.PositionIncrementAttribute; +import org.apache.lucene.analysis.tokenattributes.TermAttribute; +import org.apache.lucene.analysis.tokenattributes.TypeAttribute; + +import com.ibm.icu.lang.UCharacter; +import com.ibm.icu.lang.UScript; +import com.ibm.icu.text.BreakIterator; +import com.ibm.icu.text.RuleBasedBreakIterator; +import com.ibm.icu.util.ULocale; + +/** + * Breaks text into words according to UAX #29: Unicode Text Segmentation + * (http://www.unicode.org/reports/tr29/) + *+ * Words are broken across script boundaries and unicode boundaries based upon + * their unicode properties. + *
+ * TODO: more docs + */ +public final class ICUTokenizer extends Tokenizer { + /** + * Token type for words containing ideographic characters + */ + public static final String WORD_IDEO = "IDEO"; + + /** + * Token type for words containing kana characters + */ + public static final String WORD_KANA = "KANA"; + + /** + * Token type for words that contain letters, excluding hiragana, katakana, or + * ideographic characters. + */ + public static final String WORD_LETTER = "WORD"; + + /** + * Token type for words that appear to be numbers. + */ + public static final String WORD_NUMBER = "NUM"; + + private static final int IOBUFFER = 4096; + + private final char buffer[] = new char[IOBUFFER]; + + private int length = 0; /* true length of text in the buffer */ + + private int usableLength = 0; /* + * length of text in the buffer that can be + * evaluated safely, up to a safe end point + */ + + private int offset = 0; /* + * accumulated offset of previous buffers for this + * Reader, for correct term offsets + */ + + private final CompositeBreakIterator breaker; /* tokenizes a char[] of text */ + + private OffsetAttribute offsetAtt; + private TermAttribute termAtt; + private TypeAttribute typeAtt; + private FlagsAttribute flagsAtt; + private PositionIncrementAttribute posIncAtt; + + /** + * Construct a new ICUTokenizer that breaks text into words from the given + * Reader. + *+ * Text will be broken into words by the Unicode text segmentation algorithm: + * BreakIterator.getWordInstance(ULocale.ROOT) + *
+ *+ * The default script-specific handling is used: specifically Thai text is + * broken into words with a Thai DictionaryBasedBreakiterator + *
+ * + * @param input Reader containing text to tokenize. + */ + public ICUTokenizer(Reader input) { + this(input, null, null); + } + + /** + * Construct a new ICUTokenizer that breaks text into words from the given + * Reader, using a tailored default BreakIterator + *+ * The default script-specific handling is used: specifically Thai text is + * broken into words with a Thai DictionaryBasedBreakiterator + *
+ * + * @param input Reader containing text to tokenize. + * @param defaultBreakIterator Tailored BreakIterator that breaks text into + * words. + */ + public ICUTokenizer(Reader input, BreakIterator defaultBreakIterator) { + this(input, defaultBreakIterator, null); + } + + /** + * Construct a new ICUTokenizer that breaks text into words from the given + * Reader, using a tailored default BreakIterator and a custom mapping of + * script-specific BreakIterators + * + * @param input Reader containing text to tokenize. + * @param defaultBreakIterator Tailored BreakIterator that breaks text into + * words. + * @param scriptHandlers A {@link Map} mapping UScript codes to BreakIterators + * for script-specific handling. + */ + public ICUTokenizer(Reader input, BreakIterator defaultBreakIterator, + Map scriptHandlers) { + super(input); + + if (defaultBreakIterator == null) + defaultBreakIterator = (BreakIterator) rootBreakIterator.clone(); + + if (scriptHandlers == null) { + scriptHandlers = new HashMap(); + /* Hebrew has special punctuation handling */ + scriptHandlers.put(new Integer(UScript.HEBREW), + (BreakIterator) hebrewBreakIterator.clone()); + /* Thai has no visible word separation, use dictionary-based break iterator */ + scriptHandlers.put(new Integer(UScript.THAI), + (BreakIterator) thaiBreakIterator.clone()); + /* Khmer has no visible word separation, break into syllables */ + scriptHandlers.put(new Integer(UScript.KHMER), + (BreakIterator) khmerBreakIterator.clone()); + /* Lao has no visible word separation, break into syllables */ + scriptHandlers.put(new Integer(UScript.LAO), + (BreakIterator) laoBreakIterator.clone()); + /* Myanmar has no visible word separation, break into syllables */ + scriptHandlers.put(new Integer(UScript.MYANMAR), + (BreakIterator) myanmarBreakIterator.clone()); + } + + breaker = new CompositeBreakIterator(defaultBreakIterator, scriptHandlers); + offsetAtt = (OffsetAttribute) addAttribute(OffsetAttribute.class); + termAtt = (TermAttribute) addAttribute(TermAttribute.class); + typeAtt = (TypeAttribute) addAttribute(TypeAttribute.class); + flagsAtt = (FlagsAttribute) addAttribute(FlagsAttribute.class); + posIncAtt = (PositionIncrementAttribute) addAttribute(PositionIncrementAttribute.class); + } + + /** + * Return the token's lexical type. + *+ * For a tailored ruleset, you may want to override this method to return + * additional types. + *
+ * + * @return String containing the token type + */ + protected String getType() { + switch (breaker.getRuleStatus()) { + case RuleBasedBreakIterator.WORD_IDEO: + return ICUTokenizer.WORD_IDEO; + case RuleBasedBreakIterator.WORD_KANA: + return ICUTokenizer.WORD_KANA; + case RuleBasedBreakIterator.WORD_LETTER: + return ICUTokenizer.WORD_LETTER; + case RuleBasedBreakIterator.WORD_NUMBER: + return ICUTokenizer.WORD_NUMBER; + default: /* RuleBasedBreakIterator.WORD_NONE, not possible */ + return "NONE"; + } + } + + public boolean incrementToken() throws IOException { + clearAttributes(); + if (length == 0) + refill(); + while (!incrementTokenBuffer()) { + refill(); + if (length <= 0) // no more bytes to read; + return false; + } + return true; + } + + public void reset(Reader input) throws IOException { + super.reset(input); + breaker.setText(buffer, 0, 0); + length = usableLength = offset = 0; + } + + /* + * The method is as follows: + * + * The ICU RBBI implementation for the ROOT locale is used. This tokenizes + * text based upon the longest matching rule, and because of this, isn't + * friendly to a Reader. + * + * Text is read from the input stream in 4kB chunks. Within a 4kB chunk of + * text, the last unambiguous break point is found (in this implementation: + * white space character) Any remaining characters represent possible partial + * words, so are appended to the front of the next chunk. + * + * There is the possibility that there are no unambiguous break points within + * an entire 4kB chunk of text (binary data) Currently, this could cause what + * should really be an 8kB word to be parsed as two 4kB words instead. This + * keeps the code simple, but other possibilities could be: A. increasing the + * buffer size and reading more data, hoping to find an unambiguous boundary, + * at the risk of OOM error. B. discarding the chunk entirely and instead + * looking for a set of unambiguous start-end boundaries in future text. + * + * Option B is closer to what other Lucene analyzers do, but would increase + * code complexity significantly. + */ + + /** + * Returns the last unambiguous break position in the text. + * + * @return position of character, or -1 if one does not exist + */ + private int findSafeEnd() { + for (int i = length - 1; i >= 0; i--) + if (UCharacter.isWhitespace(buffer[i])) + return i + 1; + return -1; + } + + /** + * Refill the buffer, accumulating the offset and setting usableLength to the + * last unambiguous break position + * + * @throws IOException + */ + private void refill() throws IOException { + offset += usableLength; + int leftover = length - usableLength; + System.arraycopy(buffer, usableLength, buffer, 0, leftover); + int requested = buffer.length - leftover; + int returned = input.read(buffer, leftover, requested); + length = returned < 0 ? leftover : returned + leftover; + if (returned < requested) /* reader has been emptied, process the rest */ + usableLength = length; + else { /* still more data to be read, find a safe-stopping place */ + usableLength = findSafeEnd(); + if (usableLength < 0) + usableLength = length; /* + * more than IOBUFFER of text without space, + * gonna possibly truncate tokens + */ + } + + breaker.setText(buffer, 0, Math.max(0, usableLength)); + } + + /* + * return true if there is a token from the buffer, or null if it is + * exhausted. + */ + private boolean incrementTokenBuffer() { + int start = breaker.current(); + if (start == BreakIterator.DONE) + return false; // BreakIterator exhausted + + // find the next set of boundaries, skipping over non-tokens (rule status 0) + int end = breaker.next(); + while (start != BreakIterator.DONE && breaker.getRuleStatus() == 0) { + start = end; + end = breaker.next(); + } + + if (start == BreakIterator.DONE) + return false; // BreakIterator exhausted + + int length = end - start; + termAtt.setTermBuffer(buffer, start, length); + offsetAtt.setOffset(correctOffset(offset + start), correctOffset(offset + end)); + typeAtt.setType(getType()); + flagsAtt.setFlags(breaker.getScriptCode()); + + return true; + } + + /* + * the default breakiterators in use. these can be expensive to + * instantiate, cheap to clone. + */ + private static final BreakIterator rootBreakIterator = BreakIterator + .getWordInstance(ULocale.ROOT); + + private static final BreakIterator thaiBreakIterator = BreakIterator + .getWordInstance(new ULocale("th_TH")); + + private static final BreakIterator hebrewBreakIterator = readBreakIterator("Hebrew.brk"); + private static final BreakIterator khmerBreakIterator = readBreakIterator("Khmer.brk"); + private static final BreakIterator laoBreakIterator = new LaoBreakIterator(readBreakIterator("Lao.brk")); + private static final BreakIterator myanmarBreakIterator = readBreakIterator("Myanmar.brk"); + + private static RuleBasedBreakIterator readBreakIterator(String filename) { + InputStream is = ICUTokenizer.class.getResourceAsStream(filename); + try { + RuleBasedBreakIterator bi = RuleBasedBreakIterator.getInstanceFromCompiledRules(is); + is.close(); + return bi; + } catch (IOException e) { + throw new RuntimeException(e); + } + } +} Property changes on: contrib\icu\src\java\org\apache\lucene\icu\tokenizer\ICUTokenizer.java ___________________________________________________________________ Added: svn:eol-style + native Index: contrib/icu/src/java/org/apache/lucene/icu/tokenizer/LaoBreakIterator.java =================================================================== --- contrib/icu/src/java/org/apache/lucene/icu/tokenizer/LaoBreakIterator.java (revision 0) +++ contrib/icu/src/java/org/apache/lucene/icu/tokenizer/LaoBreakIterator.java (revision 0) @@ -0,0 +1,167 @@ +package org.apache.lucene.icu.tokenizer; + +/** + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +import java.text.CharacterIterator; + +import com.ibm.icu.lang.UCharacter; +import com.ibm.icu.text.BreakIterator; +import com.ibm.icu.text.RuleBasedBreakIterator; +import com.ibm.icu.text.UnicodeSet; + +public class LaoBreakIterator extends BreakIterator { + RuleBasedBreakIterator rules; + CharArrayIterator text; + + CharArrayIterator working = new CharArrayIterator(); + int workingOffset = 0; + + CharArrayIterator verifyText = new CharArrayIterator(); + RuleBasedBreakIterator verify; + + private static final UnicodeSet laoSet; + static { + laoSet = new UnicodeSet("[:Lao:]"); + laoSet.compact(); + laoSet.freeze(); + } + + public LaoBreakIterator(RuleBasedBreakIterator rules) { + this.rules = rules; + this.verify = (RuleBasedBreakIterator) rules.clone(); + } + + public int current() { + int current = rules.current(); + return current == BreakIterator.DONE ? BreakIterator.DONE : workingOffset + current; + } + + public int first() { + working.setText(this.text.getText(), this.text.getStart(), this.text.getLength()); + rules.setText(working); + workingOffset = 0; + int first = rules.first(); + return first == BreakIterator.DONE ? BreakIterator.DONE : workingOffset + first; + } + + public int following(int offset) { + // TODO Auto-generated method stub + return 0; + } + + public CharacterIterator getText() { + return text; + } + + public int last() { + // TODO Auto-generated method stub + return 0; + } + + public int next() { + int current = current(); + int next = rules.next(); + if (next == BreakIterator.DONE) + return next; + else + next += workingOffset; + + char c = working.current(); + int following = rules.next(); // lookahead + if (following != BreakIterator.DONE) { + following += workingOffset; + if (rules.getRuleStatus() == 0 && laoSet.contains(c) && verifyPushBack(current, next)) { + workingOffset = next - 1; + working.setText(text.getText(), text.getStart() + workingOffset, text.getLength() - workingOffset); + return next - 1; + } + rules.previous(); // undo the lookahead + } + + return next; + } + + public int next(int n) { + if (n < 0) + throw new UnsupportedOperationException("Backwards traversal is unsupported"); + + int result = current(); + while (n > 0) { + result = next(); + --n; + } + return result; + } + + public int previous() { + throw new UnsupportedOperationException("Backwards traversal is unsupported"); + } + + public void setText(CharacterIterator text) { + if (!(text instanceof CharArrayIterator)) + throw new UnsupportedOperationException("unsupported CharacterIterator"); + this.text = (CharArrayIterator) text; + ccReorder(this.text.getText(), this.text.getStart(), this.text.getLength()); + working.setText(this.text.getText(), this.text.getStart(), this.text.getLength()); + rules.setText(working); + workingOffset = 0; + } + + public void setText(String newText) { + CharArrayIterator ci = new CharArrayIterator(); + ci.setText(newText.toCharArray(), 0, newText.length()); + setText(ci); + } + + private boolean verifyPushBack(int current, int next) { + int shortenedSyllable = next - current - 1; + + verifyText.setText(text.getText(), text.getStart() + current, shortenedSyllable); + verify.setText(verifyText); + if (verify.next() != shortenedSyllable || verify.getRuleStatus() == 0) + return false; + + + verifyText.setText(text.getText(), text.getStart() + next - 1, text.getLength() - next + 1); + verify.setText(verifyText); + + return (verify.next() != BreakIterator.DONE && verify.getRuleStatus() != 0); + } + + // TODO: only bubblesort around runs of combining marks, instead of the entire text. + private void ccReorder(char[] text, int start, int length) { + int prevCC = 0; + boolean reordered; + do { + reordered = false; + for (int i = start; i < start + length; i++) { + final char c = text[i]; + final int cc = UCharacter.getCombiningClass(c); + if (cc > 0 && cc < prevCC) { + // swap + text[i] = text[i - 1]; + text[i - 1] = c; + reordered = true; + } else { + prevCC = cc; + } + } + + } while (reordered == true); + } +} Property changes on: contrib\icu\src\java\org\apache\lucene\icu\tokenizer\LaoBreakIterator.java ___________________________________________________________________ Added: svn:eol-style + native Index: contrib/icu/src/java/org/apache/lucene/icu/tokenizer/ScriptIterator.java =================================================================== --- contrib/icu/src/java/org/apache/lucene/icu/tokenizer/ScriptIterator.java (revision 0) +++ contrib/icu/src/java/org/apache/lucene/icu/tokenizer/ScriptIterator.java (revision 0) @@ -0,0 +1,183 @@ +package org.apache.lucene.icu.tokenizer; + +/** + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +import com.ibm.icu.lang.UCharacter; +import com.ibm.icu.lang.UScript; +import com.ibm.icu.text.UTF16; + +/** + * An iterator that locates ISO 15924 script boundaries in text. This is not the + * same as simply looking at the Unicode block, or even the Script property. + * Some characters are 'common' across multiple scripts, and some 'inherit' the + * script value of text surrounding them. + * + * This is similar to ICU (internal-only) UScriptRun, with the following + * differences: + * + * Doesn't attempt to match paired punctuation. For tokenization purposes, this + * is not necessary. Its also quite expensive. Non-spacing marks inherit the + * script of their base character, following recommendations from UTR #24. + * + * TODO: Runs of Han ideographs and Hiragana/Katakana need to be mapped to + * Japanese script. TODO: Runs of Han ideographs and Hangul need to be mapped to + * Korean script. + */ +final class ScriptIterator { + private char text[]; + + private int start; + + private int limit; + + private int index; + + private int scriptStart; + + private int scriptLimit; + + private int scriptCode; + + /** + * Get the start of this script run + * + * @return start position of script run + */ + int getScriptStart() { + return scriptStart; + } + + /** + * Get the index of the first character after the end of this script run + * + * @return position of the first character after this script run + */ + int getScriptLimit() { + return scriptLimit; + } + + /** + * Get the UScript script code for this script run + * + * @return code for the script of the current run + */ + int getScriptCode() { + return scriptCode; + } + + /** + * Iterates to the next script run, returning true if one exists. + * + * @return true if there is another script run, false otherwise. + */ + boolean next() { + if (scriptLimit >= limit) + return false; + + scriptCode = UScript.COMMON; + scriptStart = scriptLimit; + + /* + * TODO: Consider optimizing BMP with this idiom: + * http://www.icu-project.org/docs/papers/supplementaries_iuc21.ppt + * + * for (int i = 0; i < s.length(); ++i) { int c = s.charAt(i); if (0xD800 <= + * c && c <= 0xDBFF) { c = UTF16.charAt(s, i); i += UTF16.getCharCount(c) - + * 1; } if (UCharacter.isLetter(c)) { doSomething(c); } } + */ + + while (index < limit) { + final int ch = UTF16.charAt(text, start, limit, index - start); + final int sc = getScript(ch); + + /* + * From UTR #24: Implementations that determine the boundaries between + * characters of given scripts should never break between a non-spacing + * mark and its base character. Thus for boundary determinations and + * similar sorts of processing, a non-spacing mark — whatever its script + * value — should inherit the script value of its base character. + */ + + if (isSameScript(scriptCode, sc) + || UCharacter.getType(ch) == UCharacter.NON_SPACING_MARK) { + index += UTF16.getCharCount(ch); + + /* + * Inherited or Common becomes the script code of the surrounding text. + */ + + if (scriptCode <= UScript.INHERITED && sc > UScript.INHERITED) { + scriptCode = sc; + } + + } else { + break; + } + } + + scriptLimit = index; + return true; + } + + /* + * Determine if two scripts are compatible. + */ + private static boolean isSameScript(int scriptOne, int scriptTwo) { + return scriptOne <= UScript.INHERITED || scriptTwo <= UScript.INHERITED + || scriptOne == scriptTwo; + } + + /** + * Set a new region of text to be examined by this iterator + * + * @param text text buffer to examine + * @param start offset into buffer + * @param length maximum length to examine + */ + void setText(char text[], int start, int length) { + this.text = text; + this.start = start; + this.index = start; + this.limit = start + length; + this.scriptStart = start; + this.scriptLimit = start; + this.scriptCode = UScript.INVALID_CODE; + } + + /* + * linear array access fast-path for basic latin case, greatest 128 ints ever + * spent + */ + private static final int basicLatin[] = new int[128]; + + static { + for (int i = 0; i < basicLatin.length; i++) + basicLatin[i] = UScript.getScript(i); + } + + /* + * An accelerated version of UScript.getScript() Basic latin is an array + * lookup. + */ + private static int getScript(int codepoint) { + if (0 <= codepoint && codepoint < basicLatin.length) + return basicLatin[codepoint]; + else + return UScript.getScript(codepoint); + } +} Property changes on: contrib\icu\src\java\org\apache\lucene\icu\tokenizer\ScriptIterator.java ___________________________________________________________________ Added: svn:eol-style + native Index: contrib/icu/src/java/org/apache/lucene/icu/tokenizer/package.html =================================================================== --- contrib/icu/src/java/org/apache/lucene/icu/tokenizer/package.html (revision 0) +++ contrib/icu/src/java/org/apache/lucene/icu/tokenizer/package.html (revision 0) @@ -0,0 +1,22 @@ + + + + +Tokenizer that breaks text into words with the Unicode Text Segmentation algorithm. + + \ No newline at end of file Property changes on: contrib\icu\src\java\org\apache\lucene\icu\tokenizer\package.html ___________________________________________________________________ Added: svn:eol-style + native Index: contrib/icu/src/resources/org/apache/lucene/icu/tokenizer/Hebrew.rbbi =================================================================== --- contrib/icu/src/resources/org/apache/lucene/icu/tokenizer/Hebrew.rbbi (revision 0) +++ contrib/icu/src/resources/org/apache/lucene/icu/tokenizer/Hebrew.rbbi (revision 0) @@ -0,0 +1,61 @@ +# +# Licensed to the Apache Software Foundation (ASF) under one or more +# contributor license agreements. See the NOTICE file distributed with +# this work for additional information regarding copyright ownership. +# The ASF licenses this file to You under the Apache License, Version 2.0 +# (the "License"); you may not use this file except in compliance with +# the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +# +# +# This is an example of rule tailoring for Hebrew. +# In this example the single-quote is added to the Extend category +# The double-quote is added to the MidLetter category. +# +!!chain; +$CR = [\p{Word_Break = CR}]; +$LF = [\p{Word_Break = LF}]; +$Newline = [\p{Word_Break = Newline}]; +$Extend = [\p{Word_Break = Extend}\u0027]; +$Format = [\p{Word_Break = Format}]; +$ALetter = [\p{Word_Break = ALetter}]; +$MidNumLet = [\p{Word_Break = MidNumLet}]; +$MidLetter = [\p{Word_Break = MidLetter}\u0022]; +$MidNum = [\p{Word_Break = MidNum}]; +$Numeric = [\p{Word_Break = Numeric}]; +$ExtendNumLet = [\p{Word_Break = ExtendNumLet}]; +$dictionary = [:LineBreak = Complex_Context:]; +$Control = [\p{Grapheme_Cluster_Break = Control}]; +$ALetterPlus = [$ALetter [$dictionary-$Extend-$Control]]; + +$ALetterEx = $ALetterPlus ($Extend | $Format)*; +$MidNumLetEx = $MidNumLet ($Extend | $Format)*; +$MidLetterEx = $MidLetter ($Extend | $Format)*; +$MidNumEx = $MidNum ($Extend | $Format)*; +$NumericEx = $Numeric ($Extend | $Format)*; +$ExtendNumLetEx = $ExtendNumLet ($Extend | $Format)*; + +!!forward; + +$CR $LF; +[^$CR $LF $Newline]? ($Extend | $Format)+; +$NumericEx {100}; +$ALetterEx {200}; +$ALetterEx $ALetterEx {200}; +$ALetterEx ($MidLetterEx | $MidNumLetEx) $ALetterEx {200}; +$NumericEx $NumericEx {100}; +$ALetterEx $NumericEx {200}; +$NumericEx $ALetterEx {200}; +$NumericEx ($MidNumEx | $MidNumLetEx) $NumericEx {100}; +$ALetterEx $ExtendNumLetEx {200}; +$NumericEx $ExtendNumLetEx {100}; +$ExtendNumLetEx $ExtendNumLetEx {200}; +$ExtendNumLetEx $ALetterEx {200}; +$ExtendNumLetEx $NumericEx {100}; Index: contrib/icu/src/resources/org/apache/lucene/icu/tokenizer/Khmer.rbbi =================================================================== --- contrib/icu/src/resources/org/apache/lucene/icu/tokenizer/Khmer.rbbi (revision 0) +++ contrib/icu/src/resources/org/apache/lucene/icu/tokenizer/Khmer.rbbi (revision 0) @@ -0,0 +1,70 @@ +# +# Licensed to the Apache Software Foundation (ASF) under one or more +# contributor license agreements. See the NOTICE file distributed with +# this work for additional information regarding copyright ownership. +# The ASF licenses this file to You under the Apache License, Version 2.0 +# (the "License"); you may not use this file except in compliance with +# the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +# +# +# Parses Khmer text, with orthographic syllable as token. +# ICUBigramFilter will form bigrams of these syllables downstream. +# This is an example of how one can use a tailored ruleset. +# The definition of Khmer orthographic syllable is taken directly from the Unicode Standard. +# +!!chain; +# +# B = base character (consonant, independent vowel, etc) +$KhmerBase = [\u1780-\u17B3]; +# R = robat +$KhmerRobat = [\u17CC]; +# C = consonant shifter +$KhmerShifter = [\u17C9\u17CA]; +# S = subscript consonant or independent vowel sign +$KhmerSub = ([\u17D2] $KhmerBase); +# V = dependent vowel sign +$KhmerVowel = [\u17B4-\u17C5]; +# Z = zero-width joiner or non-joiner +$KhmerZWC = [\u200C\u200D]; +# O = any other sign +$KhmerSign = [\u17C6-\u17C8\u17CB\u17CD-\u17D1\u17DC\u17DD]; + +$WordJoin = [:Line_Break=Word_Joiner:]; + +$KhmerSyllableEx = $KhmerBase ($KhmerRobat | $KhmerShifter)? ($KhmerSub ($KhmerRobat)?)* (($KhmerZWC)? $KhmerVowel)? ($KhmerSign)? ($KhmerSub)?; + +$KhmerJoinedSyllableEx = $KhmerSyllableEx ($WordJoin $KhmerSyllableEx)*; + +# +# default numerical definitions +# +$Extend = [\p{Word_Break = Extend}]; +$Format = [\p{Word_Break = Format}]; +$MidNumLet = [\p{Word_Break = MidNumLet}]; +$MidNum = [\p{Word_Break = MidNum}]; +$Numeric = [\p{Word_Break = Numeric}]; +$ExtendNumLet = [\p{Word_Break = ExtendNumLet}]; +$MidNumLetEx = $MidNumLet ($Extend | $Format)*; +$MidNumEx = $MidNum ($Extend | $Format)*; +$NumericEx = $Numeric ($Extend | $Format)*; +$ExtendNumLetEx = $ExtendNumLet ($Extend | $Format)*; + +!!forward; +$KhmerJoinedSyllableEx {200}; + +# +# default numeric rules +# +$NumericEx {100}; +$NumericEx $NumericEx {100}; +$NumericEx ($MidNumEx | $MidNumLetEx) $NumericEx {100}; +$NumericEx $ExtendNumLetEx {100}; +$ExtendNumLetEx $NumericEx {100}; Index: contrib/icu/src/resources/org/apache/lucene/icu/tokenizer/Lao.rbbi =================================================================== --- contrib/icu/src/resources/org/apache/lucene/icu/tokenizer/Lao.rbbi (revision 0) +++ contrib/icu/src/resources/org/apache/lucene/icu/tokenizer/Lao.rbbi (revision 0) @@ -0,0 +1,215 @@ +# +# Licensed to the Apache Software Foundation (ASF) under one or more +# contributor license agreements. See the NOTICE file distributed with +# this work for additional information regarding copyright ownership. +# The ASF licenses this file to You under the Apache License, Version 2.0 +# (the "License"); you may not use this file except in compliance with +# the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +# +# +# Parses Lao text, with syllable as token. +# This is an example of how one can use a tailored ruleset. +# ICUBigramFilter can then be used to form bigrams of these syllables downstream +# +# The definition of Lao syllable is based from: +# +# Syllabification of Lao Script for Line Breaking +# Phonpasit Phissamay, Valaxay Dalolay, Chitaphone Chanhsililath, Oulaiphone Silimasak, +# Sarmad Hussain, Nadir Durrani, Science Technology and Environment Agency, CRULP +# http://www.panl10n.net/english/final%20reports/pdf%20files/Laos/LAO06.pdf +# http://www.panl10n.net/Presentations/Cambodia/Phonpassit/LineBreakingAlgo.pdf +# +# NOTE: +# There are some ambiguities in Lao syllabification without additional processing, as mentioned in the paper. +# For this reason, this RBBI grammar really only works with LaoBreakIterator +# For example, what appears to be a final consonant might instead be part of the next syllable. +# The first rule will match in a greedy fashion, leaving an illegal sequence that matches no rules. +# The first rule greedily matches ກວ່າດ, but then ອກ is encountered, which is illegal. +# +# What LaoBreakIterator does, according to the paper: +# 1. backtrack and remove the ດ from the last syllable, placing it on the current syllable. +# 2. verify the modified previous syllable (ກວ່າ ) is still legal. +# 3. verify the modified current syllable (ດອກ) is now legal. +# 4. If 2 or 3 fails, then restore the ດ to the last syllable and skip the current character. +# +# Finally, LaoBreakIterator also takes care of the second concern mentioned in the paper. +# This is the issue of combining marks being in the wrong order (typos) +# +# TODO: Since we turned off rule-chaining, write the numeric rules in unchained form. +# +# Syllable structure, where X is the nuclear consonant: +# +# +----+ +# | X5 | +# +----+ +# | X4 | +# +----+----+----+----+----+----+----+-----+ +# | X0 | X1 | X | X6 | X7 | X8 | X9 | X10 | +# +----+----+----+----+----+----+----+-----+ +# | X2 | +# +----+ +# | X3 | +# +----+ +# +# X0 represents a vowel which occurs before the nuclear consonant. +# It can always define the beginning of syllable. +$X0 = [\u0EC0-\u0EC4]; +# X1 is a combination consonant which comes before the nuclear consonant, +# but only if nuclear consonant is one of {ງ ຍ ລ ວ ຼ ມ ນ ຣ} +$X1 = [\u0EAB]; +# X represents the nuclear consonant. +$X = [\u0E81-\u0EAE\u0EDC\u0EDD]; +# X2 is a combination consonant which comes after the nuclear consonant, +# which is placed under or next to the nuclear consonant. +$X2 = [\u0EBC\u0EA3\u0EA7\u0EA5]; +# X3 represents a vowel which occurs under the nuclear consonant. +$X3 = [\u0EB8\u0EB9]; +# X4 represents a vowel which occurs above the nuclear consonant. +$X4 = [\u0EB4-\u0EB7\u0ECD\u0EBB\u0EB1]; +# X5 represents a tone mark which occurs above the nuclear consonant or upper vowel. +$X5 = [\u0EC8-\u0ECB]; +# X6 represents a consonant vowel, which occurs after the nuclear consonant. +# It functions when the syllable doesn’t have any vowels. And it always exists with X8. +$X6 = [\u0EA7\u0EAD\u0EBD]; +# X7 represents a final vowel. +# However X7_1 always represents the end of syllable and it never exists with tone mark. +$X7 = [\u0EB0\u0EB2\u0EB3]; +# X8 represents an alternate consonant. +$X8 = [\u0E81\u0E87\u0E8D\u0E94\u0E99\u0EA1\u0E9A\u0EA7]; +# X9 represents alternate consonants to pronounce foreign terms, it always exist with X10_3. +$X9 = [\u0E88\u0EAA\u0E8A\u0E9E\u0E9F\u0EA5]; +# X10 represents a sign mark. +# It always occurs at the end of a syllable, but mostly people keep it separate from syllable. +$X10 = [\u0EAF\u0EC6\u0ECC]; + +# Section 1 +$X0_1 = [\u0EC0]; +$X4_1_2 = [\u0EB4\u0EB5]; +$X4_3_4 = [\u0EB6\u0EB7]; +$X4_6 = [\u0EBB]; +$X4_7 = [\u0EB1]; +$X6_2 = [\u0EAD]; +$X6_3 = [\u0EBD]; +$X7_1 = [\u0EB0]; +$X7_2 = [\u0EB2]; +$X10_1 = [\u0EAF]; +$X10_2 = [\u0EC6]; +$X10_3 = [\u0ECC]; + +$Rule1_1 = $X0_1 ($X1)? $X ($X2)? ($X5)? ($X8)? ($X9 $X10_3)? ($X10_2)? ($X10_1)?; +$Rule1_2 = $X0_1 ($X1)? $X ($X2)? $X4_1_2 ($X5)? ($X8)? ($X9 $X10_3)? ($X10_2)? ($X10_1)?; +$Rule1_3 = $X0_1 ($X1)? $X ($X2)? $X4_3_4 ($X5)? $X6_2 ($X8)? ($X9 $X10_3)? ($X10_2)? ($X10_1)?; +$Rule1_4 = $X0_1 ($X1)? $X ($X2)? ($X7_2)? $X7_1; +$Rule1_5 = $X0_1 ($X1)? $X ($X2)? $X4_6 ($X5)? $X7_2 ($X9 $X10_3)? ($X10_2)? ($X10_1)?; +$Rule1_6 = $X0_1 ($X1)? $X ($X2)? $X4_7 ($X5)? $X8 ($X9 $X10_3)? ($X10_2)? ($X10_1)?; +$Rule1_7 = $X0_1 ($X1)? $X ($X2)? ($X4_7)? ($X5)? $X6_3 ($X9 $X10_3)? ($X10_2)? ($X10_1)?; + +$Rule1 = ($Rule1_1 | $Rule1_2 | $Rule1_3 | $Rule1_4 | $Rule1_5 | $Rule1_6 | $Rule1_7); + +# Section 2 +$X0_2 = [\u0EC1]; + +$Rule2_1 = $X0_2 ($X1)? $X ($X2)? ($X5)? ($X6)? ($X8)? ($X9 $X10_3)? ($X10_2)? ($X10_1)?; +$Rule2_2 = $X0_2 ($X1)? $X ($X2)? $X7_1; +$Rule2_3 = $X0_2 ($X1)? $X ($X2)? $X4_7 ($X5)? $X8 ($X9 $X10_3)? ($X10_2)? ($X10_1)?; + +$Rule2 = ($Rule2_1 | $Rule2_2 | $Rule2_3); + +# Section 3 +$X0_3 = [\u0EC2]; +$X8_3 = [\u0E8D]; +$X8_8 = [\u0EA7]; + +$Rule3_1 = $X0_3 ($X1)? $X ($X2)? ($X5)? ($X8)? ($X9 $X10_3)? ($X10_2)? ($X10_1)?; +$Rule3_2 = $X0_3 ($X1)? $X ($X2)? $X7_1; +$Rule3_3 = $X0_3 ($X1)? $X ($X2)? $X4_7 ($X5)? ($X8_3 | $X8_8); + +$Rule3 = ($Rule3_1 | $Rule3_2 | $Rule3_3); + +# Section 4 +$X0_4 = [\u0EC4]; +$X6_1 = [\u0EA7]; + +$Rule4 = $X0_4 ($X1)? $X ($X2)? ($X5)? ($X6_1)? ($X9 $X10_3)? ($X10_2)? ($X10_1)?; + +# Section 5 +$X0_5 = [\u0EC3]; + +$Rule5 = $X0_5 ($X1)? $X ($X2)? ($X5)? ($X6_1)? ($X9 $X10_3)? ($X10_2)? ($X10_1)?; + +# Section 6 +$Rule6 = ($X1)? $X ($X2)? $X3 ($X5)? ($X8)? ($X9 $X10_3)? ($X10_2)? ($X10_1)?; + +# Section 7 +$X4_1_4 = [\u0EB4-\u0EB7]; + +$Rule7 = ($X1)? $X ($X2)? $X4_1_4 ($X5)? ($X8)? ($X9 $X10_3)? ($X10_2)? ($X10_1)?; + +# Section 8 +$X4_5 = [\u0ECD]; + +$Rule8 = ($X1)? $X ($X2)? $X4_5 ($X5)? ($X7_2)? ($X9 $X10_3)? ($X10_2)? ($X10_1)?; + +# Section 9 + +$Rule9_1 = ($X1)? $X ($X2)? $X4_6 ($X5)? $X8 ($X9 $X10_3)? ($X10_2)? ($X10_1)?; +$Rule9_2 = ($X1)? $X ($X2)? $X4_6 ($X5)? $X6_1 $X7_1; + +$Rule9 = ($Rule9_1 | $Rule9_2); + +# Section 10 +$Rule10 = ($X1)? $X ($X2)? $X4_7 ($X5)? ($X6_1)? $X8 ($X9 $X10_3)? ($X10_2)? ($X10_1)?; + +# Section 11 +$Rule11 = ($X1)? $X ($X2)? ($X5)? $X6 $X8 ($X9 $X10_3)? ($X10_2)? ($X10_1)?; + +# Section 12 +$Rule12 = ($X1)? $X ($X2)? ($X5)? $X7_1; + +# Section 13 +$Rule13 = ($X1)? $X ($X2)? ($X5)? $X7_2 ($X8)? ($X9 $X10_3)? ($X10_2)? ($X10_1)?; + +# Section 14 +$X7_3 = [\u0EB3]; + +$Rule14 = ($X1)? $X ($X2)? ($X5)? $X7_3 ($X9 $X10_3)? ($X10_2)? ($X10_1)?; + +$LaoSyllableEx = ($Rule1 | $Rule2 | $Rule3 | $Rule4 | $Rule5 | $Rule6 | $Rule7 | $Rule8 | $Rule9 | $Rule10 | $Rule11 | $Rule12 | $Rule13 | $Rule14); + +$WordJoin = [:Line_Break=Word_Joiner:]; + +$LaoJoinedSyllableEx = $LaoSyllableEx ($WordJoin $LaoSyllableEx)*; + +# +# default numerical definitions +# +$Extend = [\p{Word_Break = Extend}]; +$Format = [\p{Word_Break = Format}]; +$MidNumLet = [\p{Word_Break = MidNumLet}]; +$MidNum = [\p{Word_Break = MidNum}]; +$Numeric = [\p{Word_Break = Numeric}]; +$ExtendNumLet = [\p{Word_Break = ExtendNumLet}]; +$MidNumLetEx = $MidNumLet ($Extend | $Format)*; +$MidNumEx = $MidNum ($Extend | $Format)*; +$NumericEx = $Numeric ($Extend | $Format)*; +$ExtendNumLetEx = $ExtendNumLet ($Extend | $Format)*; + +!!forward; +$LaoJoinedSyllableEx {200}; + +# +# default numeric rules +# +$NumericEx {100}; +$NumericEx $NumericEx {100}; +$NumericEx ($MidNumEx | $MidNumLetEx) $NumericEx {100}; +$NumericEx $ExtendNumLetEx {100}; +$ExtendNumLetEx $NumericEx {100}; Index: contrib/icu/src/resources/org/apache/lucene/icu/tokenizer/Myanmar.rbbi =================================================================== --- contrib/icu/src/resources/org/apache/lucene/icu/tokenizer/Myanmar.rbbi (revision 0) +++ contrib/icu/src/resources/org/apache/lucene/icu/tokenizer/Myanmar.rbbi (revision 0) @@ -0,0 +1,58 @@ +# +# Licensed to the Apache Software Foundation (ASF) under one or more +# contributor license agreements. See the NOTICE file distributed with +# this work for additional information regarding copyright ownership. +# The ASF licenses this file to You under the Apache License, Version 2.0 +# (the "License"); you may not use this file except in compliance with +# the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. +# +# +# Parses Myanmar text, with syllable as token. +# ICUBigramFilter can be used to form bigrams of these syllables downstream. +# +!!chain; +# + +$Cons = [[:Other_Letter:]&[:Myanmar:]]; +$Virama = [\u1039]; +$Asat = [\u103A]; + +$WordJoin = [:Line_Break=Word_Joiner:]; +# +# default numerical definitions +# +$Extend = [\p{Word_Break = Extend}]; +$Format = [\p{Word_Break = Format}]; +$MidNumLet = [\p{Word_Break = MidNumLet}]; +$MidNum = [\p{Word_Break = MidNum}]; +$Numeric = [\p{Word_Break = Numeric}]; +$ExtendNumLet = [\p{Word_Break = ExtendNumLet}]; +$MidNumLetEx = $MidNumLet ($Extend | $Format)*; +$MidNumEx = $MidNum ($Extend | $Format)*; +$NumericEx = $Numeric ($Extend | $Format)*; +$ExtendNumLetEx = $ExtendNumLet ($Extend | $Format)*; + +$ConsEx = $Cons ($Extend | $Format)*; +$AsatEx = $Cons $Asat ($Virama $ConsEx)? ($Extend | $Format)*; +$MyanmarSyllableEx = $ConsEx ($Virama $ConsEx)? ($AsatEx)*; +$MyanmarJoinedSyllableEx = $MyanmarSyllableEx ($WordJoin $MyanmarSyllableEx)*; + +!!forward; +$MyanmarJoinedSyllableEx {200}; + +# +# default numeric rules +# +$NumericEx {100}; +$NumericEx $NumericEx {100}; +$NumericEx ($MidNumEx | $MidNumLetEx) $NumericEx {100}; +$NumericEx $ExtendNumLetEx {100}; +$ExtendNumLetEx $NumericEx {100}; Index: contrib/icu/src/test/org/apache/lucene/icu/TestICUAnalyzer.java =================================================================== --- contrib/icu/src/test/org/apache/lucene/icu/TestICUAnalyzer.java (revision 0) +++ contrib/icu/src/test/org/apache/lucene/icu/TestICUAnalyzer.java (revision 0) @@ -0,0 +1,208 @@ +package org.apache.lucene.icu; + +/** + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +import org.apache.lucene.analysis.Analyzer; +import org.apache.lucene.analysis.BaseTokenStreamTestCase; + +public class TestICUAnalyzer extends BaseTokenStreamTestCase { + private Analyzer a = new ICUAnalyzer(); + + /* simple tests from a few sample different writing systems and languages + * TODO: more tests + */ + + public void testArmenian() throws Exception { + assertAnalyzesTo(a, "Վիքիպեդիայի 13 միլիոն հոդվածները (4,600` հայերեն վիքիպեդիայում) գրվել են կամավորների կողմից ու համարյա բոլոր հոդվածները կարող է խմբագրել ցանկաց մարդ ով կարող է բացել Վիքիպեդիայի կայքը։", + new String[] { "վիքիպեդիայի", "13", "միլիոն", "հոդվածները", "4,600", "հայերեն", "վիքիպեդիայում", "գրվել", "են", "կամավորների", "կողմից", + "ու", "համարյա", "բոլոր", "հոդվածները", "կարող", "է", "խմբագրել", "ցանկաց", "մարդ", "ով", "կարող", "է", "բացել", "վիքիպեդիայի", "կայքը" } ); + } + + public void testAmharic() throws Exception { + assertAnalyzesTo(a, "ዊኪፔድያ የባለ ብዙ ቋንቋ የተሟላ ትክክለኛና ነጻ መዝገበ ዕውቀት (ኢንሳይክሎፒዲያ) ነው። ማንኛውም", + new String[] { "ዊኪፔድያ", "የባለ", "ብዙ", "ቋንቋ", "የተሟላ", "ትክክለኛና", "ነጻ", "መዝገበ", "ዕውቀት", "ኢንሳይክሎፒዲያ", "ነው", "ማንኛውም" } ); + } + + public void testArabic() throws Exception { + assertAnalyzesTo(a, "الفيلم الوثائقي الأول عن ويكيبيديا يسمى \"الحقيقة بالأرقام: قصة ويكيبيديا\" (بالإنجليزية: Truth in Numbers: The Wikipedia Story)، سيتم إطلاقه في 2008.", + new String[] { "الفيلم", "الوثائقي", "الأول", "عن", "ويكيبيديا", "يسمى", "الحقيقة", "بالأرقام", "قصة", "ويكيبيديا", + "بالإنجليزية", "truth", "in", "numbers", "the", "wikipedia", "story", "سيتم", "إطلاقه", "في", "2008" } ); + } + + public void testAramaic() throws Exception { + assertAnalyzesTo(a, "ܘܝܩܝܦܕܝܐ (ܐܢܓܠܝܐ: Wikipedia) ܗܘ ܐܝܢܣܩܠܘܦܕܝܐ ܚܐܪܬܐ ܕܐܢܛܪܢܛ ܒܠܫܢ̈ܐ ܣܓܝܐ̈ܐ܂ ܫܡܗ ܐܬܐ ܡܢ ܡ̈ܠܬܐ ܕ\"ܘܝܩܝ\" ܘ\"ܐܝܢܣܩܠܘܦܕܝܐ\"܀", + new String[] { "ܘܝܩܝܦܕܝܐ", "ܐܢܓܠܝܐ", "wikipedia", "ܗܘ", "ܐܝܢܣܩܠܘܦܕܝܐ", "ܚܐܪܬܐ", "ܕܐܢܛܪܢܛ", "ܒܠܫܢ̈ܐ", "ܣܓܝܐ̈ܐ", "ܫܡܗ", + "ܐܬܐ", "ܡܢ", "ܡ̈ܠܬܐ", "ܕ", "ܘܝܩܝ", "ܘ", "ܐܝܢܣܩܠܘܦܕܝܐ"}); + } + + public void testBengali() throws Exception { + assertAnalyzesTo(a, "এই বিশ্বকোষ পরিচালনা করে উইকিমিডিয়া ফাউন্ডেশন (একটি অলাভজনক সংস্থা)। উইকিপিডিয়ার শুরু ১৫ জানুয়ারি, ২০০১ সালে। এখন পর্যন্ত ২০০টিরও বেশী ভাষায় উইকিপিডিয়া রয়েছে।", + new String[] { "এই", "বিশ্বকোষ", "পরিচালনা", "করে", "উইকিমিডিয়া", "ফাউন্ডেশন", "একটি", "অলাভজনক", "সংস্থা", "উইকিপিডিয়ার", + "শুরু", "15", "জানুয়ারি", "2001", "সালে", "এখন", "পর্যন্ত", "200টিরও", "বেশী", "ভাষায়", "উইকিপিডিয়া", "রয়েছে" }); + } + + public void testFarsi() throws Exception { + assertAnalyzesTo(a, "ویکی پدیای انگلیسی در تاریخ ۲۵ دی ۱۳۷۹ به صورت مکملی برای دانشنامهٔ تخصصی نوپدیا نوشته شد.", + new String[] { "ویکی", "پدیای", "انگلیسی", "در", "تاریخ", "25", "دی", "1379", "به", "صورت", "مکملی", + "برای", "دانشنامهٔ", "تخصصی", "نوپدیا", "نوشته", "شد" }); + } + + public void testGreek() throws Exception { + assertAnalyzesTo(a, "Γράφεται σε συνεργασία από εθελοντές με το λογισμικό wiki, κάτι που σημαίνει ότι άρθρα μπορεί να προστεθούν ή να αλλάξουν από τον καθένα.", + new String[] { "γράφεται", "σε", "συνεργασία", "από", "εθελοντέσ", "με", "το", "λογισμικό", "wiki", "κάτι", "που", + "σημαίνει", "ότι", "άρθρα", "μπορεί", "να", "προστεθούν", "ή", "να", "αλλάξουν", "από", "τον", "καθένα" }); + } + + public void testLao() throws Exception { + assertAnalyzesTo(a, "ກວ່າດອກ", new String[] { "ກວ່າ", "ດອກ", "ກວ່າດອກ" }); + } + + public void testThai() throws Exception { + assertAnalyzesTo(a, "การที่ได้ต้องแสดงว่างานดี. แล้วเธอจะไปไหน? ๑๒๓๔", + new String[] { "การ", "ที่", "ได้", "ต้อง", "แสดง", "ว่า", "งาน", "ดี", "แล้ว", "เธอ", "จะ", "ไป", "ไหน", "1234"}); + } + + public void testTibetan() throws Exception { + assertAnalyzesTo(a, "སྣོན་མཛོད་དང་ལས་འདིས་བོད་ཡིག་མི་ཉམས་གོང་འཕེལ་དུ་གཏོང་བར་ཧ་ཅང་དགེ་མཚན་མཆིས་སོ། །", + new String[] { "སྣོན", "མཛོད", "དང", "ལས", "འདིས", "བོད", "ཡིག", "མི", "ཉམས", "གོང", "འཕེལ", "དུ", "གཏོང", "བར", "ཧ", "ཅང", "དགེ", "མཚན", "མཆིས", "སོ" }); + } + + /* + * For chinese, use bigrams + * TODO: why do full-width numerics have no word-break prop? + */ + public void testChinese() throws Exception { + assertAnalyzesTo(a, "我是中国人。 １２３４ Ｔｅｓｔｓ ", + new String[] { "我", "是", "我是", "中", "是中", "国", "中国", "人", "国人", "tests"}); + } + + public void testEmpty() throws Exception { + assertAnalyzesTo(a, "", new String[] {}); + assertAnalyzesTo(a, ".", new String[] {}); + assertAnalyzesTo(a, " ", new String[] {}); + } + + /* test various jira issues this analyzer is related to */ + + public void testLUCENE1032() throws Exception { + /* + * Some of our Japanese customers are reporting errors when performing searches using half width characters. + * The desired behavior is that a document containing half width characters should be returned when performing + * a search using full width equivalents or when searching by the half width character itself. + * Currently, a search will not return any matches for half width characters. + * + * ICUAnalyzer normalizes to NFKC by default. + */ + byte[] fullWidthKa = new byte[]{(byte) 0xE3, (byte) 0x82, (byte) 0xAB}; + byte[] halfWidthKa = new byte[]{(byte) 0xEF, (byte) 0xBD, (byte) 0xB6}; + + assertAnalyzesTo(a, new String(halfWidthKa, "UTF-8"), new String[] { new String(fullWidthKa, "UTF-8") }); + assertAnalyzesTo(a, new String(fullWidthKa, "UTF-8"), new String[] { new String(fullWidthKa, "UTF-8") }); + } + + public void testLUCENE1215() throws Exception { + /* + * New in java 6, we have java.text.Normalizer that supports Unicode Standard Annex #15 normalization. + * + * FYI: The java6 impl is String-only, and does not provide quickCheck, although isNormalized MIGHT invoke quickCheck. + * + * ICUAnalyzer normalizes to NFKC by default (though this can be changed). + * See tests for ICUNormalizationFilter. + */ + + assertAnalyzesTo(a, "ﴳﴺﰧ", new String[] { "طمطمطم" }); // arabic presentation forms + } + + public void testLUCENE1545() throws Exception { + /* + * Standard analyzer does not correctly tokenize combining character U+0364 COMBINING LATIN SMALL LETTRE E. + * The word "moͤchte" is incorrectly tokenized into "mo" "chte", the combining character is lost. + * Expected result is only on token "moͤchte". + * + * ICUAnalyzer implements Unicode Text Segmentation, which never separates a combining mark from its base character. + */ + assertAnalyzesTo(a, "moͤchte", new String[] { "moͤchte" }); + } + + // public void testLUCENE1161() throws Exception { + /* + * It would be useful, in the StandardTokenizer, to be able to have more control over in-word punctuation is handled. + * For instance, it is not always desirable to split on dashes or other punctuation. + * + * Figure out nice example of this with a tailored ruleset. + * Its possible to even do something such as provide DBBI for hyphenation, etc etc. + */ + // } + + /* Tests from StandardAnalyzer, just to show behavior is similar */ + public void testAlphanumericSA() throws Exception { + // alphanumeric tokens + assertAnalyzesTo(a, "B2B", new String[]{"b2b"}); + assertAnalyzesTo(a, "2B", new String[]{"2b"}); + } + + public void testDelimitersSA() throws Exception { + // other delimiters: "-", "/", "," + assertAnalyzesTo(a, "some-dashed-phrase", new String[]{"some", "dashed", "phrase"}); + assertAnalyzesTo(a, "dogs,chase,cats", new String[]{"dogs", "chase", "cats"}); + assertAnalyzesTo(a, "ac/dc", new String[]{"ac", "dc"}); + } + + public void testApostrophesSA() throws Exception { + // internal apostrophes: O'Reilly, you're, O'Reilly's + assertAnalyzesTo(a, "O'Reilly", new String[]{"o'reilly"}); + assertAnalyzesTo(a, "you're", new String[]{"you're"}); + assertAnalyzesTo(a, "she's", new String[]{"she's"}); + assertAnalyzesTo(a, "Jim's", new String[]{"jim's"}); + assertAnalyzesTo(a, "don't", new String[]{"don't"}); + assertAnalyzesTo(a, "O'Reilly's", new String[]{"o'reilly's"}); + } + + public void testNumericSA() throws Exception { + // floating point, serial, model numbers, ip addresses, etc. + // every other segment must have at least one digit + assertAnalyzesTo(a, "21.35", new String[]{"21.35"}); + assertAnalyzesTo(a, "R2D2 C3PO", new String[]{"r2d2", "c3po"}); + assertAnalyzesTo(a, "216.239.63.104", new String[]{"216.239.63.104"}); + assertAnalyzesTo(a, "216.239.63.104", new String[]{"216.239.63.104"}); + assertAnalyzesTo(a, "২০৬৭০৩", new String[]{"206703"}); + } + + public void testTextWithNumbersSA() throws Exception { + // numbers + assertAnalyzesTo(a, "David has 5000 bones", new String[]{"david", "has", "5000", "bones"}); + } + + public void testVariousTextSA() throws Exception { + // various + assertAnalyzesTo(a, "C embedded developers wanted", new String[]{"c", "embedded", "developers", "wanted"}); + assertAnalyzesTo(a, "foo bar FOO BAR", new String[]{"foo", "bar", "foo", "bar"}); + assertAnalyzesTo(a, "foo bar . FOO <> BAR", new String[]{"foo", "bar", "foo", "bar"}); + assertAnalyzesTo(a, "\"QUOTED\" word", new String[]{"quoted", "word"}); + } + + public void testKoreanSA() throws Exception { + // Korean words + assertAnalyzesTo(a, "안녕하세요 한글입니다", new String[]{"안녕하세요", "한글입니다"}); + } + + public void testReusableTokenStream() throws Exception { + assertAnalyzesToReuse(a, "སྣོན་མཛོད་དང་ལས་འདིས་བོད་ཡིག་མི་ཉམས་གོང་འཕེལ་དུ་གཏོང་བར་ཧ་ཅང་དགེ་མཚན་མཆིས་སོ། །", + new String[] { "སྣོན", "མཛོད", "དང", "ལས", "འདིས", "བོད", "ཡིག", "མི", "ཉམས", "གོང", "འཕེལ", "དུ", "གཏོང", "བར", "ཧ", "ཅང", "དགེ", "མཚན", "མཆིས", "སོ" }); + } +} Property changes on: contrib\icu\src\test\org\apache\lucene\icu\TestICUAnalyzer.java ___________________________________________________________________ Added: svn:eol-style + native Index: contrib/icu/src/test/org/apache/lucene/icu/TestICUBigramFilter.java =================================================================== --- contrib/icu/src/test/org/apache/lucene/icu/TestICUBigramFilter.java (revision 0) +++ contrib/icu/src/test/org/apache/lucene/icu/TestICUBigramFilter.java (revision 0) @@ -0,0 +1,50 @@ +package org.apache.lucene.icu; + +/** + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +import java.io.StringReader; + +import org.apache.lucene.analysis.BaseTokenStreamTestCase; +import org.apache.lucene.analysis.TokenStream; +import org.apache.lucene.icu.tokenizer.ICUTokenizer; + +import com.ibm.icu.lang.UScript; +import com.ibm.icu.text.Normalizer; + +public class TestICUBigramFilter extends BaseTokenStreamTestCase { + + public void testBasicFunctionality() throws Exception { + TokenStream ts = new ICUBigramFilter(new ICUTokenizer(new StringReader("我购买了")), + Normalizer.NFC, new int [] { UScript.HAN }); + assertTokenStreamContents(ts, new String[] { "我", "购", "我购", "买", "购买", "了", "买了" }); + } + + + public void testMixedScripts() throws Exception { + TokenStream ts = new ICUBigramFilter(new ICUTokenizer(new StringReader("test我购买了abcd")), + Normalizer.NFC, new int [] { UScript.HAN }); + assertTokenStreamContents(ts, new String[] { "test", "我", "购", "我购", "买", "购买", "了", "买了", "abcd" }); + } + + public void testMixedScripts2() throws Exception { + TokenStream ts = new ICUBigramFilter(new ICUTokenizer(new StringReader("test我购买了ချောတဲ့လူabcd")), + Normalizer.NFC, new int [] { UScript.HAN, UScript.MYANMAR }); + assertTokenStreamContents(ts, new String[] { "test", "我", "购", "我购", "买", "购买", "了", "买了", + "ချော", "တဲ့", "ချောတဲ့", "လူ", "တဲ့လူ", "abcd" }); + } +} Property changes on: contrib\icu\src\test\org\apache\lucene\icu\TestICUBigramFilter.java ___________________________________________________________________ Added: svn:eol-style + native Index: contrib/icu/src/test/org/apache/lucene/icu/TestICUCaseFoldingFilter.java =================================================================== --- contrib/icu/src/test/org/apache/lucene/icu/TestICUCaseFoldingFilter.java (revision 0) +++ contrib/icu/src/test/org/apache/lucene/icu/TestICUCaseFoldingFilter.java (revision 0) @@ -0,0 +1,196 @@ +package org.apache.lucene.icu; + +/** + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +import java.io.IOException; +import java.io.StringReader; + +import org.apache.lucene.analysis.BaseTokenStreamTestCase; +import org.apache.lucene.analysis.KeywordTokenizer; +import org.apache.lucene.analysis.TokenFilter; +import org.apache.lucene.analysis.TokenStream; +import org.apache.lucene.analysis.Tokenizer; + +import com.ibm.icu.lang.UCharacter; +import com.ibm.icu.text.Normalizer; +import com.ibm.icu.text.UnicodeSet; +import com.ibm.icu.text.UnicodeSetIterator; +import com.ibm.icu.text.Normalizer.Mode; + +public class TestICUCaseFoldingFilter extends BaseTokenStreamTestCase { + private Tokenizer tokenizer = new KeywordTokenizer(new StringReader("")); + + private TokenFilter nfkcFold = new ICUCaseFoldingFilter(tokenizer, Normalizer.NFKC); + private TokenFilter nfcFold = new ICUCaseFoldingFilter(tokenizer, Normalizer.NFC); + private TokenFilter nfkdFold = new ICUCaseFoldingFilter(tokenizer, Normalizer.NFKD); + private TokenFilter nfdFold = new ICUCaseFoldingFilter(tokenizer, Normalizer.NFD); + private TokenFilter noneFold = new ICUCaseFoldingFilter(tokenizer, Normalizer.NONE); + + private TokenFilter nfkcNorm = new ICUNormalizationFilter(nfkcFold, Normalizer.NFKC); + private TokenFilter nfcNorm = new ICUNormalizationFilter(nfcFold, Normalizer.NFC); + private TokenFilter nfkdNorm = new ICUNormalizationFilter(nfkdFold, Normalizer.NFKD); + private TokenFilter nfdNorm = new ICUNormalizationFilter(nfdFold, Normalizer.NFD); + private TokenFilter noneNorm = new ICUNormalizationFilter(noneFold, Normalizer.NONE); + + /** + * Some basic case-folding, including a multi-codepoint folding and a supplementary codepoint folding. + */ + public void testBasicFunctionality() throws Exception { + checkToken(Normalizer.NONE, "LuCeNE", "lucene"); // latin + checkToken(Normalizer.NONE, "Ruß", "russ"); // german + checkToken(Normalizer.NONE, "𐐖", "𐐾"); // suppl. codepoint + } + + /** + * Validate the case-folder folds all unicode codepoints the same way as the UCharacter String-based method. + */ + public void testUnicodeSet() throws Exception { + UnicodeSet set = new UnicodeSet(UnicodeSet.MIN_VALUE, UnicodeSet.MAX_VALUE); + for (UnicodeSetIterator it = new UnicodeSetIterator(set); it.next();) { + String string = it.getString(); + checkToken(Normalizer.NONE, string, icuFold(string)); + } + } + + /** + * Case-fold the entire set of unicode codepoints as one huge term, and validate it against the the UCharacter String-based method + */ + public void testUnicodeSetOneTerm() throws Exception { + String allUnicode = unicodeSetString(); + checkToken(Normalizer.NONE, allUnicode, icuFold(allUnicode)); + } + + + /** + * Validate closure under normalization mode NFC. This is the default for the case-folding algorithm, but it should still work! + */ + public void testUnicodeClosureNFC() throws Exception { + UnicodeSet set = new UnicodeSet(UnicodeSet.MIN_VALUE, UnicodeSet.MAX_VALUE); + for (UnicodeSetIterator it = new UnicodeSetIterator(set); it.next();) { + String string = it.getString(); + checkTokenClosure(Normalizer.NFC, string, icuNFC(icuFold(icuNFC(icuFold(string))))); + } + } + + /** + * Validate closure under normalization mode NFD. This is the default for the case-folding algorithm, but it should still work! + */ + public void testUnicodeClosureNFD() throws Exception { + UnicodeSet set = new UnicodeSet(UnicodeSet.MIN_VALUE, UnicodeSet.MAX_VALUE); + for (UnicodeSetIterator it = new UnicodeSetIterator(set); it.next();) { + String string = it.getString(); + checkTokenClosure(Normalizer.NFD, string, icuNFD(icuFold(icuNFD(icuFold(string))))); + } + } + + /** + * Validate closure under normalization mode NFKC. + * In this case the filter will apply NFKC_Closure set to prevent from having to normalize, fold, normalize, fold. + */ + public void testUnicodeClosureNFKC() throws Exception { + UnicodeSet set = new UnicodeSet(UnicodeSet.MIN_VALUE, UnicodeSet.MAX_VALUE); + for (UnicodeSetIterator it = new UnicodeSetIterator(set); it.next();) { + String string = it.getString(); + checkTokenClosure(Normalizer.NFKC, string, icuNFKC(icuFold(icuNFKC(icuFold(string))))); + } + } + + /** + * Validate closure under normalization mode NFKD. + * In this case the filter will apply NFKC_Closure set to prevent from having to normalize, fold, normalize, fold. + */ + public void testUnicodeClosureNFKD() throws Exception { + UnicodeSet set = new UnicodeSet(UnicodeSet.MIN_VALUE, UnicodeSet.MAX_VALUE); + for (UnicodeSetIterator it = new UnicodeSetIterator(set); it.next();) { + String string = it.getString(); + checkTokenClosure(Normalizer.NFKD, string, icuNFKD(icuFold(icuNFKD(icuFold(string))))); + } + } + + String unicodeSetString() { + StringBuilder sb = new StringBuilder(); + UnicodeSet set = new UnicodeSet(UnicodeSet.MIN_VALUE, UnicodeSet.MAX_VALUE); + for (UnicodeSetIterator it = new UnicodeSetIterator(set); it.next();) + sb.append(it.getString()); + return sb.toString(); + } + + String icuFold(String s) { + return UCharacter.foldCase(s, true); + } + + String icuNFKD(String s) { + return Normalizer.normalize(s, Normalizer.NFKD); + } + + String icuNFD(String s) { + return Normalizer.normalize(s, Normalizer.NFD); + } + + String icuNFC(String s) { + return Normalizer.normalize(s, Normalizer.NFC); + } + + String icuNFKC(String s) { + return Normalizer.normalize(s, Normalizer.NFKC); + } + + TokenFilter getFilterForMode(Mode mode) { + if (mode == Normalizer.NFC) + return nfcFold; + else if (mode == Normalizer.NFD) + return nfdFold; + else if (mode == Normalizer.NFKD) + return nfkdFold; + else if (mode == Normalizer.NFKC) + return nfkcFold; + else if (mode == Normalizer.NONE) + return noneFold; + else + throw new UnsupportedOperationException("this test class does not support testing this mode!"); + } + + TokenFilter getFilterForModeClosure(Mode mode) { + if (mode == Normalizer.NFC) + return nfcNorm; + else if (mode == Normalizer.NFD) + return nfdNorm; + else if (mode == Normalizer.NFKD) + return nfkdNorm; + else if (mode == Normalizer.NFKC) + return nfkcNorm; + else if (mode == Normalizer.NONE) + return noneNorm; + else + throw new UnsupportedOperationException("this test class does not support testing this mode!"); + } + + void checkToken(Mode mode, String input, String expected) throws IOException { + tokenizer.reset(new StringReader(input)); + TokenStream filter = getFilterForMode(mode); + filter.reset(); + assertTokenStreamContents(filter, new String[] { expected }); + } + + void checkTokenClosure(Mode mode, String input, String expected) throws IOException { + tokenizer.reset(new StringReader(input)); + TokenStream filter = getFilterForModeClosure(mode); + filter.reset(); + assertTokenStreamContents(filter, new String[] { expected }); + } +} Property changes on: contrib\icu\src\test\org\apache\lucene\icu\TestICUCaseFoldingFilter.java ___________________________________________________________________ Added: svn:eol-style + native Index: contrib/icu/src/test/org/apache/lucene/icu/TestICUFormatFilter.java =================================================================== --- contrib/icu/src/test/org/apache/lucene/icu/TestICUFormatFilter.java (revision 0) +++ contrib/icu/src/test/org/apache/lucene/icu/TestICUFormatFilter.java (revision 0) @@ -0,0 +1,31 @@ +package org.apache.lucene.icu; + +/** + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +import java.io.StringReader; + +import org.apache.lucene.analysis.BaseTokenStreamTestCase; +import org.apache.lucene.analysis.KeywordTokenizer; +import org.apache.lucene.analysis.TokenStream; + +public class TestICUFormatFilter extends BaseTokenStreamTestCase { + public void testBasicFunctionality() throws Exception { + TokenStream ts = new ICUFormatFilter(new KeywordTokenizer(new StringReader("क्ष"))); + assertTokenStreamContents(ts, new String[] { "क्ष" }); + } +} Property changes on: contrib\icu\src\test\org\apache\lucene\icu\TestICUFormatFilter.java ___________________________________________________________________ Added: svn:eol-style + native Index: contrib/icu/src/test/org/apache/lucene/icu/TestICUNormalizationFilter.java =================================================================== --- contrib/icu/src/test/org/apache/lucene/icu/TestICUNormalizationFilter.java (revision 0) +++ contrib/icu/src/test/org/apache/lucene/icu/TestICUNormalizationFilter.java (revision 0) @@ -0,0 +1,109 @@ +package org.apache.lucene.icu; + +/** + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +import java.io.StringReader; + +import org.apache.lucene.analysis.BaseTokenStreamTestCase; +import org.apache.lucene.analysis.KeywordTokenizer; +import org.apache.lucene.analysis.TokenFilter; +import org.apache.lucene.analysis.TokenStream; +import org.apache.lucene.analysis.Tokenizer; + +import com.ibm.icu.text.Normalizer; +import com.ibm.icu.text.UnicodeSet; +import com.ibm.icu.text.UnicodeSetIterator; +import com.ibm.icu.text.Normalizer.Mode; + + +/** + * TODO: maybe just use the Unicode normalization test suite? + * + */ +public class TestICUNormalizationFilter extends BaseTokenStreamTestCase { + private Tokenizer tokenizer = new KeywordTokenizer(new StringReader("")); + + private TokenFilter nfkcNorm = new ICUNormalizationFilter(tokenizer, Normalizer.NFKC); + private TokenFilter nfcNorm = new ICUNormalizationFilter(tokenizer, Normalizer.NFC); + private TokenFilter nfkdNorm = new ICUNormalizationFilter(tokenizer, Normalizer.NFKD); + private TokenFilter nfdNorm = new ICUNormalizationFilter(tokenizer, Normalizer.NFD); + private TokenFilter noneNorm = new ICUNormalizationFilter(tokenizer, Normalizer.NONE); + + TokenFilter getFilterForMode(Mode mode) { + if (mode == Normalizer.NFC) + return nfcNorm; + else if (mode == Normalizer.NFD) + return nfdNorm; + else if (mode == Normalizer.NFKD) + return nfkdNorm; + else if (mode == Normalizer.NFKC) + return nfkcNorm; + else if (mode == Normalizer.NONE) + return noneNorm; + else + throw new UnsupportedOperationException("this test class does not support testing this mode!"); + } + + /** + * Validate the given String normalizes correctly in the provided mode against the ICU normalizer. + * + */ + void check(String input, Normalizer.Mode mode) throws Exception { + String expected = Normalizer.normalize(input, mode); + tokenizer.reset(new StringReader(input)); + TokenStream filter = getFilterForMode(mode); + assertTokenStreamContents(filter, new String[] { expected }); + } + + /** + * Validate all unicode codepoints against the given mode. + */ + void unicodeCompliance(Normalizer.Mode mode) throws Exception { + UnicodeSet set = new UnicodeSet(UnicodeSet.MIN_VALUE, UnicodeSet.MAX_VALUE); + for (UnicodeSetIterator it = new UnicodeSetIterator(set); it.next();) + check(it.getString(), mode); + } + + public void testComplianceNFC() throws Exception { + unicodeCompliance(Normalizer.NFC); + } + + public void testComplianceNFKC() throws Exception { + unicodeCompliance(Normalizer.NFKC); + } + + public void testComplianceNFD() throws Exception { + unicodeCompliance(Normalizer.NFD); + } + + public void testComplianceNFKD() throws Exception { + unicodeCompliance(Normalizer.NFKD); + } + + public void testComplianceNone() throws Exception { + unicodeCompliance(Normalizer.NONE); + } + + public void testHugeExpansion() throws Exception { + StringBuilder sb = new StringBuilder(); + for (int i = 0; i < 1000; i++) + sb.append('\uFDFA'); + String hugeString = sb.toString(); + check(hugeString, Normalizer.NFKC); + } +} Property changes on: contrib\icu\src\test\org\apache\lucene\icu\TestICUNormalizationFilter.java ___________________________________________________________________ Added: svn:eol-style + native Index: contrib/icu/src/test/org/apache/lucene/icu/TestICUTransformFilter.java =================================================================== --- contrib/icu/src/test/org/apache/lucene/icu/TestICUTransformFilter.java (revision 0) +++ contrib/icu/src/test/org/apache/lucene/icu/TestICUTransformFilter.java (revision 0) @@ -0,0 +1,92 @@ +package org.apache.lucene.icu; + +/** + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +import java.io.IOException; +import java.io.Reader; +import java.io.StringReader; + +import org.apache.lucene.analysis.Analyzer; +import org.apache.lucene.analysis.BaseTokenStreamTestCase; +import org.apache.lucene.analysis.KeywordTokenizer; +import org.apache.lucene.analysis.TokenStream; +import org.apache.lucene.icu.tokenizer.ICUTokenizer; + +import com.ibm.icu.text.Transliterator; +import com.ibm.icu.text.UnicodeSet; + + +/** + * Test the ICUTransformFilter with some basic examples. + */ +public class TestICUTransformFilter extends BaseTokenStreamTestCase { + + public void testBasicFunctionality() throws Exception { + checkToken(Transliterator.getInstance("Traditional-Simplified"), "簡化字", "简化字"); + checkToken(Transliterator.getInstance("Katakana-Hiragana"), "ヒラガナ", "ひらがな"); + checkToken(Transliterator.getInstance("Fullwidth-Halfwidth"), "アルアノリウ", "ｱﾙｱﾉﾘｳ"); + checkToken(Transliterator.getInstance("Any-Latin"), "Αλφαβητικός Κατάλογος", "Alphabētikós Katálogos"); + checkToken(Transliterator.getInstance("NFD; [:Nonspacing Mark:] Remove"), "Alphabētikós Katálogos", "Alphabetikos Katalogos"); + } + + public void testCustomFunctionality() throws Exception { + String rules = "a > b; b > c;"; // convert a's to b's and b's to c's + checkToken(Transliterator.createFromRules("test", rules, Transliterator.FORWARD), "abacadaba", "bcbcbdbcb"); + } + + public void testOptimizer() throws Exception { + String rules = "a > b; b > c;"; // convert a's to b's and b's to c's + Transliterator custom = Transliterator.createFromRules("test", rules, Transliterator.FORWARD); + assertTrue(custom.getFilter() == null); + new ICUTransformFilter(new KeywordTokenizer(new StringReader("")), custom); + assertTrue(custom.getFilter().equals(new UnicodeSet("[ab]"))); + } + + public void testOptimizerSurrogate() throws Exception { + String rules = "\\U00020087 > x;"; // convert CJK UNIFIED IDEOGRAPH-20087 to an x + Transliterator custom = Transliterator.createFromRules("test", rules, Transliterator.FORWARD); + assertTrue(custom.getFilter() == null); + new ICUTransformFilter(new KeywordTokenizer(new StringReader("")), custom); + assertTrue(custom.getFilter().equals(new UnicodeSet("[\\U00020087]"))); + } + + public void testLUCENE1343() throws Exception { + /* + * A replacement for ISOLatin1AccentFilter that does a more thorough job of removing diacritical marks or non-spacing modifiers. + * This issue also had a normalization impl, but no quickCheck. + * + * ICUAnalyzer does NOT remove any accents by default!!!! + * This is language-specific/usually incorrect behavior, but if you want to do this kind of thing, the components are here. + * + * If you want to do some additional custom mappings, you can create a custom Transliterator. + */ + Analyzer a = new Analyzer() { + public TokenStream tokenStream(String fieldName, Reader reader) { + return new ICUTransformFilter( + new ICUTokenizer(reader), Transliterator.getInstance("NFKD; [:Nonspacing Mark:] Remove; NFC")); + } }; + + assertAnalyzesTo(a, "Sorcie\u0300re sorcière Pe\u0301rez Matilde Pérez A\u0308\uFB03ne Äffine ", + new String[] { "Sorciere","sorciere","Perez", "Matilde", "Perez", "Affine", "Affine" }); + } + + private void checkToken(Transliterator transform, String input, String expected) throws IOException { + TokenStream ts = new ICUTransformFilter(new KeywordTokenizer((new StringReader(input))), transform); + assertTokenStreamContents(ts, new String[] { expected }); + } +} Property changes on: contrib\icu\src\test\org\apache\lucene\icu\TestICUTransformFilter.java ___________________________________________________________________ Added: svn:eol-style + native Index: contrib/icu/src/test/org/apache/lucene/icu/tokenizer/TestCharArrayIterator.java =================================================================== --- contrib/icu/src/test/org/apache/lucene/icu/tokenizer/TestCharArrayIterator.java (revision 0) +++ contrib/icu/src/test/org/apache/lucene/icu/tokenizer/TestCharArrayIterator.java (revision 0) @@ -0,0 +1,108 @@ +package org.apache.lucene.icu.tokenizer; + +/** + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +import java.text.CharacterIterator; + +import org.apache.lucene.util.LuceneTestCase; + +public class TestCharArrayIterator extends LuceneTestCase { + public void testBasicUsage() { + CharArrayIterator ci = new CharArrayIterator(); + ci.setText("testing".toCharArray(), 0, "testing".length()); + assertEquals(0, ci.getBeginIndex()); + assertEquals(7, ci.getEndIndex()); + assertEquals(0, ci.getIndex()); + assertEquals('t', ci.current()); + assertEquals('e', ci.next()); + assertEquals('g', ci.last()); + assertEquals('n', ci.previous()); + assertEquals('t', ci.first()); + assertEquals(CharacterIterator.DONE, ci.previous()); + } + + public void testFirst() { + CharArrayIterator ci = new CharArrayIterator(); + ci.setText("testing".toCharArray(), 0, "testing".length()); + ci.next(); + // Sets the position to getBeginIndex() and returns the character at that position. + assertEquals('t', ci.first()); + assertEquals(ci.getBeginIndex(), ci.getIndex()); + // or DONE if the text is empty + ci.setText(new char[] {}, 0, 0); + assertEquals(CharacterIterator.DONE, ci.first()); + } + + public void testLast() { + CharArrayIterator ci = new CharArrayIterator(); + ci.setText("testing".toCharArray(), 0, "testing".length()); + // Sets the position to getEndIndex()-1 (getEndIndex() if the text is empty) + // and returns the character at that position. + assertEquals('g', ci.last()); + assertEquals(ci.getIndex(), ci.getEndIndex() - 1); + // or DONE if the text is empty + ci.setText(new char[] {}, 0, 0); + assertEquals(CharacterIterator.DONE, ci.last()); + assertEquals(ci.getEndIndex(), ci.getIndex()); + } + + public void testCurrent() { + CharArrayIterator ci = new CharArrayIterator(); + // Gets the character at the current position (as returned by getIndex()). + ci.setText("testing".toCharArray(), 0, "testing".length()); + assertEquals('t', ci.current()); + ci.last(); + ci.next(); + // or DONE if the current position is off the end of the text. + assertEquals(CharacterIterator.DONE, ci.current()); + } + + public void testNext() { + CharArrayIterator ci = new CharArrayIterator(); + ci.setText("te".toCharArray(), 0, 2); + // Increments the iterator's index by one and returns the character at the new index. + assertEquals('e', ci.next()); + assertEquals(1, ci.getIndex()); + // or DONE if the new position is off the end of the text range. + assertEquals(CharacterIterator.DONE, ci.next()); + assertEquals(ci.getEndIndex(), ci.getIndex()); + } + + public void testSetIndex() { + CharArrayIterator ci = new CharArrayIterator(); + ci.setText("test".toCharArray(), 0, "test".length()); + try { + ci.setIndex(5); + fail(); + } catch (Exception e) { + assertTrue(e instanceof IllegalArgumentException); + } + } + + public void testClone() { + char text[] = "testing".toCharArray(); + CharArrayIterator ci = new CharArrayIterator(); + ci.next(); + CharArrayIterator ci2 = (CharArrayIterator) ci.clone(); + assertEquals(ci.getIndex(), ci2.getIndex()); + assertEquals(ci.next(), ci2.next()); + assertEquals(ci.last(), ci2.last()); + } + + +} Property changes on: contrib\icu\src\test\org\apache\lucene\icu\tokenizer\TestCharArrayIterator.java ___________________________________________________________________ Added: svn:eol-style + native Index: contrib/icu/src/test/org/apache/lucene/icu/tokenizer/TestICUTokenizer.java =================================================================== --- contrib/icu/src/test/org/apache/lucene/icu/tokenizer/TestICUTokenizer.java (revision 0) +++ contrib/icu/src/test/org/apache/lucene/icu/tokenizer/TestICUTokenizer.java (revision 0) @@ -0,0 +1,58 @@ +package org.apache.lucene.icu.tokenizer; + +/** + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +import java.io.IOException; +import java.io.StringReader; + +import org.apache.lucene.analysis.BaseTokenStreamTestCase; + +import java.util.Arrays; + +public class TestICUTokenizer extends BaseTokenStreamTestCase { + + public void testHugeDoc() throws IOException { + StringBuilder sb = new StringBuilder(); + char whitespace[] = new char[4094]; + Arrays.fill(whitespace, ' '); + sb.append(whitespace); + sb.append("testing 1234"); + String input = sb.toString(); + ICUTokenizer tokenizer = new ICUTokenizer(new StringReader(input)); + assertTokenStreamContents(tokenizer, new String[] { "testing", "1234" }); + } + + public void testHugeTerm2() throws IOException { + StringBuilder sb = new StringBuilder(); + for (int i = 0; i < 40960; i++) { + sb.append('a'); + } + String input = sb.toString(); + ICUTokenizer tokenizer = new ICUTokenizer(new StringReader(input)); + char token[] = new char[4096]; + Arrays.fill(token, 'a'); + String expectedToken = new String(token); + String expected[] = { + expectedToken, expectedToken, expectedToken, + expectedToken, expectedToken, expectedToken, + expectedToken, expectedToken, expectedToken, + expectedToken + }; + assertTokenStreamContents(tokenizer, expected); + } +} Property changes on: contrib\icu\src\test\org\apache\lucene\icu\tokenizer\TestICUTokenizer.java ___________________________________________________________________ Added: svn:eol-style + native Index: contrib/icu/src/test/org/apache/lucene/icu/tokenizer/TestLaoBreakIterator.java =================================================================== --- contrib/icu/src/test/org/apache/lucene/icu/tokenizer/TestLaoBreakIterator.java (revision 0) +++ contrib/icu/src/test/org/apache/lucene/icu/tokenizer/TestLaoBreakIterator.java (revision 0) @@ -0,0 +1,77 @@ +package org.apache.lucene.icu.tokenizer; + +/** + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +import java.io.InputStream; + +import org.apache.lucene.util.LuceneTestCase; + +import com.ibm.icu.lang.UCharacter; +import com.ibm.icu.text.BreakIterator; +import com.ibm.icu.text.RuleBasedBreakIterator; +import com.ibm.icu.text.UTF16; + +public class TestLaoBreakIterator extends LuceneTestCase { + private LaoBreakIterator wordIterator; + + protected void setUp() throws Exception { + super.setUp(); + InputStream is = TestLaoBreakIterator.class.getResourceAsStream("Lao.brk"); + wordIterator = new LaoBreakIterator(RuleBasedBreakIterator.getInstanceFromCompiledRules(is)); + is.close(); + } + + protected void assertBreaksTo(BreakIterator iterator, String sourceText, String tokens[]) { + char text[] = sourceText.toCharArray(); + CharArrayIterator ci = new CharArrayIterator(); + ci.setText(text, 0, text.length); + iterator.setText(ci); + + for (int i = 0; i < tokens.length; i++) { + int start, end; + do { + start = iterator.current(); + end = iterator.next(); + } while (end != BreakIterator.DONE && !isWord(text, start, end)); + assertTrue(start != BreakIterator.DONE); + assertTrue(end != BreakIterator.DONE); + assertEquals(tokens[i], new String(text, start, end - start)); + } + + assertTrue(iterator.next() == BreakIterator.DONE); + } + + protected boolean isWord(char text[], int start, int end) { + int codepoint; + for (int i = start; i < end; i += UTF16.getCharCount(codepoint)) { + codepoint = UTF16.charAt(text, 0, end, start); + + if (UCharacter.isLetterOrDigit(codepoint)) + return true; + } + + return false; + } + + public void testBasicUsage() throws Exception { + assertBreaksTo(wordIterator, "ກວ່າດອກ", new String[] { "ກວ່າ", "ດອກ" }); + assertBreaksTo(wordIterator, "ຜູ້ເຂົ້າ", new String[] { "ຜູ້", "ເຂົ້າ" }); + assertBreaksTo(wordIterator, "", new String[] {}); + } + +} Property changes on: contrib\icu\src\test\org\apache\lucene\icu\tokenizer\TestLaoBreakIterator.java ___________________________________________________________________ Added: svn:eol-style + native Index: contrib/icu/src/tools/java/com/ibm/icu/text/RBBIRuleBuilderTunnel.java =================================================================== --- contrib/icu/src/tools/java/com/ibm/icu/text/RBBIRuleBuilderTunnel.java (revision 0) +++ contrib/icu/src/tools/java/com/ibm/icu/text/RBBIRuleBuilderTunnel.java (revision 0) @@ -0,0 +1,27 @@ +package com.ibm.icu.text; + +/** + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +import java.io.IOException; +import java.io.OutputStream; + +public class RBBIRuleBuilderTunnel { + public static void compileRules(String rules, OutputStream os) throws IOException { + RBBIRuleBuilder.compileRules(rules, os); + } +} Property changes on: contrib\icu\src\tools\java\com\ibm\icu\text\RBBIRuleBuilderTunnel.java ___________________________________________________________________ Added: svn:eol-style + native Index: contrib/icu/src/tools/java/org/apache/lucene/icu/RBBIRuleCompiler.java =================================================================== --- contrib/icu/src/tools/java/org/apache/lucene/icu/RBBIRuleCompiler.java (revision 0) +++ contrib/icu/src/tools/java/org/apache/lucene/icu/RBBIRuleCompiler.java (revision 0) @@ -0,0 +1,89 @@ +package org.apache.lucene.icu; + +import java.io.BufferedReader; +import java.io.File; +import java.io.FileInputStream; +import java.io.FileOutputStream; +import java.io.FilenameFilter; +import java.io.IOException; +import java.io.InputStream; +import java.io.InputStreamReader; + +import com.ibm.icu.text.RBBIRuleBuilderTunnel; +import com.ibm.icu.text.RuleBasedBreakIterator; + +/** + * Licensed to the Apache Software Foundation (ASF) under one or more + * contributor license agreements. See the NOTICE file distributed with + * this work for additional information regarding copyright ownership. + * The ASF licenses this file to You under the Apache License, Version 2.0 + * (the "License"); you may not use this file except in compliance with + * the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +/** + * + * + */ +public class RBBIRuleCompiler { + + static String getRules(File ruleFile) throws IOException { + StringBuffer rules = new StringBuffer(); + InputStream in = new FileInputStream(ruleFile); + BufferedReader cin = new BufferedReader(new InputStreamReader(in, "UTF-8")); + String line = null; + while ((line = cin.readLine()) != null) { + if (!line.startsWith("#")) + rules.append(line); + rules.append('\n'); + } + cin.close(); + in.close(); + return rules.toString(); + } + + static void compile(File file) throws IOException { + if (file.isDirectory()) { + File files[] = file.listFiles(new FilenameFilter() { + public boolean accept(File dir, String name) { + return name.endsWith("rbbi") || new File(dir, name).isDirectory(); + }}); + for (int i = 0; i < files.length; i++) + compile(files[i]); + } else { + File outputFile = new File(file.getParent(), file.getName().replaceAll("rbbi$", "brk")); + String rules = getRules(file); + System.err.println("Compiling " + file.getName() + " to " + outputFile.getName()); + /* + * if there is a syntax error, compileRules() may succeed. + * the way to check is to try to instantiate from the string. + * additionally if the rules are invalid, you can get a useful syntax error. + */ + try { + new RuleBasedBreakIterator(rules); + } catch (IllegalArgumentException e) { + /* do this intentionally, so you don't get a massive stack trace + * instead, get a useful syntax error! + */ + System.err.println(e.getMessage()); + System.exit(1); + } + FileOutputStream os = new FileOutputStream(outputFile); + RBBIRuleBuilderTunnel.compileRules(rules, os); + os.close(); + } + } + + public static void main(String args[]) throws Exception { + compile(new File(".")); + System.exit(0); + } +} Property changes on: contrib\icu\src\tools\java\org\apache\lucene\icu\RBBIRuleCompiler.java ___________________________________________________________________ Added: svn:eol-style + native