magibney commented on a change in pull request #892: LUCENE-8972: Add ICUTransformCharFilter, to support pre-tokenizer ICU text transformation URL: https://github.com/apache/lucene-solr/pull/892#discussion_r341304654
########## File path: lucene/analysis/icu/src/java/org/apache/lucene/analysis/icu/ICUTransformCharFilter.java ########## @@ -0,0 +1,384 @@ +/* + * 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. + */ + +package org.apache.lucene.analysis.icu; + +import java.io.IOException; +import java.io.Reader; + +import com.ibm.icu.text.ReplaceableString; +import com.ibm.icu.text.Transliterator; +import com.ibm.icu.text.Transliterator.Position; +import com.ibm.icu.text.UTF16; + +import org.apache.lucene.analysis.CharFilter; +import org.apache.lucene.analysis.charfilter.BaseCharFilter; +import org.apache.lucene.util.ArrayUtil; + +/** + * A {@link CharFilter} that transforms text with ICU. + * <p> + * 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. + * </p> + * <p> + * Some useful transformations for search are built-in: + * <ul> + * <li>Conversion from Traditional to Simplified Chinese characters + * <li>Conversion from Hiragana to Katakana + * <li>Conversion from Fullwidth to Halfwidth forms. + * <li>Script conversions, for example Serbian Cyrillic to Latin + * </ul> + * <p> + * Example usage: <blockquote>stream = new ICUTransformCharFilter(reader, + * Transliterator.getInstance("Traditional-Simplified"));</blockquote> + * <br> + * For more details, see the <a + * href="http://userguide.icu-project.org/transforms/general">ICU User + * Guide</a>. + */ +public final class ICUTransformCharFilter extends BaseCharFilter { + + // Transliterator to transform the text + private final Transliterator transform; + + // Reusable position object + private final Position position = new Position(); + + private static final int READ_BUFFER_SIZE = 1024; + private final char[] tmpBuffer = new char[READ_BUFFER_SIZE]; + + private static final int INITIAL_TRANSLITERATE_BUFFER_SIZE = 1024; + private final StringBuffer buffer = new StringBuffer(INITIAL_TRANSLITERATE_BUFFER_SIZE); + private final ReplaceableString replaceable = new ReplaceableString(buffer); + + private static final int BUFFER_PRUNE_THRESHOLD = 1024; + + private int outputCursor = 0; + private boolean inputFinished = false; + private int charCount = 0; + private int offsetDiffAdjust = 0; + + private static final int HARD_MAX_ROLLBACK_BUFFER_CAPACITY = Integer.highestOneBit(Integer.MAX_VALUE); + static final int DEFAULT_MAX_ROLLBACK_BUFFER_CAPACITY = 8192; + private final int maxRollbackBufferCapacity; + + private static final int DEFAULT_INITIAL_ROLLBACK_BUFFER_CAPACITY = 4; // must be power of 2 + private char[] rollbackBuffer; + private int rollbackBufferSize = 0; + + ICUTransformCharFilter(Reader in, Transliterator transform) { + this(in, transform, DEFAULT_MAX_ROLLBACK_BUFFER_CAPACITY); + } + + /** + * Construct new {@link ICUTransformCharFilter} with the specified {@link Transliterator}, backed by + * the specified {@link Reader}. + * @param in input source + * @param transform used to perform transliteration + * @param maxRollbackBufferCapacityHint used to control the maximum size to which this + * {@link ICUTransformCharFilter} will buffer and rollback partial transliteration of input sequences. + * The provided hint will be converted to an enforced limit of "the greatest power of 2 (excluding '1') + * less than or equal to the specified value". It is illegal to specify a negative value. There is no + * power of 2 greater than <code>Integer.highestOneBit(Integer.MAX_VALUE))</code>, so to prevent overflow, values + * in this range will resolve to an enforced limit of <code>Integer.highestOneBit(Integer.MAX_VALUE))</code>. + * Specifying "0" (or "1", in practice) disables rollback. Larger values can in some cases yield more accurate + * transliteration, at the cost of performance and resolution/accuracy of offset correction. + * This is intended primarily as a failsafe, with a relatively large default value of {@value ICUTransformCharFilter#DEFAULT_MAX_ROLLBACK_BUFFER_CAPACITY}. + * See comments "To understand the need for rollback" in private method: + * {@link Transliterator#filteredTransliterate(com.ibm.icu.text.Replaceable, Position, boolean, boolean)} + */ + ICUTransformCharFilter(Reader in, Transliterator transform, int maxRollbackBufferCapacityHint) { + super(in); + this.transform = ICUTransformFilter.optimizeForCommonCase(transform); + if (maxRollbackBufferCapacityHint < 0) { + throw new IllegalArgumentException("It is illegal to request negative rollback buffer max capacity"); + } else if (maxRollbackBufferCapacityHint >= HARD_MAX_ROLLBACK_BUFFER_CAPACITY) { + // arg is positive, so user wants the largest possible buffer capacity limit + // we know what that is (static), protecting for overflow. + this.maxRollbackBufferCapacity = HARD_MAX_ROLLBACK_BUFFER_CAPACITY; + this.rollbackBuffer = new char[DEFAULT_INITIAL_ROLLBACK_BUFFER_CAPACITY]; + } else { + // greatest power of 2 (excluding "1") less than or equal to the specified hint + this.maxRollbackBufferCapacity = Integer.highestOneBit(maxRollbackBufferCapacityHint - 1) << 1; + if (this.maxRollbackBufferCapacity == 0) { + this.rollbackBuffer = null; + } else { + this.rollbackBuffer = new char[DEFAULT_INITIAL_ROLLBACK_BUFFER_CAPACITY]; + } + } + } + + /** + * Reads characters into a portion of an array. This method will block until some input is available, an I/O error + * occurs, or the end of the stream is reached. + * + * @param cbuf + * Destination buffer + * @param off + * Offset at which to start storing characters + * @param len + * Maximum number of characters to read + * @return The number of characters read, or -1 if the end of the stream has been reached + * @throws IOException + * If an I/O error occurs + */ + @Override + public int read(char[] cbuf, int off, int len) throws IOException { + if (off < 0) throw new IndexOutOfBoundsException("specified negative array offset"); + if (off >= cbuf.length) throw new IndexOutOfBoundsException("specified offset exceeds buffer length"); + if (len <= 0) throw new IndexOutOfBoundsException("non-positive length specified"); + if (len > cbuf.length - off) throw new IndexOutOfBoundsException("requested end array index exceeds buffer length"); + + // !inputFinished || output remains to be flushed + while (!inputFinished || position.start > outputCursor) { + if (position.start > outputCursor) { + return outputFromResultBuffer(cbuf, off, len); + } + + int resLen = readFromIoNormalizeUptoBoundary(); + if (resLen > 0) { + return outputFromResultBuffer(cbuf, off, len); + } + + if (!readInputToBuffer()) { Review comment: Yes, that's weird. This is because `readFromIoNormalizeUptoBoundary()` is the method that does the incremental transliteration, and we don't know _a priori_ how many characters are needed for an arbitrary transliterator to determine that it has an exact match. For example, think a dictionary-based transliterator that has two terms that share a very long common prefix -- we'd have to read input until we get past the shared prefix, which could be arbitrarily long ... and if it's longer than `READ_BUFFER_SIZE`, a single call to `readInputToBuffer()` would not supply enough characters. N.b., this is different from `getMaximumContextLength()`, which says nothing about how much must be provided to the _active_ input buffer. I found this surprising as well! ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org With regards, Apache Git Services --------------------------------------------------------------------- To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For additional commands, e-mail: issues-h...@lucene.apache.org