Normalizer2.java

// © 2016 and later: Unicode, Inc. and others.
// License & terms of use: http://www.unicode.org/copyright.html
/*
 *******************************************************************************
 *   Copyright (C) 2009-2016, International Business Machines
 *   Corporation and others.  All Rights Reserved.
 *******************************************************************************
 */

package com.ibm.icu.text;

import com.ibm.icu.impl.ICUBinary;
import com.ibm.icu.impl.Norm2AllModes;
import com.ibm.icu.util.ICUUncheckedIOException;
import java.io.IOException;
import java.io.InputStream;
import java.nio.ByteBuffer;

/**
 * Unicode normalization functionality for standard Unicode normalization or for using custom
 * mapping tables. All instances of this class are unmodifiable/immutable. The Normalizer2 class is
 * not intended for public subclassing.
 *
 * <p>The primary functions are to produce a normalized string and to detect whether a string is
 * already normalized. The most commonly used normalization forms are those defined in
 * https://www.unicode.org/reports/tr15/ However, this API supports additional normalization forms
 * for specialized purposes. For example, NFKC_Casefold is provided via getInstance("nfkc_cf",
 * COMPOSE) and can be used in implementations of UTS #46.
 *
 * <p>Not only are the standard compose and decompose modes supplied, but additional modes are
 * provided as documented in the Mode enum.
 *
 * <p>Some of the functions in this class identify normalization boundaries. At a normalization
 * boundary, the portions of the string before it and starting from it do not interact and can be
 * handled independently.
 *
 * <p>The spanQuickCheckYes() stops at a normalization boundary. When the goal is a normalized
 * string, then the text before the boundary can be copied, and the remainder can be processed with
 * normalizeSecondAndAppend().
 *
 * <p>The hasBoundaryBefore(), hasBoundaryAfter() and isInert() functions test whether a character
 * is guaranteed to be at a normalization boundary, regardless of context. This is used for moving
 * from one normalization boundary to the next or preceding boundary, and for performing iterative
 * normalization.
 *
 * <p>Iterative normalization is useful when only a small portion of a longer string needs to be
 * processed. For example, in ICU, iterative normalization is used by the
 * NormalizationTransliterator (to avoid replacing already-normalized text) and
 * ucol_nextSortKeyPart() (to process only the substring for which sort key bytes are computed).
 *
 * <p>The set of normalization boundaries returned by these functions may not be complete: There may
 * be more boundaries that could be returned. Different functions may return different boundaries.
 *
 * @stable ICU 4.4
 * @author Markus W. Scherer
 */
public abstract class Normalizer2 {
    /**
     * Constants for normalization modes. For details about standard Unicode normalization forms and
     * about the algorithms which are also used with custom mapping tables see
     * https://www.unicode.org/reports/tr15/
     *
     * @stable ICU 4.4
     */
    public enum Mode {
        /**
         * Decomposition followed by composition. Same as standard NFC when using an "nfc" instance.
         * Same as standard NFKC when using an "nfkc" instance. For details about standard Unicode
         * normalization forms see https://www.unicode.org/reports/tr15/
         *
         * @stable ICU 4.4
         */
        COMPOSE,
        /**
         * Map, and reorder canonically. Same as standard NFD when using an "nfc" instance. Same as
         * standard NFKD when using an "nfkc" instance. For details about standard Unicode
         * normalization forms see https://www.unicode.org/reports/tr15/
         *
         * @stable ICU 4.4
         */
        DECOMPOSE,
        /**
         * "Fast C or D" form. If a string is in this form, then further decomposition <i>without
         * reordering</i> would yield the same form as DECOMPOSE. Text in "Fast C or D" form can be
         * processed efficiently with data tables that are "canonically closed", that is, that
         * provide equivalent data for equivalent text, without having to be fully normalized.<br>
         * Not a standard Unicode normalization form.<br>
         * Not a unique form: Different FCD strings can be canonically equivalent.<br>
         * For details see http://www.unicode.org/notes/tn5/#FCD
         *
         * @stable ICU 4.4
         */
        FCD,
        /**
         * Compose only contiguously. Also known as "FCC" or "Fast C Contiguous". The result will
         * often but not always be in NFC. The result will conform to FCD which is useful for
         * processing.<br>
         * Not a standard Unicode normalization form.<br>
         * For details see http://www.unicode.org/notes/tn5/#FCC
         *
         * @stable ICU 4.4
         */
        COMPOSE_CONTIGUOUS
    };

    /**
     * Returns a Normalizer2 instance for Unicode NFC normalization. Same as getInstance(null,
     * "nfc", Mode.COMPOSE). Returns an unmodifiable singleton instance.
     *
     * @return the requested Normalizer2, if successful
     * @stable ICU 49
     */
    public static Normalizer2 getNFCInstance() {
        return Norm2AllModes.getNFCInstance().comp;
    }

    /**
     * Returns a Normalizer2 instance for Unicode NFD normalization. Same as getInstance(null,
     * "nfc", Mode.DECOMPOSE). Returns an unmodifiable singleton instance.
     *
     * @return the requested Normalizer2, if successful
     * @stable ICU 49
     */
    public static Normalizer2 getNFDInstance() {
        return Norm2AllModes.getNFCInstance().decomp;
    }

    /**
     * Returns a Normalizer2 instance for Unicode NFKC normalization. Same as getInstance(null,
     * "nfkc", Mode.COMPOSE). Returns an unmodifiable singleton instance.
     *
     * @return the requested Normalizer2, if successful
     * @stable ICU 49
     */
    public static Normalizer2 getNFKCInstance() {
        return Norm2AllModes.getNFKCInstance().comp;
    }

    /**
     * Returns a Normalizer2 instance for Unicode NFKD normalization. Same as getInstance(null,
     * "nfkc", Mode.DECOMPOSE). Returns an unmodifiable singleton instance.
     *
     * @return the requested Normalizer2, if successful
     * @stable ICU 49
     */
    public static Normalizer2 getNFKDInstance() {
        return Norm2AllModes.getNFKCInstance().decomp;
    }

    /**
     * Returns a Normalizer2 instance for Unicode toNFKC_Casefold() normalization which is
     * equivalent to applying the NFKC_Casefold mappings and then NFC. See
     * https://www.unicode.org/reports/tr44/#NFKC_Casefold
     *
     * <p>Same as getInstance(null, "nfkc_cf", Mode.COMPOSE). Returns an unmodifiable singleton
     * instance.
     *
     * @return the requested Normalizer2, if successful
     * @stable ICU 49
     */
    public static Normalizer2 getNFKCCasefoldInstance() {
        return Norm2AllModes.getNFKC_CFInstance().comp;
    }

    /**
     * Returns a Normalizer2 instance for a variant of Unicode toNFKC_Casefold() normalization which
     * is equivalent to applying the NFKC_Simple_Casefold mappings and then NFC. See
     * https://www.unicode.org/reports/tr44/#NFKC_Simple_Casefold
     *
     * <p>Same as getInstance(null, "nfkc_scf", Mode.COMPOSE). Returns an unmodifiable singleton
     * instance.
     *
     * @return the requested Normalizer2, if successful
     * @stable ICU 74
     */
    public static Normalizer2 getNFKCSimpleCasefoldInstance() {
        return Norm2AllModes.getNFKC_SCFInstance().comp;
    }

    /**
     * Returns a Normalizer2 instance which uses the specified data file (an ICU data file if
     * data=null, or else custom binary data) and which composes or decomposes text according to the
     * specified mode. Returns an unmodifiable singleton instance.
     *
     * <ul>
     *   <li>Use data=null for data files that are part of ICU's own data.
     *   <li>Use name="nfc" and COMPOSE/DECOMPOSE for Unicode standard NFC/NFD.
     *   <li>Use name="nfkc" and COMPOSE/DECOMPOSE for Unicode standard NFKC/NFKD.
     *   <li>Use name="nfkc_cf" and COMPOSE for Unicode standard NFKC_CF=NFKC_Casefold.
     * </ul>
     *
     * If data!=null, then the binary data is read once and cached using the provided name as the
     * key. If you know or expect the data to be cached already, you can use data!=null for non-ICU
     * data as well.
     *
     * <p>Any {@link java.io.IOException} is wrapped into a {@link
     * com.ibm.icu.util.ICUUncheckedIOException}.
     *
     * @param data the binary, big-endian normalization (.nrm file) data, or null for ICU data
     * @param name "nfc" or "nfkc" or "nfkc_cf" or "nfkc_scf" or name of custom data file
     * @param mode normalization mode (compose or decompose etc.)
     * @return the requested Normalizer2, if successful
     * @stable ICU 4.4
     */
    public static Normalizer2 getInstance(InputStream data, String name, Mode mode) {
        // TODO: If callers really use this API, then we should add an overload that takes a
        // ByteBuffer.
        ByteBuffer bytes = null;
        if (data != null) {
            try {
                bytes = ICUBinary.getByteBufferFromInputStreamAndCloseStream(data);
            } catch (IOException e) {
                throw new ICUUncheckedIOException(e);
            }
        }
        Norm2AllModes all2Modes = Norm2AllModes.getInstance(bytes, name);
        switch (mode) {
            case COMPOSE:
                return all2Modes.comp;
            case DECOMPOSE:
                return all2Modes.decomp;
            case FCD:
                return all2Modes.fcd;
            case COMPOSE_CONTIGUOUS:
                return all2Modes.fcc;
            default:
                return null; // will not occur
        }
    }

    /**
     * Returns the normalized form of the source string.
     *
     * @param src source string
     * @return normalized src
     * @stable ICU 4.4
     */
    public String normalize(CharSequence src) {
        if (src instanceof String) {
            // Fastpath: Do not construct a new String if the src is a String
            // and is already normalized.
            int spanLength = spanQuickCheckYes(src);
            if (spanLength == src.length()) {
                return (String) src;
            }
            if (spanLength != 0) {
                StringBuilder sb = new StringBuilder(src.length()).append(src, 0, spanLength);
                return normalizeSecondAndAppend(sb, src.subSequence(spanLength, src.length()))
                        .toString();
            }
        }
        return normalize(src, new StringBuilder(src.length())).toString();
    }

    /**
     * Writes the normalized form of the source string to the destination string (replacing its
     * contents) and returns the destination string. The source and destination strings must be
     * different objects.
     *
     * @param src source string
     * @param dest destination string; its contents is replaced with normalized src
     * @return dest
     * @stable ICU 4.4
     */
    public abstract StringBuilder normalize(CharSequence src, StringBuilder dest);

    /**
     * Writes the normalized form of the source string to the destination Appendable and returns the
     * destination Appendable. The source and destination strings must be different objects.
     *
     * <p>Any {@link java.io.IOException} is wrapped into a {@link
     * com.ibm.icu.util.ICUUncheckedIOException}.
     *
     * @param src source string
     * @param dest destination Appendable; gets normalized src appended
     * @return dest
     * @stable ICU 4.6
     */
    public abstract Appendable normalize(CharSequence src, Appendable dest);

    /**
     * Appends the normalized form of the second string to the first string (merging them at the
     * boundary) and returns the first string. The result is normalized if the first string was
     * normalized. The first and second strings must be different objects.
     *
     * @param first string, should be normalized
     * @param second string, will be normalized
     * @return first
     * @stable ICU 4.4
     */
    public abstract StringBuilder normalizeSecondAndAppend(
            StringBuilder first, CharSequence second);

    /**
     * Appends the second string to the first string (merging them at the boundary) and returns the
     * first string. The result is normalized if both the strings were normalized. The first and
     * second strings must be different objects.
     *
     * @param first string, should be normalized
     * @param second string, should be normalized
     * @return first
     * @stable ICU 4.4
     */
    public abstract StringBuilder append(StringBuilder first, CharSequence second);

    /**
     * Gets the decomposition mapping of c. Roughly equivalent to normalizing the String form of c
     * on a DECOMPOSE Normalizer2 instance, but much faster, and except that this function returns
     * null if c does not have a decomposition mapping in this instance's data. This function is
     * independent of the mode of the Normalizer2.
     *
     * @param c code point
     * @return c's decomposition mapping, if any; otherwise null
     * @stable ICU 4.6
     */
    public abstract String getDecomposition(int c);

    /**
     * Gets the raw decomposition mapping of c.
     *
     * <p>This is similar to the getDecomposition() method but returns the raw decomposition mapping
     * as specified in UnicodeData.txt or (for custom data) in the mapping files processed by the
     * gennorm2 tool. By contrast, getDecomposition() returns the processed, recursively-decomposed
     * version of this mapping.
     *
     * <p>When used on a standard NFKC Normalizer2 instance, getRawDecomposition() returns the
     * Unicode Decomposition_Mapping (dm) property.
     *
     * <p>When used on a standard NFC Normalizer2 instance, it returns the Decomposition_Mapping
     * only if the Decomposition_Type (dt) is Canonical (Can); in this case, the result contains
     * either one or two code points (=1..4 Java chars).
     *
     * <p>This function is independent of the mode of the Normalizer2. The default implementation
     * returns null.
     *
     * @param c code point
     * @return c's raw decomposition mapping, if any; otherwise null
     * @stable ICU 49
     */
    public String getRawDecomposition(int c) {
        return null;
    }

    /**
     * Performs pairwise composition of a &amp; b and returns the composite if there is one.
     *
     * <p>Returns a composite code point c only if c has a two-way mapping to a+b. In standard
     * Unicode normalization, this means that c has a canonical decomposition to a+b and c does not
     * have the Full_Composition_Exclusion property.
     *
     * <p>This function is independent of the mode of the Normalizer2. The default implementation
     * returns a negative value.
     *
     * @param a A (normalization starter) code point.
     * @param b Another code point.
     * @return The non-negative composite code point if there is one; otherwise a negative value.
     * @stable ICU 49
     */
    public int composePair(int a, int b) {
        return -1;
    }

    /**
     * Gets the combining class of c. The default implementation returns 0 but all standard
     * implementations return the Unicode Canonical_Combining_Class value.
     *
     * @param c code point
     * @return c's combining class
     * @stable ICU 49
     */
    public int getCombiningClass(int c) {
        return 0;
    }

    /**
     * Tests if the string is normalized. Internally, in cases where the quickCheck() method would
     * return "maybe" (which is only possible for the two COMPOSE modes) this method resolves to
     * "yes" or "no" to provide a definitive result, at the cost of doing more work in those cases.
     *
     * @param s input string
     * @return true if s is normalized
     * @stable ICU 4.4
     */
    public abstract boolean isNormalized(CharSequence s);

    /**
     * Tests if the string is normalized. For the two COMPOSE modes, the result could be "maybe" in
     * cases that would take a little more work to resolve definitively. Use spanQuickCheckYes() and
     * normalizeSecondAndAppend() for a faster combination of quick check + normalization, to avoid
     * re-checking the "yes" prefix.
     *
     * @param s input string
     * @return the quick check result
     * @stable ICU 4.4
     */
    public abstract Normalizer.QuickCheckResult quickCheck(CharSequence s);

    /**
     * Returns the end of the normalized substring of the input string. In other words, with <code>
     * end=spanQuickCheckYes(s);</code> the substring <code>s.subSequence(0, end)</code> will pass
     * the quick check with a "yes" result.
     *
     * <p>The returned end index is usually one or more characters before the "no" or "maybe"
     * character: The end index is at a normalization boundary. (See the class documentation for
     * more about normalization boundaries.)
     *
     * <p>When the goal is a normalized string and most input strings are expected to be normalized
     * already, then call this method, and if it returns a prefix shorter than the input string,
     * copy that prefix and use normalizeSecondAndAppend() for the remainder.
     *
     * @param s input string
     * @return "yes" span end index
     * @stable ICU 4.4
     */
    public abstract int spanQuickCheckYes(CharSequence s);

    /**
     * Tests if the character always has a normalization boundary before it, regardless of context.
     * If true, then the character does not normalization-interact with preceding characters. In
     * other words, a string containing this character can be normalized by processing portions
     * before this character and starting from this character independently. This is used for
     * iterative normalization. See the class documentation for details.
     *
     * @param c character to test
     * @return true if c has a normalization boundary before it
     * @stable ICU 4.4
     */
    public abstract boolean hasBoundaryBefore(int c);

    /**
     * Tests if the character always has a normalization boundary after it, regardless of context.
     * If true, then the character does not normalization-interact with following characters. In
     * other words, a string containing this character can be normalized by processing portions up
     * to this character and after this character independently. This is used for iterative
     * normalization. See the class documentation for details.
     *
     * <p>Note that this operation may be significantly slower than hasBoundaryBefore().
     *
     * @param c character to test
     * @return true if c has a normalization boundary after it
     * @stable ICU 4.4
     */
    public abstract boolean hasBoundaryAfter(int c);

    /**
     * Tests if the character is normalization-inert. If true, then the character does not change,
     * nor normalization-interact with preceding or following characters. In other words, a string
     * containing this character can be normalized by processing portions before this character and
     * after this character independently. This is used for iterative normalization. See the class
     * documentation for details.
     *
     * <p>Note that this operation may be significantly slower than hasBoundaryBefore().
     *
     * @param c character to test
     * @return true if c is normalization-inert
     * @stable ICU 4.4
     */
    public abstract boolean isInert(int c);

    /**
     * Sole constructor. (For invocation by subclass constructors, typically implicit.)
     *
     * @internal
     * @deprecated This API is ICU internal only.
     */
    @Deprecated
    protected Normalizer2() {}
}