Home | History | Annotate | Download | only in impl
      1 /* GENERATED SOURCE. DO NOT MODIFY. */
      2 //  2016 and later: Unicode, Inc. and others.
      3 // License & terms of use: http://www.unicode.org/copyright.html#License
      4 /*
      5  *******************************************************************************
      6  * Copyright (C) 2009-2014, International Business Machines Corporation and
      7  * others. All Rights Reserved.
      8  *******************************************************************************
      9  */
     10 
     11 package android.icu.impl;
     12 
     13 import java.io.DataOutputStream;
     14 import java.io.IOException;
     15 import java.io.OutputStream;
     16 import java.nio.ByteBuffer;
     17 
     18 /**
     19  * @author aheninger
     20  *
     21  * A read-only Trie2, holding 32 bit data values.
     22  *
     23  * A Trie2 is a highly optimized data structure for mapping from Unicode
     24  * code points (values ranging from 0 to 0x10ffff) to a 16 or 32 bit value.
     25  *
     26  * See class Trie2 for descriptions of the API for accessing the contents of a trie.
     27  *
     28  * The fundamental data access methods are declared final in this class, with
     29  * the intent that applications might gain a little extra performance, when compared
     30  * with calling the same methods via the abstract UTrie2 base class.
     31  * @hide Only a subset of ICU is exposed in Android
     32  */
     33 
     34 public class Trie2_32 extends Trie2 {
     35 
     36     /**
     37      * Internal constructor, not for general use.
     38      */
     39     Trie2_32() {
     40     }
     41 
     42 
     43     /**
     44      * Create a Trie2 from its serialized form.  Inverse of utrie2_serialize().
     45      * The serialized format is identical between ICU4C and ICU4J, so this function
     46      * will work with serialized Trie2s from either.
     47      *
     48      * The serialized Trie2 in the bytes may be in either little or big endian byte order.
     49      * This allows using serialized Tries from ICU4C without needing to consider the
     50      * byte order of the system that created them.
     51      *
     52      * @param bytes a byte buffer to the serialized form of a UTrie2.
     53      * @return An unserialized Trie_32, ready for use.
     54      * @throws IllegalArgumentException if the stream does not contain a serialized Trie2.
     55      * @throws IOException if a read error occurs in the buffer.
     56      * @throws ClassCastException if the bytes contains a serialized Trie2_16
     57      */
     58     public static Trie2_32 createFromSerialized(ByteBuffer bytes) throws IOException {
     59         return (Trie2_32) Trie2.createFromSerialized(bytes);
     60     }
     61 
     62     /**
     63      * Get the value for a code point as stored in the Trie2.
     64      *
     65      * @param codePoint the code point
     66      * @return the value
     67      */
     68     @Override
     69     public final int get(int codePoint) {
     70         int value;
     71         int ix;
     72 
     73         if (codePoint >= 0) {
     74             if (codePoint < 0x0d800 || (codePoint > 0x0dbff && codePoint <= 0x0ffff)) {
     75                 // Ordinary BMP code point, excluding leading surrogates.
     76                 // BMP uses a single level lookup.  BMP index starts at offset 0 in the Trie2 index.
     77                 // 32 bit data is stored in the index array itself.
     78                 ix = index[codePoint >> UTRIE2_SHIFT_2];
     79                 ix = (ix << UTRIE2_INDEX_SHIFT) + (codePoint & UTRIE2_DATA_MASK);
     80                 value = data32[ix];
     81                 return value;
     82             }
     83             if (codePoint <= 0xffff) {
     84                 // Lead Surrogate Code Point.  A Separate index section is stored for
     85                 // lead surrogate code units and code points.
     86                 //   The main index has the code unit data.
     87                 //   For this function, we need the code point data.
     88                 // Note: this expression could be refactored for slightly improved efficiency, but
     89                 //       surrogate code points will be so rare in practice that it's not worth it.
     90                 ix = index[UTRIE2_LSCP_INDEX_2_OFFSET + ((codePoint - 0xd800) >> UTRIE2_SHIFT_2)];
     91                 ix = (ix << UTRIE2_INDEX_SHIFT) + (codePoint & UTRIE2_DATA_MASK);
     92                 value = data32[ix];
     93                 return value;
     94             }
     95             if (codePoint < highStart) {
     96                 // Supplemental code point, use two-level lookup.
     97                 ix = (UTRIE2_INDEX_1_OFFSET - UTRIE2_OMITTED_BMP_INDEX_1_LENGTH) + (codePoint >> UTRIE2_SHIFT_1);
     98                 ix = index[ix];
     99                 ix += (codePoint >> UTRIE2_SHIFT_2) & UTRIE2_INDEX_2_MASK;
    100                 ix = index[ix];
    101                 ix = (ix << UTRIE2_INDEX_SHIFT) + (codePoint & UTRIE2_DATA_MASK);
    102                 value = data32[ix];
    103                 return value;
    104             }
    105             if (codePoint <= 0x10ffff) {
    106                 value = data32[highValueIndex];
    107                 return value;
    108             }
    109         }
    110 
    111         // Fall through.  The code point is outside of the legal range of 0..0x10ffff.
    112         return errorValue;
    113     }
    114 
    115 
    116     /**
    117      * Get a Trie2 value for a UTF-16 code unit.
    118      *
    119      * This function returns the same value as get() if the input
    120      * character is outside of the lead surrogate range
    121      *
    122      * There are two values stored in a Trie2 for inputs in the lead
    123      * surrogate range.  This function returns the alternate value,
    124      * while Trie2.get() returns the main value.
    125      *
    126      * @param codeUnit a 16 bit code unit or lead surrogate value.
    127      * @return the value
    128      */
    129     @Override
    130     public int getFromU16SingleLead(char codeUnit){
    131         int value;
    132         int ix;
    133 
    134         ix = index[codeUnit >> UTRIE2_SHIFT_2];
    135         ix = (ix << UTRIE2_INDEX_SHIFT) + (codeUnit & UTRIE2_DATA_MASK);
    136         value = data32[ix];
    137         return value;
    138 
    139     }
    140 
    141     /**
    142      * Serialize a Trie2_32 onto an OutputStream.
    143      *
    144      * A Trie2 can be serialized multiple times.
    145      * The serialized data is compatible with ICU4C UTrie2 serialization.
    146      * Trie2 serialization is unrelated to Java object serialization.
    147      *
    148      * @param os the stream to which the serialized Trie2 data will be written.
    149      * @return the number of bytes written.
    150      * @throw IOException on an error writing to the OutputStream.
    151      */
    152     public int serialize(OutputStream os) throws IOException {
    153         DataOutputStream dos = new DataOutputStream(os);
    154         int  bytesWritten = 0;
    155 
    156         bytesWritten += serializeHeader(dos);
    157         for (int i=0; i<dataLength; i++) {
    158             dos.writeInt(data32[i]);
    159         }
    160         bytesWritten += dataLength*4;
    161         return bytesWritten;
    162     }
    163 
    164     /**
    165      * @return the number of bytes of the serialized trie
    166      */
    167     public int getSerializedLength() {
    168         return 16+header.indexLength*2+dataLength*4;
    169     }
    170 
    171     /**
    172      * Given a starting code point, find the last in a range of code points,
    173      * all with the same value.
    174      *
    175      * This function is part of the implementation of iterating over the
    176      * Trie2's contents.
    177      * @param startingCP The code point at which to begin looking.
    178      * @return The last code point with the same value as the starting code point.
    179      */
    180     @Override
    181     int rangeEnd(int startingCP, int limit, int value) {
    182         int   cp = startingCP;
    183         int   block = 0;
    184         int   index2Block = 0;
    185 
    186         // Loop runs once for each of
    187         //   - a partial data block
    188         //   - a reference to the null (default) data block.
    189         //   - a reference to the index2 null block
    190 
    191       outerLoop:
    192         for (;;) {
    193             if (cp >= limit) {
    194                 break;
    195             }
    196             if (cp < 0x0d800 || (cp > 0x0dbff && cp <= 0x0ffff)) {
    197                 // Ordinary BMP code point, excluding leading surrogates.
    198                 // BMP uses a single level lookup.  BMP index starts at offset 0 in the Trie2 index.
    199                 // 16 bit data is stored in the index array itself.
    200                 index2Block = 0;
    201                 block       = index[cp >> UTRIE2_SHIFT_2] << UTRIE2_INDEX_SHIFT;
    202             } else if (cp < 0xffff) {
    203                 // Lead Surrogate Code Point, 0xd800 <= cp < 0xdc00
    204                 index2Block = UTRIE2_LSCP_INDEX_2_OFFSET;
    205                 block       = index[index2Block + ((cp - 0xd800) >> UTRIE2_SHIFT_2)] << UTRIE2_INDEX_SHIFT;
    206             } else if (cp < highStart) {
    207                 // Supplemental code point, use two-level lookup.
    208                 int ix = (UTRIE2_INDEX_1_OFFSET - UTRIE2_OMITTED_BMP_INDEX_1_LENGTH) + (cp >> UTRIE2_SHIFT_1);
    209                 index2Block = index[ix];
    210                 block = index[index2Block + ((cp >> UTRIE2_SHIFT_2) & UTRIE2_INDEX_2_MASK)] << UTRIE2_INDEX_SHIFT;
    211             } else  {
    212                 // Code point above highStart.
    213                 if (value == data32[highValueIndex]) {
    214                     cp = limit;
    215                 }
    216                 break;
    217             }
    218 
    219             if (index2Block == index2NullOffset) {
    220                 if (value != initialValue) {
    221                     break;
    222                 }
    223                 cp += UTRIE2_CP_PER_INDEX_1_ENTRY;
    224             } else if (block == dataNullOffset) {
    225                 // The block at dataNullOffset has all values == initialValue.
    226                 // Because Trie2 iteration always proceeds in ascending order, we will always
    227                 //   encounter a null block at its beginning, and can skip over
    228                 //   a number of code points equal to the length of the block.
    229                 if (value != initialValue) {
    230                     break;
    231                 }
    232                 cp += UTRIE2_DATA_BLOCK_LENGTH;
    233             } else {
    234                 // Current position refers to an ordinary data block.
    235                 // Walk over the data entries, checking the values.
    236                 int startIx = block + (cp & UTRIE2_DATA_MASK);
    237                 int limitIx = block + UTRIE2_DATA_BLOCK_LENGTH;
    238                 for (int ix = startIx; ix<limitIx; ix++) {
    239                     if (data32[ix] != value) {
    240                         // We came to an entry with a different value.
    241                         //   We are done.
    242                         cp += (ix - startIx);
    243                         break outerLoop;
    244                     }
    245                 }
    246                 // The ordinary data block contained our value until its end.
    247                 //  Advance the current code point, and continue the outer loop.
    248                 cp += limitIx - startIx;
    249             }
    250         }
    251         if (cp > limit) {
    252             cp = limit;
    253         }
    254 
    255         return cp - 1;
    256     }
    257 
    258 }
    259 
    260