com.ibm.icu.impl.RBBIDataWrapper Maven / Gradle / Ivy
Show all versions of icu4j Show documentation
// © 2016 and later: Unicode, Inc. and others.
// License & terms of use: http://www.unicode.org/copyright.html#License
/*
*******************************************************************************
* Copyright (C) 1996-2015, International Business Machines Corporation and
* others. All Rights Reserved.
*******************************************************************************
*/
package com.ibm.icu.impl;
import java.io.DataOutputStream;
import java.io.IOException;
import java.nio.ByteBuffer;
import java.util.Arrays;
import com.ibm.icu.impl.ICUBinary.Authenticate;
import com.ibm.icu.text.RuleBasedBreakIterator;
/**
* Internal class used for Rule Based Break Iterators.
* This class provides access to the compiled break rule data, as
* it is stored in a .brk file. Refer to the file common/rbbidata.h from
* ICU4C for further details.
*/
public final class RBBIDataWrapper {
/**
* A RBBI State Transition table, the form of the data used at run time in Java.
* These can be created from stored ICU data, or built from rules.
* The structure corresponds closely to struct RBBIStateTable in ICU4C.
*/
static public class RBBIStateTable {
/**
* Number of states (rows) in this table.
*/
public int fNumStates;
/**
* Length of a table row in bytes. Note mismatch with table data, which is short[].
*/
public int fRowLen;
/**
* Option Flags for this state table.
*/
public int fFlags;
/**
* Option Flags for this state table.
*/
public int fReserved;
/**
* Linear array of next state values, accessed as short[state, char_class]
*/
public short[] fTable;
public RBBIStateTable() {
}
static RBBIStateTable get(ByteBuffer bytes, int length) throws IOException {
if (length == 0) {
return null;
}
if (length < 16) {
throw new IOException("Invalid RBBI state table length.");
}
RBBIStateTable This = new RBBIStateTable();
This.fNumStates = bytes.getInt();
This.fRowLen = bytes.getInt();
This.fFlags = bytes.getInt();
This.fReserved = bytes.getInt();
int lengthOfShorts = length - 16; // length in bytes.
This.fTable = ICUBinary.getShorts(bytes, lengthOfShorts / 2, lengthOfShorts & 1);
return This;
}
public int put(DataOutputStream bytes) throws IOException {
bytes.writeInt(fNumStates);
bytes.writeInt(fRowLen);
bytes.writeInt(fFlags);
bytes.writeInt(fReserved);
int tableLen = fRowLen * fNumStates / 2; // fRowLen is bytes.
for (int i = 0; i < tableLen; i++) {
bytes.writeShort(fTable[i]);
}
int bytesWritten = 16 + fRowLen * fNumStates; // total bytes written,
// including 16 for the header.
while (bytesWritten % 8 != 0) {
bytes.writeByte(0);
++bytesWritten;
}
return bytesWritten;
}
/**
* {@inheritDoc}
*/
@Override
public boolean equals (Object other) {
if (other == this) {
return true;
}
if (!(other instanceof RBBIStateTable)) {
return false;
}
RBBIStateTable otherST = (RBBIStateTable)other;
if (fNumStates != otherST.fNumStates) return false;
if (fRowLen != otherST.fRowLen) return false;
if (fFlags != otherST.fFlags) return false;
if (fReserved != otherST.fReserved) return false;
return Arrays.equals(fTable, otherST.fTable);
}
}
/**
* Equals helper for state tables, including null handling.
*/
static public boolean equals(RBBIStateTable left, RBBIStateTable right) {
if (left == right) {
return true;
}
if (left == null || right == null) {
return false;
}
return left.equals(right);
}
//
// These fields are the ready-to-use compiled rule data, as
// read from the file.
//
public RBBIDataHeader fHeader;
public RBBIStateTable fFTable;
public RBBIStateTable fRTable;
public Trie2 fTrie;
public String fRuleSource;
public int fStatusTable[];
public static final int DATA_FORMAT = 0x42726b20; // "Brk "
public static final int FORMAT_VERSION = 0x05000000; // 4.0.0.0
private static final class IsAcceptable implements Authenticate {
@Override
public boolean isDataVersionAcceptable(byte version[]) {
int intVersion = (version[0] << 24) + (version[1] << 16) + (version[2] << 8) + version[3];
return intVersion == FORMAT_VERSION;
}
}
private static final IsAcceptable IS_ACCEPTABLE = new IsAcceptable();
//
// Indexes to fields in the ICU4C style binary form of the RBBI Data Header
// Used by the rule compiler when flattening the data.
//
public final static int DH_SIZE = 20;
public final static int DH_MAGIC = 0;
public final static int DH_FORMATVERSION = 1;
public final static int DH_LENGTH = 2;
public final static int DH_CATCOUNT = 3;
public final static int DH_FTABLE = 4;
public final static int DH_FTABLELEN = 5;
public final static int DH_RTABLE = 6;
public final static int DH_RTABLELEN = 7;
public final static int DH_TRIE = 8;
public final static int DH_TRIELEN = 9;
public final static int DH_RULESOURCE = 10;
public final static int DH_RULESOURCELEN = 11;
public final static int DH_STATUSTABLE = 12;
public final static int DH_STATUSTABLELEN = 13;
// Index offsets to the fields in a state table row.
// Corresponds to struct RBBIStateTableRow in the C version.
//
/**
* offset to the "accepting" field in a state table row.
*/
public final static int ACCEPTING = 0;
/**
* offset to the "lookahead" field in a state table row.
*/
public final static int LOOKAHEAD = 1;
/**
* offset to the "tagIndex" field in a state table row.
*/
public final static int TAGIDX = 2;
/**
* offset to the reserved field in a state table row.
*/
public final static int RESERVED = 3;
/**
* offset to the start of the next states array in a state table row.
*/
public final static int NEXTSTATES = 4;
// Bit selectors for the "FLAGS" field of the state table header
// enum RBBIStateTableFlags in the C version.
//
public final static int RBBI_LOOKAHEAD_HARD_BREAK = 1;
public final static int RBBI_BOF_REQUIRED = 2;
/**
* Data Header. A struct-like class with the fields from the RBBI data file header.
* Not intended for public use, declared public for testing purposes only.
*/
public final static class RBBIDataHeader {
int fMagic; // == 0xbla0
byte[] fFormatVersion; // For ICU 3.4 and later.
int fLength; // Total length in bytes of this RBBI Data,
// including all sections, not just the header.
/**
* the number of character categories.
*/
public int fCatCount; // Number of character categories.
//
// Offsets and sizes of each of the subsections within the RBBI data.
// All offsets are bytes from the start of the RBBIDataHeader.
// All sizes are in bytes.
//
int fFTable; // forward state transition table.
int fFTableLen;
int fRTable; // Offset to the reverse state transition table.
int fRTableLen;
int fTrie; // Offset to Trie data for character categories
int fTrieLen;
int fRuleSource; // Offset to the source for for the break
int fRuleSourceLen; // rules. Stored UChar *.
int fStatusTable; // Offset to the table of rule status values
int fStatusTableLen;
public RBBIDataHeader() {
fMagic = 0;
fFormatVersion = new byte[4];
}
}
/**
* RBBI State Table Indexing Function. Given a state number, return the
* array index of the start of the state table row for that state.
*/
public int getRowIndex(int state){
return state * (fHeader.fCatCount + 4);
}
RBBIDataWrapper() {
}
/**
* Get an RBBIDataWrapper from an InputStream onto a pre-compiled set
* of RBBI rules.
*/
public static RBBIDataWrapper get(ByteBuffer bytes) throws IOException {
RBBIDataWrapper This = new RBBIDataWrapper();
ICUBinary.readHeader(bytes, DATA_FORMAT, IS_ACCEPTABLE);
// Read in the RBBI data header...
This.fHeader = new RBBIDataHeader();
This.fHeader.fMagic = bytes.getInt();
This.fHeader.fFormatVersion[0] = bytes.get();
This.fHeader.fFormatVersion[1] = bytes.get();
This.fHeader.fFormatVersion[2] = bytes.get();
This.fHeader.fFormatVersion[3] = bytes.get();
This.fHeader.fLength = bytes.getInt();
This.fHeader.fCatCount = bytes.getInt();
This.fHeader.fFTable = bytes.getInt();
This.fHeader.fFTableLen = bytes.getInt();
This.fHeader.fRTable = bytes.getInt();
This.fHeader.fRTableLen = bytes.getInt();
This.fHeader.fTrie = bytes.getInt();
This.fHeader.fTrieLen = bytes.getInt();
This.fHeader.fRuleSource = bytes.getInt();
This.fHeader.fRuleSourceLen = bytes.getInt();
This.fHeader.fStatusTable = bytes.getInt();
This.fHeader.fStatusTableLen = bytes.getInt();
ICUBinary.skipBytes(bytes, 6 * 4); // uint32_t fReserved[6];
if (This.fHeader.fMagic != 0xb1a0 || !IS_ACCEPTABLE.isDataVersionAcceptable(This.fHeader.fFormatVersion)) {
throw new IOException("Break Iterator Rule Data Magic Number Incorrect, or unsupported data version.");
}
// Current position in the buffer.
int pos = DH_SIZE * 4; // offset of end of header, which has DH_SIZE fields, all int32_t (4 bytes)
//
// Read in the Forward state transition table as an array of shorts.
//
// Quick Sanity Check
if (This.fHeader.fFTable < pos || This.fHeader.fFTable > This.fHeader.fLength) {
throw new IOException("Break iterator Rule data corrupt");
}
// Skip over any padding preceding this table
ICUBinary.skipBytes(bytes, This.fHeader.fFTable - pos);
pos = This.fHeader.fFTable;
This.fFTable = RBBIStateTable.get(bytes, This.fHeader.fFTableLen);
pos += This.fHeader.fFTableLen;
//
// Read in the Reverse state table
//
// Skip over any padding in the file
ICUBinary.skipBytes(bytes, This.fHeader.fRTable - pos);
pos = This.fHeader.fRTable;
// Create & fill the table itself.
This.fRTable = RBBIStateTable.get(bytes, This.fHeader.fRTableLen);
pos += This.fHeader.fRTableLen;
//
// Unserialize the Character categories TRIE
// Because we can't be absolutely certain where the Trie deserialize will
// leave the buffer, leave position unchanged.
// The seek to the start of the next item following the TRIE will get us
// back in sync.
//
ICUBinary.skipBytes(bytes, This.fHeader.fTrie - pos); // seek buffer from end of
pos = This.fHeader.fTrie; // previous section to the start of the trie
bytes.mark(); // Mark position of start of TRIE in the input
// and tell Java to keep the mark valid so long
// as we don't go more than 100 bytes past the
// past the end of the TRIE.
This.fTrie = Trie2.createFromSerialized(bytes); // Deserialize the TRIE, leaving buffer
// at an unknown position, preceding the
// padding between TRIE and following section.
bytes.reset(); // Move buffer back to marked position at
// the start of the serialized TRIE. Now our
// "pos" variable and the buffer are in
// agreement.
//
// Read the Rule Status Table
//
if (pos > This.fHeader.fStatusTable) {
throw new IOException("Break iterator Rule data corrupt");
}
ICUBinary.skipBytes(bytes, This.fHeader.fStatusTable - pos);
pos = This.fHeader.fStatusTable;
This.fStatusTable = ICUBinary.getInts(
bytes, This.fHeader.fStatusTableLen / 4, This.fHeader.fStatusTableLen & 3);
pos += This.fHeader.fStatusTableLen;
//
// Put the break rule source into a String
//
if (pos > This.fHeader.fRuleSource) {
throw new IOException("Break iterator Rule data corrupt");
}
ICUBinary.skipBytes(bytes, This.fHeader.fRuleSource - pos);
pos = This.fHeader.fRuleSource;
This.fRuleSource = ICUBinary.getString(
bytes, This.fHeader.fRuleSourceLen / 2, This.fHeader.fRuleSourceLen & 1);
if (RuleBasedBreakIterator.fDebugEnv!=null && RuleBasedBreakIterator.fDebugEnv.indexOf("data")>=0) {
This.dump(System.out);
}
return This;
}
/** Debug function to display the break iterator data. */
public void dump(java.io.PrintStream out) {
if (fFTable == null) {
// There is no table. Fail early for testing purposes.
throw new NullPointerException();
}
out.println("RBBI Data Wrapper dump ...");
out.println();
out.println("Forward State Table");
dumpTable(out, fFTable);
out.println("Reverse State Table");
dumpTable(out, fRTable);
dumpCharCategories(out);
out.println("Source Rules: " + fRuleSource);
}
/** Fixed width int-to-string conversion. */
static public String intToString(int n, int width) {
StringBuilder dest = new StringBuilder(width);
dest.append(n);
while (dest.length() < width) {
dest.insert(0, ' ');
}
return dest.toString();
}
/** Fixed width int-to-string conversion. */
static public String intToHexString(int n, int width) {
StringBuilder dest = new StringBuilder(width);
dest.append(Integer.toHexString(n));
while (dest.length() < width) {
dest.insert(0, ' ');
}
return dest.toString();
}
/** Dump a state table. (A full set of RBBI rules has 4 state tables.) */
private void dumpTable(java.io.PrintStream out, RBBIStateTable table) {
if (table == null || table.fTable.length == 0) {
out.println(" -- null -- ");
} else {
int n;
int state;
StringBuilder header = new StringBuilder(" Row Acc Look Tag");
for (n=0; n fHeader.fCatCount) {
out.println("Error, bad category " + Integer.toHexString(category) +
" for char " + Integer.toHexString(char32));
break;
}
if (category == lastCat ) {
rangeEnd = char32;
} else {
if (lastCat >= 0) {
if (catStrings[lastCat].length() > lastNewline[lastCat] + 70) {
lastNewline[lastCat] = catStrings[lastCat].length() + 10;
catStrings[lastCat] += "\n ";
}
catStrings[lastCat] += " " + Integer.toHexString(rangeStart);
if (rangeEnd != rangeStart) {
catStrings[lastCat] += "-" + Integer.toHexString(rangeEnd);
}
}
lastCat = category;
rangeStart = rangeEnd = char32;
}
}
catStrings[lastCat] += " " + Integer.toHexString(rangeStart);
if (rangeEnd != rangeStart) {
catStrings[lastCat] += "-" + Integer.toHexString(rangeEnd);
}
for (category = 0; category <= fHeader.fCatCount; category ++) {
out.println (intToString(category, 5) + " " + catStrings[category]);
}
out.println();
}
}