001/* 002 * Copyright (C) 2009 The Guava Authors 003 * 004 * Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except 005 * in compliance with the License. You may obtain a copy of the License at 006 * 007 * http://www.apache.org/licenses/LICENSE-2.0 008 * 009 * Unless required by applicable law or agreed to in writing, software distributed under the License 010 * is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express 011 * or implied. See the License for the specific language governing permissions and limitations under 012 * the License. 013 */ 014 015package com.google.common.primitives; 016 017import static com.google.common.base.Preconditions.checkArgument; 018import static com.google.common.base.Preconditions.checkNotNull; 019import static com.google.common.base.Preconditions.checkPositionIndexes; 020import static java.util.Objects.requireNonNull; 021 022import com.google.common.annotations.GwtIncompatible; 023import com.google.common.annotations.J2ktIncompatible; 024import com.google.common.annotations.VisibleForTesting; 025import com.google.errorprone.annotations.CanIgnoreReturnValue; 026import java.nio.ByteOrder; 027import java.util.Arrays; 028import java.util.Comparator; 029import sun.misc.Unsafe; 030 031/** 032 * Static utility methods pertaining to {@code byte} primitives that interpret values as 033 * <i>unsigned</i> (that is, any negative value {@code b} is treated as the positive value {@code 034 * 256 + b}). The corresponding methods that treat the values as signed are found in {@link 035 * SignedBytes}, and the methods for which signedness is not an issue are in {@link Bytes}. 036 * 037 * <p>See the Guava User Guide article on <a 038 * href="https://github.com/google/guava/wiki/PrimitivesExplained">primitive utilities</a>. 039 * 040 * @author Kevin Bourrillion 041 * @author Martin Buchholz 042 * @author Hiroshi Yamauchi 043 * @author Louis Wasserman 044 * @since 1.0 045 */ 046@J2ktIncompatible 047@GwtIncompatible 048@ElementTypesAreNonnullByDefault 049public final class UnsignedBytes { 050 private UnsignedBytes() {} 051 052 /** 053 * The largest power of two that can be represented as an unsigned {@code byte}. 054 * 055 * @since 10.0 056 */ 057 public static final byte MAX_POWER_OF_TWO = (byte) 0x80; 058 059 /** 060 * The largest value that fits into an unsigned byte. 061 * 062 * @since 13.0 063 */ 064 public static final byte MAX_VALUE = (byte) 0xFF; 065 066 private static final int UNSIGNED_MASK = 0xFF; 067 068 /** 069 * Returns the value of the given byte as an integer, when treated as unsigned. That is, returns 070 * {@code value + 256} if {@code value} is negative; {@code value} itself otherwise. 071 * 072 * <p><b>Java 8 users:</b> use {@link Byte#toUnsignedInt(byte)} instead. 073 * 074 * @since 6.0 075 */ 076 public static int toInt(byte value) { 077 return value & UNSIGNED_MASK; 078 } 079 080 /** 081 * Returns the {@code byte} value that, when treated as unsigned, is equal to {@code value}, if 082 * possible. 083 * 084 * @param value a value between 0 and 255 inclusive 085 * @return the {@code byte} value that, when treated as unsigned, equals {@code value} 086 * @throws IllegalArgumentException if {@code value} is negative or greater than 255 087 */ 088 @CanIgnoreReturnValue 089 public static byte checkedCast(long value) { 090 checkArgument(value >> Byte.SIZE == 0, "out of range: %s", value); 091 return (byte) value; 092 } 093 094 /** 095 * Returns the {@code byte} value that, when treated as unsigned, is nearest in value to {@code 096 * value}. 097 * 098 * @param value any {@code long} value 099 * @return {@code (byte) 255} if {@code value >= 255}, {@code (byte) 0} if {@code value <= 0}, and 100 * {@code value} cast to {@code byte} otherwise 101 */ 102 public static byte saturatedCast(long value) { 103 if (value > toInt(MAX_VALUE)) { 104 return MAX_VALUE; // -1 105 } 106 if (value < 0) { 107 return (byte) 0; 108 } 109 return (byte) value; 110 } 111 112 /** 113 * Compares the two specified {@code byte} values, treating them as unsigned values between 0 and 114 * 255 inclusive. For example, {@code (byte) -127} is considered greater than {@code (byte) 127} 115 * because it is seen as having the value of positive {@code 129}. 116 * 117 * @param a the first {@code byte} to compare 118 * @param b the second {@code byte} to compare 119 * @return a negative value if {@code a} is less than {@code b}; a positive value if {@code a} is 120 * greater than {@code b}; or zero if they are equal 121 */ 122 public static int compare(byte a, byte b) { 123 return toInt(a) - toInt(b); 124 } 125 126 /** 127 * Returns the least value present in {@code array}, treating values as unsigned. 128 * 129 * @param array a <i>nonempty</i> array of {@code byte} values 130 * @return the value present in {@code array} that is less than or equal to every other value in 131 * the array according to {@link #compare} 132 * @throws IllegalArgumentException if {@code array} is empty 133 */ 134 public static byte min(byte... array) { 135 checkArgument(array.length > 0); 136 int min = toInt(array[0]); 137 for (int i = 1; i < array.length; i++) { 138 int next = toInt(array[i]); 139 if (next < min) { 140 min = next; 141 } 142 } 143 return (byte) min; 144 } 145 146 /** 147 * Returns the greatest value present in {@code array}, treating values as unsigned. 148 * 149 * @param array a <i>nonempty</i> array of {@code byte} values 150 * @return the value present in {@code array} that is greater than or equal to every other value 151 * in the array according to {@link #compare} 152 * @throws IllegalArgumentException if {@code array} is empty 153 */ 154 public static byte max(byte... array) { 155 checkArgument(array.length > 0); 156 int max = toInt(array[0]); 157 for (int i = 1; i < array.length; i++) { 158 int next = toInt(array[i]); 159 if (next > max) { 160 max = next; 161 } 162 } 163 return (byte) max; 164 } 165 166 /** 167 * Returns a string representation of x, where x is treated as unsigned. 168 * 169 * @since 13.0 170 */ 171 public static String toString(byte x) { 172 return toString(x, 10); 173 } 174 175 /** 176 * Returns a string representation of {@code x} for the given radix, where {@code x} is treated as 177 * unsigned. 178 * 179 * @param x the value to convert to a string. 180 * @param radix the radix to use while working with {@code x} 181 * @throws IllegalArgumentException if {@code radix} is not between {@link Character#MIN_RADIX} 182 * and {@link Character#MAX_RADIX}. 183 * @since 13.0 184 */ 185 public static String toString(byte x, int radix) { 186 checkArgument( 187 radix >= Character.MIN_RADIX && radix <= Character.MAX_RADIX, 188 "radix (%s) must be between Character.MIN_RADIX and Character.MAX_RADIX", 189 radix); 190 // Benchmarks indicate this is probably not worth optimizing. 191 return Integer.toString(toInt(x), radix); 192 } 193 194 /** 195 * Returns the unsigned {@code byte} value represented by the given decimal string. 196 * 197 * @throws NumberFormatException if the string does not contain a valid unsigned {@code byte} 198 * value 199 * @throws NullPointerException if {@code string} is null (in contrast to {@link 200 * Byte#parseByte(String)}) 201 * @since 13.0 202 */ 203 @CanIgnoreReturnValue 204 public static byte parseUnsignedByte(String string) { 205 return parseUnsignedByte(string, 10); 206 } 207 208 /** 209 * Returns the unsigned {@code byte} value represented by a string with the given radix. 210 * 211 * @param string the string containing the unsigned {@code byte} representation to be parsed. 212 * @param radix the radix to use while parsing {@code string} 213 * @throws NumberFormatException if the string does not contain a valid unsigned {@code byte} with 214 * the given radix, or if {@code radix} is not between {@link Character#MIN_RADIX} and {@link 215 * Character#MAX_RADIX}. 216 * @throws NullPointerException if {@code string} is null (in contrast to {@link 217 * Byte#parseByte(String)}) 218 * @since 13.0 219 */ 220 @CanIgnoreReturnValue 221 public static byte parseUnsignedByte(String string, int radix) { 222 int parse = Integer.parseInt(checkNotNull(string), radix); 223 // We need to throw a NumberFormatException, so we have to duplicate checkedCast. =( 224 if (parse >> Byte.SIZE == 0) { 225 return (byte) parse; 226 } else { 227 throw new NumberFormatException("out of range: " + parse); 228 } 229 } 230 231 /** 232 * Returns a string containing the supplied {@code byte} values separated by {@code separator}. 233 * For example, {@code join(":", (byte) 1, (byte) 2, (byte) 255)} returns the string {@code 234 * "1:2:255"}. 235 * 236 * @param separator the text that should appear between consecutive values in the resulting string 237 * (but not at the start or end) 238 * @param array an array of {@code byte} values, possibly empty 239 */ 240 public static String join(String separator, byte... array) { 241 checkNotNull(separator); 242 if (array.length == 0) { 243 return ""; 244 } 245 246 // For pre-sizing a builder, just get the right order of magnitude 247 StringBuilder builder = new StringBuilder(array.length * (3 + separator.length())); 248 builder.append(toInt(array[0])); 249 for (int i = 1; i < array.length; i++) { 250 builder.append(separator).append(toString(array[i])); 251 } 252 return builder.toString(); 253 } 254 255 /** 256 * Returns a comparator that compares two {@code byte} arrays <a 257 * href="http://en.wikipedia.org/wiki/Lexicographical_order">lexicographically</a>. That is, it 258 * compares, using {@link #compare(byte, byte)}), the first pair of values that follow any common 259 * prefix, or when one array is a prefix of the other, treats the shorter array as the lesser. For 260 * example, {@code [] < [0x01] < [0x01, 0x7F] < [0x01, 0x80] < [0x02]}. Values are treated as 261 * unsigned. 262 * 263 * <p>The returned comparator is inconsistent with {@link Object#equals(Object)} (since arrays 264 * support only identity equality), but it is consistent with {@link 265 * java.util.Arrays#equals(byte[], byte[])}. 266 * 267 * @since 2.0 268 */ 269 public static Comparator<byte[]> lexicographicalComparator() { 270 return LexicographicalComparatorHolder.BEST_COMPARATOR; 271 } 272 273 @VisibleForTesting 274 static Comparator<byte[]> lexicographicalComparatorJavaImpl() { 275 return LexicographicalComparatorHolder.PureJavaComparator.INSTANCE; 276 } 277 278 /** 279 * Provides a lexicographical comparator implementation; either a Java implementation or a faster 280 * implementation based on {@link Unsafe}. 281 * 282 * <p>Uses reflection to gracefully fall back to the Java implementation if {@code Unsafe} isn't 283 * available. 284 */ 285 @VisibleForTesting 286 static class LexicographicalComparatorHolder { 287 static final String UNSAFE_COMPARATOR_NAME = 288 LexicographicalComparatorHolder.class.getName() + "$UnsafeComparator"; 289 290 static final Comparator<byte[]> BEST_COMPARATOR = getBestComparator(); 291 292 @VisibleForTesting 293 enum UnsafeComparator implements Comparator<byte[]> { 294 INSTANCE; 295 296 static final boolean BIG_ENDIAN = ByteOrder.nativeOrder().equals(ByteOrder.BIG_ENDIAN); 297 298 /* 299 * The following static final fields exist for performance reasons. 300 * 301 * In UnsignedBytesBenchmark, accessing the following objects via static final fields is the 302 * fastest (more than twice as fast as the Java implementation, vs ~1.5x with non-final static 303 * fields, on x86_32) under the Hotspot server compiler. The reason is obviously that the 304 * non-final fields need to be reloaded inside the loop. 305 * 306 * And, no, defining (final or not) local variables out of the loop still isn't as good 307 * because the null check on the theUnsafe object remains inside the loop and 308 * BYTE_ARRAY_BASE_OFFSET doesn't get constant-folded. 309 * 310 * The compiler can treat static final fields as compile-time constants and can constant-fold 311 * them while (final or not) local variables are run time values. 312 */ 313 314 static final Unsafe theUnsafe = getUnsafe(); 315 316 /** The offset to the first element in a byte array. */ 317 static final int BYTE_ARRAY_BASE_OFFSET = theUnsafe.arrayBaseOffset(byte[].class); 318 319 static { 320 // fall back to the safer pure java implementation unless we're in 321 // a 64-bit JVM with an 8-byte aligned field offset. 322 if (!("64".equals(System.getProperty("sun.arch.data.model")) 323 && (BYTE_ARRAY_BASE_OFFSET % 8) == 0 324 // sanity check - this should never fail 325 && theUnsafe.arrayIndexScale(byte[].class) == 1)) { 326 throw new Error(); // force fallback to PureJavaComparator 327 } 328 } 329 330 /** 331 * Returns a sun.misc.Unsafe. Suitable for use in a 3rd party package. Replace with a simple 332 * call to Unsafe.getUnsafe when integrating into a jdk. 333 * 334 * @return a sun.misc.Unsafe 335 */ 336 private static sun.misc.Unsafe getUnsafe() { 337 try { 338 return sun.misc.Unsafe.getUnsafe(); 339 } catch (SecurityException e) { 340 // that's okay; try reflection instead 341 } 342 try { 343 return java.security.AccessController.doPrivileged( 344 new java.security.PrivilegedExceptionAction<sun.misc.Unsafe>() { 345 @Override 346 public sun.misc.Unsafe run() throws Exception { 347 Class<sun.misc.Unsafe> k = sun.misc.Unsafe.class; 348 for (java.lang.reflect.Field f : k.getDeclaredFields()) { 349 f.setAccessible(true); 350 Object x = f.get(null); 351 if (k.isInstance(x)) { 352 return k.cast(x); 353 } 354 } 355 throw new NoSuchFieldError("the Unsafe"); 356 } 357 }); 358 } catch (java.security.PrivilegedActionException e) { 359 throw new RuntimeException("Could not initialize intrinsics", e.getCause()); 360 } 361 } 362 363 @Override 364 public int compare(byte[] left, byte[] right) { 365 int stride = 8; 366 int minLength = Math.min(left.length, right.length); 367 int strideLimit = minLength & ~(stride - 1); 368 int i; 369 370 /* 371 * Compare 8 bytes at a time. Benchmarking on x86 shows a stride of 8 bytes is no slower 372 * than 4 bytes even on 32-bit. On the other hand, it is substantially faster on 64-bit. 373 */ 374 for (i = 0; i < strideLimit; i += stride) { 375 long lw = theUnsafe.getLong(left, BYTE_ARRAY_BASE_OFFSET + (long) i); 376 long rw = theUnsafe.getLong(right, BYTE_ARRAY_BASE_OFFSET + (long) i); 377 if (lw != rw) { 378 if (BIG_ENDIAN) { 379 return UnsignedLongs.compare(lw, rw); 380 } 381 382 /* 383 * We want to compare only the first index where left[index] != right[index]. This 384 * corresponds to the least significant nonzero byte in lw ^ rw, since lw and rw are 385 * little-endian. Long.numberOfTrailingZeros(diff) tells us the least significant 386 * nonzero bit, and zeroing out the first three bits of L.nTZ gives us the shift to get 387 * that least significant nonzero byte. 388 */ 389 int n = Long.numberOfTrailingZeros(lw ^ rw) & ~0x7; 390 return ((int) ((lw >>> n) & UNSIGNED_MASK)) - ((int) ((rw >>> n) & UNSIGNED_MASK)); 391 } 392 } 393 394 // The epilogue to cover the last (minLength % stride) elements. 395 for (; i < minLength; i++) { 396 int result = UnsignedBytes.compare(left[i], right[i]); 397 if (result != 0) { 398 return result; 399 } 400 } 401 return left.length - right.length; 402 } 403 404 @Override 405 public String toString() { 406 return "UnsignedBytes.lexicographicalComparator() (sun.misc.Unsafe version)"; 407 } 408 } 409 410 enum PureJavaComparator implements Comparator<byte[]> { 411 INSTANCE; 412 413 @Override 414 public int compare(byte[] left, byte[] right) { 415 int minLength = Math.min(left.length, right.length); 416 for (int i = 0; i < minLength; i++) { 417 int result = UnsignedBytes.compare(left[i], right[i]); 418 if (result != 0) { 419 return result; 420 } 421 } 422 return left.length - right.length; 423 } 424 425 @Override 426 public String toString() { 427 return "UnsignedBytes.lexicographicalComparator() (pure Java version)"; 428 } 429 } 430 431 /** 432 * Returns the Unsafe-using Comparator, or falls back to the pure-Java implementation if unable 433 * to do so. 434 */ 435 static Comparator<byte[]> getBestComparator() { 436 try { 437 Class<?> theClass = Class.forName(UNSAFE_COMPARATOR_NAME); 438 439 // requireNonNull is safe because the class is an enum. 440 Object[] constants = requireNonNull(theClass.getEnumConstants()); 441 442 // yes, UnsafeComparator does implement Comparator<byte[]> 443 @SuppressWarnings("unchecked") 444 Comparator<byte[]> comparator = (Comparator<byte[]>) constants[0]; 445 return comparator; 446 } catch (Throwable t) { // ensure we really catch *everything* 447 return lexicographicalComparatorJavaImpl(); 448 } 449 } 450 } 451 452 private static byte flip(byte b) { 453 return (byte) (b ^ 0x80); 454 } 455 456 /** 457 * Sorts the array, treating its elements as unsigned bytes. 458 * 459 * @since 23.1 460 */ 461 public static void sort(byte[] array) { 462 checkNotNull(array); 463 sort(array, 0, array.length); 464 } 465 466 /** 467 * Sorts the array between {@code fromIndex} inclusive and {@code toIndex} exclusive, treating its 468 * elements as unsigned bytes. 469 * 470 * @since 23.1 471 */ 472 public static void sort(byte[] array, int fromIndex, int toIndex) { 473 checkNotNull(array); 474 checkPositionIndexes(fromIndex, toIndex, array.length); 475 for (int i = fromIndex; i < toIndex; i++) { 476 array[i] = flip(array[i]); 477 } 478 Arrays.sort(array, fromIndex, toIndex); 479 for (int i = fromIndex; i < toIndex; i++) { 480 array[i] = flip(array[i]); 481 } 482 } 483 484 /** 485 * Sorts the elements of {@code array} in descending order, interpreting them as unsigned 8-bit 486 * integers. 487 * 488 * @since 23.1 489 */ 490 public static void sortDescending(byte[] array) { 491 checkNotNull(array); 492 sortDescending(array, 0, array.length); 493 } 494 495 /** 496 * Sorts the elements of {@code array} between {@code fromIndex} inclusive and {@code toIndex} 497 * exclusive in descending order, interpreting them as unsigned 8-bit integers. 498 * 499 * @since 23.1 500 */ 501 public static void sortDescending(byte[] array, int fromIndex, int toIndex) { 502 checkNotNull(array); 503 checkPositionIndexes(fromIndex, toIndex, array.length); 504 for (int i = fromIndex; i < toIndex; i++) { 505 array[i] ^= Byte.MAX_VALUE; 506 } 507 Arrays.sort(array, fromIndex, toIndex); 508 for (int i = fromIndex; i < toIndex; i++) { 509 array[i] ^= Byte.MAX_VALUE; 510 } 511 } 512}