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.lang.reflect.Field; 027import java.nio.ByteOrder; 028import java.security.AccessController; 029import java.security.PrivilegedActionException; 030import java.security.PrivilegedExceptionAction; 031import java.util.Arrays; 032import java.util.Comparator; 033import sun.misc.Unsafe; 034 035/** 036 * Static utility methods pertaining to {@code byte} primitives that interpret values as 037 * <i>unsigned</i> (that is, any negative value {@code b} is treated as the positive value {@code 038 * 256 + b}). The corresponding methods that treat the values as signed are found in {@link 039 * SignedBytes}, and the methods for which signedness is not an issue are in {@link Bytes}. 040 * 041 * <p>See the Guava User Guide article on <a 042 * href="https://github.com/google/guava/wiki/PrimitivesExplained">primitive utilities</a>. 043 * 044 * @author Kevin Bourrillion 045 * @author Martin Buchholz 046 * @author Hiroshi Yamauchi 047 * @author Louis Wasserman 048 * @since 1.0 049 */ 050@J2ktIncompatible 051@GwtIncompatible 052@ElementTypesAreNonnullByDefault 053public final class UnsignedBytes { 054 private UnsignedBytes() {} 055 056 /** 057 * The largest power of two that can be represented as an unsigned {@code byte}. 058 * 059 * @since 10.0 060 */ 061 public static final byte MAX_POWER_OF_TWO = (byte) 0x80; 062 063 /** 064 * The largest value that fits into an unsigned byte. 065 * 066 * @since 13.0 067 */ 068 public static final byte MAX_VALUE = (byte) 0xFF; 069 070 private static final int UNSIGNED_MASK = 0xFF; 071 072 /** 073 * Returns the value of the given byte as an integer, when treated as unsigned. That is, returns 074 * {@code value + 256} if {@code value} is negative; {@code value} itself otherwise. 075 * 076 * <p><b>Java 8+ users:</b> use {@link Byte#toUnsignedInt(byte)} instead. 077 * 078 * @since 6.0 079 */ 080 public static int toInt(byte value) { 081 return value & UNSIGNED_MASK; 082 } 083 084 /** 085 * Returns the {@code byte} value that, when treated as unsigned, is equal to {@code value}, if 086 * possible. 087 * 088 * @param value a value between 0 and 255 inclusive 089 * @return the {@code byte} value that, when treated as unsigned, equals {@code value} 090 * @throws IllegalArgumentException if {@code value} is negative or greater than 255 091 */ 092 @CanIgnoreReturnValue 093 public static byte checkedCast(long value) { 094 checkArgument(value >> Byte.SIZE == 0, "out of range: %s", value); 095 return (byte) value; 096 } 097 098 /** 099 * Returns the {@code byte} value that, when treated as unsigned, is nearest in value to {@code 100 * value}. 101 * 102 * @param value any {@code long} value 103 * @return {@code (byte) 255} if {@code value >= 255}, {@code (byte) 0} if {@code value <= 0}, and 104 * {@code value} cast to {@code byte} otherwise 105 */ 106 public static byte saturatedCast(long value) { 107 if (value > toInt(MAX_VALUE)) { 108 return MAX_VALUE; // -1 109 } 110 if (value < 0) { 111 return (byte) 0; 112 } 113 return (byte) value; 114 } 115 116 /** 117 * Compares the two specified {@code byte} values, treating them as unsigned values between 0 and 118 * 255 inclusive. For example, {@code (byte) -127} is considered greater than {@code (byte) 127} 119 * because it is seen as having the value of positive {@code 129}. 120 * 121 * @param a the first {@code byte} to compare 122 * @param b the second {@code byte} to compare 123 * @return a negative value if {@code a} is less than {@code b}; a positive value if {@code a} is 124 * greater than {@code b}; or zero if they are equal 125 */ 126 public static int compare(byte a, byte b) { 127 return toInt(a) - toInt(b); 128 } 129 130 /** 131 * Returns the least value present in {@code array}, treating values as unsigned. 132 * 133 * @param array a <i>nonempty</i> array of {@code byte} values 134 * @return the value present in {@code array} that is less than or equal to every other value in 135 * the array according to {@link #compare} 136 * @throws IllegalArgumentException if {@code array} is empty 137 */ 138 public static byte min(byte... array) { 139 checkArgument(array.length > 0); 140 int min = toInt(array[0]); 141 for (int i = 1; i < array.length; i++) { 142 int next = toInt(array[i]); 143 if (next < min) { 144 min = next; 145 } 146 } 147 return (byte) min; 148 } 149 150 /** 151 * Returns the greatest value present in {@code array}, treating values as unsigned. 152 * 153 * @param array a <i>nonempty</i> array of {@code byte} values 154 * @return the value present in {@code array} that is greater than or equal to every other value 155 * in the array according to {@link #compare} 156 * @throws IllegalArgumentException if {@code array} is empty 157 */ 158 public static byte max(byte... array) { 159 checkArgument(array.length > 0); 160 int max = toInt(array[0]); 161 for (int i = 1; i < array.length; i++) { 162 int next = toInt(array[i]); 163 if (next > max) { 164 max = next; 165 } 166 } 167 return (byte) max; 168 } 169 170 /** 171 * Returns a string representation of x, where x is treated as unsigned. 172 * 173 * @since 13.0 174 */ 175 public static String toString(byte x) { 176 return toString(x, 10); 177 } 178 179 /** 180 * Returns a string representation of {@code x} for the given radix, where {@code x} is treated as 181 * unsigned. 182 * 183 * @param x the value to convert to a string. 184 * @param radix the radix to use while working with {@code x} 185 * @throws IllegalArgumentException if {@code radix} is not between {@link Character#MIN_RADIX} 186 * and {@link Character#MAX_RADIX}. 187 * @since 13.0 188 */ 189 public static String toString(byte x, int radix) { 190 checkArgument( 191 radix >= Character.MIN_RADIX && radix <= Character.MAX_RADIX, 192 "radix (%s) must be between Character.MIN_RADIX and Character.MAX_RADIX", 193 radix); 194 // Benchmarks indicate this is probably not worth optimizing. 195 return Integer.toString(toInt(x), radix); 196 } 197 198 /** 199 * Returns the unsigned {@code byte} value represented by the given decimal string. 200 * 201 * @throws NumberFormatException if the string does not contain a valid unsigned {@code byte} 202 * value 203 * @throws NullPointerException if {@code string} is null (in contrast to {@link 204 * Byte#parseByte(String)}) 205 * @since 13.0 206 */ 207 @CanIgnoreReturnValue 208 public static byte parseUnsignedByte(String string) { 209 return parseUnsignedByte(string, 10); 210 } 211 212 /** 213 * Returns the unsigned {@code byte} value represented by a string with the given radix. 214 * 215 * @param string the string containing the unsigned {@code byte} representation to be parsed. 216 * @param radix the radix to use while parsing {@code string} 217 * @throws NumberFormatException if the string does not contain a valid unsigned {@code byte} with 218 * the given radix, or if {@code radix} is not between {@link Character#MIN_RADIX} and {@link 219 * Character#MAX_RADIX}. 220 * @throws NullPointerException if {@code string} is null (in contrast to {@link 221 * Byte#parseByte(String)}) 222 * @since 13.0 223 */ 224 @CanIgnoreReturnValue 225 public static byte parseUnsignedByte(String string, int radix) { 226 int parse = Integer.parseInt(checkNotNull(string), radix); 227 // We need to throw a NumberFormatException, so we have to duplicate checkedCast. =( 228 if (parse >> Byte.SIZE == 0) { 229 return (byte) parse; 230 } else { 231 throw new NumberFormatException("out of range: " + parse); 232 } 233 } 234 235 /** 236 * Returns a string containing the supplied {@code byte} values separated by {@code separator}. 237 * For example, {@code join(":", (byte) 1, (byte) 2, (byte) 255)} returns the string {@code 238 * "1:2:255"}. 239 * 240 * @param separator the text that should appear between consecutive values in the resulting string 241 * (but not at the start or end) 242 * @param array an array of {@code byte} values, possibly empty 243 */ 244 public static String join(String separator, byte... array) { 245 checkNotNull(separator); 246 if (array.length == 0) { 247 return ""; 248 } 249 250 // For pre-sizing a builder, just get the right order of magnitude 251 StringBuilder builder = new StringBuilder(array.length * (3 + separator.length())); 252 builder.append(toInt(array[0])); 253 for (int i = 1; i < array.length; i++) { 254 builder.append(separator).append(toString(array[i])); 255 } 256 return builder.toString(); 257 } 258 259 /** 260 * Returns a comparator that compares two {@code byte} arrays <a 261 * href="http://en.wikipedia.org/wiki/Lexicographical_order">lexicographically</a>. That is, it 262 * compares, using {@link #compare(byte, byte)}), the first pair of values that follow any common 263 * prefix, or when one array is a prefix of the other, treats the shorter array as the lesser. For 264 * example, {@code [] < [0x01] < [0x01, 0x7F] < [0x01, 0x80] < [0x02]}. Values are treated as 265 * unsigned. 266 * 267 * <p>The returned comparator is inconsistent with {@link Object#equals(Object)} (since arrays 268 * support only identity equality), but it is consistent with {@link 269 * java.util.Arrays#equals(byte[], byte[])}. 270 * 271 * @since 2.0 272 */ 273 public static Comparator<byte[]> lexicographicalComparator() { 274 return LexicographicalComparatorHolder.BEST_COMPARATOR; 275 } 276 277 @VisibleForTesting 278 static Comparator<byte[]> lexicographicalComparatorJavaImpl() { 279 return LexicographicalComparatorHolder.PureJavaComparator.INSTANCE; 280 } 281 282 /** 283 * Provides a lexicographical comparator implementation; either a Java implementation or a faster 284 * implementation based on {@link Unsafe}. 285 * 286 * <p>Uses reflection to gracefully fall back to the Java implementation if {@code Unsafe} isn't 287 * available. 288 */ 289 @VisibleForTesting 290 static class LexicographicalComparatorHolder { 291 static final String UNSAFE_COMPARATOR_NAME = 292 LexicographicalComparatorHolder.class.getName() + "$UnsafeComparator"; 293 294 static final Comparator<byte[]> BEST_COMPARATOR = getBestComparator(); 295 296 @SuppressWarnings({"SunApi", "removal"}) // b/345822163 297 @VisibleForTesting 298 enum UnsafeComparator implements Comparator<byte[]> { 299 INSTANCE; 300 301 static final boolean BIG_ENDIAN = ByteOrder.nativeOrder().equals(ByteOrder.BIG_ENDIAN); 302 303 /* 304 * The following static final fields exist for performance reasons. 305 * 306 * In UnsignedBytesBenchmark, accessing the following objects via static final fields is the 307 * fastest (more than twice as fast as the Java implementation, vs ~1.5x with non-final static 308 * fields, on x86_32) under the Hotspot server compiler. The reason is obviously that the 309 * non-final fields need to be reloaded inside the loop. 310 * 311 * And, no, defining (final or not) local variables out of the loop still isn't as good 312 * because the null check on the theUnsafe object remains inside the loop and 313 * BYTE_ARRAY_BASE_OFFSET doesn't get constant-folded. 314 * 315 * The compiler can treat static final fields as compile-time constants and can constant-fold 316 * them while (final or not) local variables are run time values. 317 */ 318 319 static final Unsafe theUnsafe = getUnsafe(); 320 321 /** The offset to the first element in a byte array. */ 322 static final int BYTE_ARRAY_BASE_OFFSET = theUnsafe.arrayBaseOffset(byte[].class); 323 324 static { 325 // fall back to the safer pure java implementation unless we're in 326 // a 64-bit JVM with an 8-byte aligned field offset. 327 if (!("64".equals(System.getProperty("sun.arch.data.model")) 328 && (BYTE_ARRAY_BASE_OFFSET % 8) == 0 329 // sanity check - this should never fail 330 && theUnsafe.arrayIndexScale(byte[].class) == 1)) { 331 throw new Error(); // force fallback to PureJavaComparator 332 } 333 } 334 335 /** 336 * Returns a sun.misc.Unsafe. Suitable for use in a 3rd party package. Replace with a simple 337 * call to Unsafe.getUnsafe when integrating into a jdk. 338 * 339 * @return a sun.misc.Unsafe 340 */ 341 private static Unsafe getUnsafe() { 342 try { 343 return Unsafe.getUnsafe(); 344 } catch (SecurityException e) { 345 // that's okay; try reflection instead 346 } 347 try { 348 return AccessController.doPrivileged( 349 new PrivilegedExceptionAction<Unsafe>() { 350 @Override 351 public Unsafe run() throws Exception { 352 Class<Unsafe> k = Unsafe.class; 353 for (Field f : k.getDeclaredFields()) { 354 f.setAccessible(true); 355 Object x = f.get(null); 356 if (k.isInstance(x)) { 357 return k.cast(x); 358 } 359 } 360 throw new NoSuchFieldError("the Unsafe"); 361 } 362 }); 363 } catch (PrivilegedActionException e) { 364 throw new RuntimeException("Could not initialize intrinsics", e.getCause()); 365 } 366 } 367 368 @Override 369 // Long.compareUnsigned is available under Android, which is what we really care about. 370 @SuppressWarnings("Java7ApiChecker") 371 public int compare(byte[] left, byte[] right) { 372 int stride = 8; 373 int minLength = Math.min(left.length, right.length); 374 int strideLimit = minLength & ~(stride - 1); 375 int i; 376 377 /* 378 * Compare 8 bytes at a time. Benchmarking on x86 shows a stride of 8 bytes is no slower 379 * than 4 bytes even on 32-bit. On the other hand, it is substantially faster on 64-bit. 380 */ 381 for (i = 0; i < strideLimit; i += stride) { 382 long lw = theUnsafe.getLong(left, BYTE_ARRAY_BASE_OFFSET + (long) i); 383 long rw = theUnsafe.getLong(right, BYTE_ARRAY_BASE_OFFSET + (long) i); 384 if (lw != rw) { 385 if (BIG_ENDIAN) { 386 return Long.compareUnsigned(lw, rw); 387 } 388 389 /* 390 * We want to compare only the first index where left[index] != right[index]. This 391 * corresponds to the least significant nonzero byte in lw ^ rw, since lw and rw are 392 * little-endian. Long.numberOfTrailingZeros(diff) tells us the least significant 393 * nonzero bit, and zeroing out the first three bits of L.nTZ gives us the shift to get 394 * that least significant nonzero byte. 395 */ 396 int n = Long.numberOfTrailingZeros(lw ^ rw) & ~0x7; 397 return ((int) ((lw >>> n) & UNSIGNED_MASK)) - ((int) ((rw >>> n) & UNSIGNED_MASK)); 398 } 399 } 400 401 // The epilogue to cover the last (minLength % stride) elements. 402 for (; i < minLength; i++) { 403 int result = UnsignedBytes.compare(left[i], right[i]); 404 if (result != 0) { 405 return result; 406 } 407 } 408 return left.length - right.length; 409 } 410 411 @Override 412 public String toString() { 413 return "UnsignedBytes.lexicographicalComparator() (sun.misc.Unsafe version)"; 414 } 415 } 416 417 enum PureJavaComparator implements Comparator<byte[]> { 418 INSTANCE; 419 420 @Override 421 public int compare(byte[] left, byte[] right) { 422 int minLength = Math.min(left.length, right.length); 423 for (int i = 0; i < minLength; i++) { 424 int result = UnsignedBytes.compare(left[i], right[i]); 425 if (result != 0) { 426 return result; 427 } 428 } 429 return left.length - right.length; 430 } 431 432 @Override 433 public String toString() { 434 return "UnsignedBytes.lexicographicalComparator() (pure Java version)"; 435 } 436 } 437 438 /** 439 * Returns the Unsafe-using Comparator, or falls back to the pure-Java implementation if unable 440 * to do so. 441 */ 442 static Comparator<byte[]> getBestComparator() { 443 try { 444 Class<?> theClass = Class.forName(UNSAFE_COMPARATOR_NAME); 445 446 // requireNonNull is safe because the class is an enum. 447 Object[] constants = requireNonNull(theClass.getEnumConstants()); 448 449 // yes, UnsafeComparator does implement Comparator<byte[]> 450 @SuppressWarnings("unchecked") 451 Comparator<byte[]> comparator = (Comparator<byte[]>) constants[0]; 452 return comparator; 453 } catch (Throwable t) { // ensure we really catch *everything* 454 return lexicographicalComparatorJavaImpl(); 455 } 456 } 457 } 458 459 private static byte flip(byte b) { 460 return (byte) (b ^ 0x80); 461 } 462 463 /** 464 * Sorts the array, treating its elements as unsigned bytes. 465 * 466 * @since 23.1 467 */ 468 public static void sort(byte[] array) { 469 checkNotNull(array); 470 sort(array, 0, array.length); 471 } 472 473 /** 474 * Sorts the array between {@code fromIndex} inclusive and {@code toIndex} exclusive, treating its 475 * elements as unsigned bytes. 476 * 477 * @since 23.1 478 */ 479 public static void sort(byte[] array, int fromIndex, int toIndex) { 480 checkNotNull(array); 481 checkPositionIndexes(fromIndex, toIndex, array.length); 482 for (int i = fromIndex; i < toIndex; i++) { 483 array[i] = flip(array[i]); 484 } 485 Arrays.sort(array, fromIndex, toIndex); 486 for (int i = fromIndex; i < toIndex; i++) { 487 array[i] = flip(array[i]); 488 } 489 } 490 491 /** 492 * Sorts the elements of {@code array} in descending order, interpreting them as unsigned 8-bit 493 * integers. 494 * 495 * @since 23.1 496 */ 497 public static void sortDescending(byte[] array) { 498 checkNotNull(array); 499 sortDescending(array, 0, array.length); 500 } 501 502 /** 503 * Sorts the elements of {@code array} between {@code fromIndex} inclusive and {@code toIndex} 504 * exclusive in descending order, interpreting them as unsigned 8-bit integers. 505 * 506 * @since 23.1 507 */ 508 public static void sortDescending(byte[] array, int fromIndex, int toIndex) { 509 checkNotNull(array); 510 checkPositionIndexes(fromIndex, toIndex, array.length); 511 for (int i = fromIndex; i < toIndex; i++) { 512 array[i] ^= Byte.MAX_VALUE; 513 } 514 Arrays.sort(array, fromIndex, toIndex); 515 for (int i = fromIndex; i < toIndex; i++) { 516 array[i] ^= Byte.MAX_VALUE; 517 } 518 } 519}