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