001/*
002 * Copyright (C) 2011 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.hash;
016
017import static com.google.common.base.Preconditions.checkArgument;
018import static com.google.common.base.Preconditions.checkNotNull;
019
020import com.google.errorprone.annotations.Immutable;
021import java.security.Key;
022import java.util.ArrayList;
023import java.util.Arrays;
024import java.util.Collections;
025import java.util.Iterator;
026import java.util.List;
027import java.util.zip.Adler32;
028import java.util.zip.CRC32;
029import java.util.zip.Checksum;
030import javax.annotation.CheckForNull;
031import javax.crypto.spec.SecretKeySpec;
032
033/**
034 * Static methods to obtain {@link HashFunction} instances, and other static hashing-related
035 * utilities.
036 *
037 * <p>A comparison of the various hash functions can be found <a
038 * href="https://docs.google.com/spreadsheets/d/1_q2EVcxA2HjcrlVMbaqXwMj31h9M5-Bqj_m8vITOwwk/">here</a>.
039 *
040 * @author Kevin Bourrillion
041 * @author Dimitris Andreou
042 * @author Kurt Alfred Kluever
043 * @since 11.0
044 */
045public final class Hashing {
046  /**
047   * Returns a general-purpose, <b>temporary-use</b>, non-cryptographic hash function. The algorithm
048   * the returned function implements is unspecified and subject to change without notice.
049   *
050   * <p><b>Warning:</b> a new random seed for these functions is chosen each time the {@code
051   * Hashing} class is loaded. <b>Do not use this method</b> if hash codes may escape the current
052   * process in any way, for example being sent over RPC, or saved to disk. For a general-purpose,
053   * non-cryptographic hash function that will never change behavior, we suggest {@link
054   * #murmur3_128}.
055   *
056   * <p>Repeated calls to this method on the same loaded {@code Hashing} class, using the same value
057   * for {@code minimumBits}, will return identically-behaving {@link HashFunction} instances.
058   *
059   * @param minimumBits a positive integer. This can be arbitrarily large. The returned {@link
060   *     HashFunction} instance may use memory proportional to this integer.
061   * @return a hash function, described above, that produces hash codes of length {@code
062   *     minimumBits} or greater
063   */
064  public static HashFunction goodFastHash(int minimumBits) {
065    int bits = checkPositiveAndMakeMultipleOf32(minimumBits);
066
067    if (bits == 32) {
068      return Murmur3_32HashFunction.GOOD_FAST_HASH_32;
069    }
070    if (bits <= 128) {
071      return Murmur3_128HashFunction.GOOD_FAST_HASH_128;
072    }
073
074    // Otherwise, join together some 128-bit murmur3s
075    int hashFunctionsNeeded = (bits + 127) / 128;
076    HashFunction[] hashFunctions = new HashFunction[hashFunctionsNeeded];
077    hashFunctions[0] = Murmur3_128HashFunction.GOOD_FAST_HASH_128;
078    int seed = GOOD_FAST_HASH_SEED;
079    for (int i = 1; i < hashFunctionsNeeded; i++) {
080      seed += 1500450271; // a prime; shouldn't matter
081      hashFunctions[i] = murmur3_128(seed);
082    }
083    return new ConcatenatedHashFunction(hashFunctions);
084  }
085
086  /**
087   * Used to randomize {@link #goodFastHash} instances, so that programs which persist anything
088   * dependent on the hash codes they produce will fail sooner.
089   */
090  @SuppressWarnings("GoodTime") // reading system time without TimeSource
091  static final int GOOD_FAST_HASH_SEED = (int) System.currentTimeMillis();
092
093  /**
094   * Returns a hash function implementing the <a
095   * href="https://github.com/aappleby/smhasher/blob/master/src/MurmurHash3.cpp">32-bit murmur3
096   * algorithm, x86 variant</a> (little-endian variant), using the given seed value, <b>with a known
097   * bug</b> as described in the deprecation text.
098   *
099   * <p>The C++ equivalent is the MurmurHash3_x86_32 function (Murmur3A), which however does not
100   * have the bug.
101   *
102   * @deprecated This implementation produces incorrect hash values from the {@link
103   *     HashFunction#hashString} method if the string contains non-BMP characters. Use {@link
104   *     #murmur3_32_fixed(int)} instead.
105   */
106  @Deprecated
107  public static HashFunction murmur3_32(int seed) {
108    return new Murmur3_32HashFunction(seed, /* supplementaryPlaneFix= */ false);
109  }
110
111  /**
112   * Returns a hash function implementing the <a
113   * href="https://github.com/aappleby/smhasher/blob/master/src/MurmurHash3.cpp">32-bit murmur3
114   * algorithm, x86 variant</a> (little-endian variant), using the given seed value, <b>with a known
115   * bug</b> as described in the deprecation text.
116   *
117   * <p>The C++ equivalent is the MurmurHash3_x86_32 function (Murmur3A), which however does not
118   * have the bug.
119   *
120   * @deprecated This implementation produces incorrect hash values from the {@link
121   *     HashFunction#hashString} method if the string contains non-BMP characters. Use {@link
122   *     #murmur3_32_fixed()} instead.
123   */
124  @Deprecated
125  public static HashFunction murmur3_32() {
126    return Murmur3_32HashFunction.MURMUR3_32;
127  }
128
129  /**
130   * Returns a hash function implementing the <a
131   * href="https://github.com/aappleby/smhasher/blob/master/src/MurmurHash3.cpp">32-bit murmur3
132   * algorithm, x86 variant</a> (little-endian variant), using the given seed value.
133   *
134   * <p>The exact C++ equivalent is the MurmurHash3_x86_32 function (Murmur3A).
135   *
136   * <p>This method is called {@code murmur3_32_fixed} because it fixes a bug in the {@code
137   * HashFunction} returned by the original {@code murmur3_32} method.
138   *
139   * @since 31.0
140   */
141  public static HashFunction murmur3_32_fixed(int seed) {
142    return new Murmur3_32HashFunction(seed, /* supplementaryPlaneFix= */ true);
143  }
144
145  /**
146   * Returns a hash function implementing the <a
147   * href="https://github.com/aappleby/smhasher/blob/master/src/MurmurHash3.cpp">32-bit murmur3
148   * algorithm, x86 variant</a> (little-endian variant), using a seed value of zero.
149   *
150   * <p>The exact C++ equivalent is the MurmurHash3_x86_32 function (Murmur3A).
151   *
152   * <p>This method is called {@code murmur3_32_fixed} because it fixes a bug in the {@code
153   * HashFunction} returned by the original {@code murmur3_32} method.
154   *
155   * @since 31.0
156   */
157  public static HashFunction murmur3_32_fixed() {
158    return Murmur3_32HashFunction.MURMUR3_32_FIXED;
159  }
160
161  /**
162   * Returns a hash function implementing the <a
163   * href="https://github.com/aappleby/smhasher/blob/master/src/MurmurHash3.cpp">128-bit murmur3
164   * algorithm, x64 variant</a> (little-endian variant), using the given seed value.
165   *
166   * <p>The exact C++ equivalent is the MurmurHash3_x64_128 function (Murmur3F).
167   */
168  public static HashFunction murmur3_128(int seed) {
169    return new Murmur3_128HashFunction(seed);
170  }
171
172  /**
173   * Returns a hash function implementing the <a
174   * href="https://github.com/aappleby/smhasher/blob/master/src/MurmurHash3.cpp">128-bit murmur3
175   * algorithm, x64 variant</a> (little-endian variant), using a seed value of zero.
176   *
177   * <p>The exact C++ equivalent is the MurmurHash3_x64_128 function (Murmur3F).
178   */
179  public static HashFunction murmur3_128() {
180    return Murmur3_128HashFunction.MURMUR3_128;
181  }
182
183  /**
184   * Returns a hash function implementing the <a href="https://131002.net/siphash/">64-bit
185   * SipHash-2-4 algorithm</a> using a seed value of {@code k = 00 01 02 ...}.
186   *
187   * @since 15.0
188   */
189  public static HashFunction sipHash24() {
190    return SipHashFunction.SIP_HASH_24;
191  }
192
193  /**
194   * Returns a hash function implementing the <a href="https://131002.net/siphash/">64-bit
195   * SipHash-2-4 algorithm</a> using the given seed.
196   *
197   * @since 15.0
198   */
199  public static HashFunction sipHash24(long k0, long k1) {
200    return new SipHashFunction(2, 4, k0, k1);
201  }
202
203  /**
204   * Returns a hash function implementing the MD5 hash algorithm (128 hash bits).
205   *
206   * @deprecated If you must interoperate with a system that requires MD5, then use this method,
207   *     despite its deprecation. But if you can choose your hash function, avoid MD5, which is
208   *     neither fast nor secure. As of January 2017, we suggest:
209   *     <ul>
210   *       <li>For security:
211   *           {@link Hashing#sha256} or a higher-level API.
212   *       <li>For speed: {@link Hashing#goodFastHash}, though see its docs for caveats.
213   *     </ul>
214   */
215  @Deprecated
216  public static HashFunction md5() {
217    return Md5Holder.MD5;
218  }
219
220  private static class Md5Holder {
221    static final HashFunction MD5 = new MessageDigestHashFunction("MD5", "Hashing.md5()");
222  }
223
224  /**
225   * Returns a hash function implementing the SHA-1 algorithm (160 hash bits).
226   *
227   * @deprecated If you must interoperate with a system that requires SHA-1, then use this method,
228   *     despite its deprecation. But if you can choose your hash function, avoid SHA-1, which is
229   *     neither fast nor secure. As of January 2017, we suggest:
230   *     <ul>
231   *       <li>For security:
232   *           {@link Hashing#sha256} or a higher-level API.
233   *       <li>For speed: {@link Hashing#goodFastHash}, though see its docs for caveats.
234   *     </ul>
235   */
236  @Deprecated
237  public static HashFunction sha1() {
238    return Sha1Holder.SHA_1;
239  }
240
241  private static class Sha1Holder {
242    static final HashFunction SHA_1 = new MessageDigestHashFunction("SHA-1", "Hashing.sha1()");
243  }
244
245  /** Returns a hash function implementing the SHA-256 algorithm (256 hash bits). */
246  public static HashFunction sha256() {
247    return Sha256Holder.SHA_256;
248  }
249
250  private static class Sha256Holder {
251    static final HashFunction SHA_256 =
252        new MessageDigestHashFunction("SHA-256", "Hashing.sha256()");
253  }
254
255  /**
256   * Returns a hash function implementing the SHA-384 algorithm (384 hash bits).
257   *
258   * @since 19.0
259   */
260  public static HashFunction sha384() {
261    return Sha384Holder.SHA_384;
262  }
263
264  private static class Sha384Holder {
265    static final HashFunction SHA_384 =
266        new MessageDigestHashFunction("SHA-384", "Hashing.sha384()");
267  }
268
269  /** Returns a hash function implementing the SHA-512 algorithm (512 hash bits). */
270  public static HashFunction sha512() {
271    return Sha512Holder.SHA_512;
272  }
273
274  private static class Sha512Holder {
275    static final HashFunction SHA_512 =
276        new MessageDigestHashFunction("SHA-512", "Hashing.sha512()");
277  }
278
279  /**
280   * Returns a hash function implementing the Message Authentication Code (MAC) algorithm, using the
281   * MD5 (128 hash bits) hash function and the given secret key.
282   *
283   * <p>If you are designing a new system that needs HMAC, prefer {@link #hmacSha256} or other
284   * future-proof algorithms <a
285   * href="https://datatracker.ietf.org/doc/html/rfc6151#section-2.3">over {@code hmacMd5}</a>.
286   *
287   * @param key the secret key
288   * @throws IllegalArgumentException if the given key is inappropriate for initializing this MAC
289   * @since 20.0
290   */
291  public static HashFunction hmacMd5(Key key) {
292    return new MacHashFunction("HmacMD5", key, hmacToString("hmacMd5", key));
293  }
294
295  /**
296   * Returns a hash function implementing the Message Authentication Code (MAC) algorithm, using the
297   * MD5 (128 hash bits) hash function and a {@link SecretKeySpec} created from the given byte array
298   * and the MD5 algorithm.
299   *
300   * <p>If you are designing a new system that needs HMAC, prefer {@link #hmacSha256} or other
301   * future-proof algorithms <a
302   * href="https://datatracker.ietf.org/doc/html/rfc6151#section-2.3">over {@code hmacMd5}</a>.
303   *
304   * @param key the key material of the secret key
305   * @since 20.0
306   */
307  public static HashFunction hmacMd5(byte[] key) {
308    return hmacMd5(new SecretKeySpec(checkNotNull(key), "HmacMD5"));
309  }
310
311  /**
312   * Returns a hash function implementing the Message Authentication Code (MAC) algorithm, using the
313   * SHA-1 (160 hash bits) hash function and the given secret key.
314   *
315   * @param key the secret key
316   * @throws IllegalArgumentException if the given key is inappropriate for initializing this MAC
317   * @since 20.0
318   */
319  public static HashFunction hmacSha1(Key key) {
320    return new MacHashFunction("HmacSHA1", key, hmacToString("hmacSha1", key));
321  }
322
323  /**
324   * Returns a hash function implementing the Message Authentication Code (MAC) algorithm, using the
325   * SHA-1 (160 hash bits) hash function and a {@link SecretKeySpec} created from the given byte
326   * array and the SHA-1 algorithm.
327   *
328   * @param key the key material of the secret key
329   * @since 20.0
330   */
331  public static HashFunction hmacSha1(byte[] key) {
332    return hmacSha1(new SecretKeySpec(checkNotNull(key), "HmacSHA1"));
333  }
334
335  /**
336   * Returns a hash function implementing the Message Authentication Code (MAC) algorithm, using the
337   * SHA-256 (256 hash bits) hash function and the given secret key.
338   *
339   * @param key the secret key
340   * @throws IllegalArgumentException if the given key is inappropriate for initializing this MAC
341   * @since 20.0
342   */
343  public static HashFunction hmacSha256(Key key) {
344    return new MacHashFunction("HmacSHA256", key, hmacToString("hmacSha256", key));
345  }
346
347  /**
348   * Returns a hash function implementing the Message Authentication Code (MAC) algorithm, using the
349   * SHA-256 (256 hash bits) hash function and a {@link SecretKeySpec} created from the given byte
350   * array and the SHA-256 algorithm.
351   *
352   * @param key the key material of the secret key
353   * @since 20.0
354   */
355  public static HashFunction hmacSha256(byte[] key) {
356    return hmacSha256(new SecretKeySpec(checkNotNull(key), "HmacSHA256"));
357  }
358
359  /**
360   * Returns a hash function implementing the Message Authentication Code (MAC) algorithm, using the
361   * SHA-512 (512 hash bits) hash function and the given secret key.
362   *
363   * @param key the secret key
364   * @throws IllegalArgumentException if the given key is inappropriate for initializing this MAC
365   * @since 20.0
366   */
367  public static HashFunction hmacSha512(Key key) {
368    return new MacHashFunction("HmacSHA512", key, hmacToString("hmacSha512", key));
369  }
370
371  /**
372   * Returns a hash function implementing the Message Authentication Code (MAC) algorithm, using the
373   * SHA-512 (512 hash bits) hash function and a {@link SecretKeySpec} created from the given byte
374   * array and the SHA-512 algorithm.
375   *
376   * @param key the key material of the secret key
377   * @since 20.0
378   */
379  public static HashFunction hmacSha512(byte[] key) {
380    return hmacSha512(new SecretKeySpec(checkNotNull(key), "HmacSHA512"));
381  }
382
383  private static String hmacToString(String methodName, Key key) {
384    return "Hashing."
385        + methodName
386        + "(Key[algorithm="
387        + key.getAlgorithm()
388        + ", format="
389        + key.getFormat()
390        + "])";
391  }
392
393  /**
394   * Returns a hash function implementing the CRC32C checksum algorithm (32 hash bits) as described
395   * by RFC 3720, Section 12.1.
396   *
397   * <p>This function is best understood as a <a
398   * href="https://en.wikipedia.org/wiki/Checksum">checksum</a> rather than a true <a
399   * href="https://en.wikipedia.org/wiki/Hash_function">hash function</a>.
400   *
401   * @since 18.0
402   */
403  public static HashFunction crc32c() {
404    return Crc32cHashFunction.CRC_32_C;
405  }
406
407  /**
408   * Returns a hash function implementing the CRC-32 checksum algorithm (32 hash bits).
409   *
410   * <p>To get the {@code long} value equivalent to {@link Checksum#getValue()} for a {@code
411   * HashCode} produced by this function, use {@link HashCode#padToLong()}.
412   *
413   * <p>This function is best understood as a <a
414   * href="https://en.wikipedia.org/wiki/Checksum">checksum</a> rather than a true <a
415   * href="https://en.wikipedia.org/wiki/Hash_function">hash function</a>.
416   *
417   * @since 14.0
418   */
419  public static HashFunction crc32() {
420    return ChecksumType.CRC_32.hashFunction;
421  }
422
423  /**
424   * Returns a hash function implementing the Adler-32 checksum algorithm (32 hash bits).
425   *
426   * <p>To get the {@code long} value equivalent to {@link Checksum#getValue()} for a {@code
427   * HashCode} produced by this function, use {@link HashCode#padToLong()}.
428   *
429   * <p>This function is best understood as a <a
430   * href="https://en.wikipedia.org/wiki/Checksum">checksum</a> rather than a true <a
431   * href="https://en.wikipedia.org/wiki/Hash_function">hash function</a>.
432   *
433   * @since 14.0
434   */
435  public static HashFunction adler32() {
436    return ChecksumType.ADLER_32.hashFunction;
437  }
438
439  @Immutable
440  enum ChecksumType implements ImmutableSupplier<Checksum> {
441    CRC_32("Hashing.crc32()") {
442      @Override
443      public Checksum get() {
444        return new CRC32();
445      }
446    },
447    ADLER_32("Hashing.adler32()") {
448      @Override
449      public Checksum get() {
450        return new Adler32();
451      }
452    };
453
454    public final HashFunction hashFunction;
455
456    ChecksumType(String toString) {
457      this.hashFunction = new ChecksumHashFunction(this, 32, toString);
458    }
459  }
460
461  /**
462   * Returns a hash function implementing FarmHash's Fingerprint64, an open-source algorithm.
463   *
464   * <p>This is designed for generating persistent fingerprints of strings. It isn't
465   * cryptographically secure, but it produces a high-quality hash with fewer collisions than some
466   * alternatives we've used in the past.
467   *
468   * <p>FarmHash fingerprints are encoded by {@link HashCode#asBytes} in little-endian order. This
469   * means {@link HashCode#asLong} is guaranteed to return the same value that
470   * farmhash::Fingerprint64() would for the same input (when compared using {@link
471   * com.google.common.primitives.UnsignedLongs}'s encoding of 64-bit unsigned numbers).
472   *
473   * <p>This function is best understood as a <a
474   * href="https://en.wikipedia.org/wiki/Fingerprint_(computing)">fingerprint</a> rather than a true
475   * <a href="https://en.wikipedia.org/wiki/Hash_function">hash function</a>.
476   *
477   * @since 20.0
478   */
479  public static HashFunction farmHashFingerprint64() {
480    return FarmHashFingerprint64.FARMHASH_FINGERPRINT_64;
481  }
482
483  /**
484   * Returns a hash function implementing the Fingerprint2011 hashing function (64 hash bits).
485   *
486   * <p>This is designed for generating persistent fingerprints of strings. It isn't
487   * cryptographically secure, but it produces a high-quality hash with few collisions. Fingerprints
488   * generated using this are byte-wise identical to those created using the C++ version, but note
489   * that this uses unsigned integers (see {@link com.google.common.primitives.UnsignedInts}).
490   * Comparisons between the two should take this into account.
491   *
492   * <p>Fingerprint2011() is a form of Murmur2 on strings up to 32 bytes and a form of CityHash for
493   * longer strings. It could have been one or the other throughout. The main advantage of the
494   * combination is that CityHash has a bunch of special cases for short strings that don't need to
495   * be replicated here. The result will never be 0 or 1.
496   *
497   * <p>This function is best understood as a <a
498   * href="https://en.wikipedia.org/wiki/Fingerprint_(computing)">fingerprint</a> rather than a true
499   * <a href="https://en.wikipedia.org/wiki/Hash_function">hash function</a>.
500   *
501   * @since 31.1
502   */
503  public static HashFunction fingerprint2011() {
504    return Fingerprint2011.FINGERPRINT_2011;
505  }
506
507  /**
508   * Assigns to {@code hashCode} a "bucket" in the range {@code [0, buckets)}, in a uniform manner
509   * that minimizes the need for remapping as {@code buckets} grows. That is, {@code
510   * consistentHash(h, n)} equals:
511   *
512   * <ul>
513   *   <li>{@code n - 1}, with approximate probability {@code 1/n}
514   *   <li>{@code consistentHash(h, n - 1)}, otherwise (probability {@code 1 - 1/n})
515   * </ul>
516   *
517   * <p>This method is suitable for the common use case of dividing work among buckets that meet the
518   * following conditions:
519   *
520   * <ul>
521   *   <li>You want to assign the same fraction of inputs to each bucket.
522   *   <li>When you reduce the number of buckets, you can accept that the most recently added
523   *       buckets will be removed first. More concretely, if you are dividing traffic among tasks,
524   *       you can decrease the number of tasks from 15 and 10, killing off the final 5 tasks, and
525   *       {@code consistentHash} will handle it. If, however, you are dividing traffic among
526   *       servers {@code alpha}, {@code bravo}, and {@code charlie} and you occasionally need to
527   *       take each of the servers offline, {@code consistentHash} will be a poor fit: It provides
528   *       no way for you to specify which of the three buckets is disappearing. Thus, if your
529   *       buckets change from {@code [alpha, bravo, charlie]} to {@code [bravo, charlie]}, it will
530   *       assign all the old {@code alpha} traffic to {@code bravo} and all the old {@code bravo}
531   *       traffic to {@code charlie}, rather than letting {@code bravo} keep its traffic.
532   * </ul>
533   *
534   * <p>See the <a href="http://en.wikipedia.org/wiki/Consistent_hashing">Wikipedia article on
535   * consistent hashing</a> for more information.
536   */
537  public static int consistentHash(HashCode hashCode, int buckets) {
538    return consistentHash(hashCode.padToLong(), buckets);
539  }
540
541  /**
542   * Assigns to {@code input} a "bucket" in the range {@code [0, buckets)}, in a uniform manner that
543   * minimizes the need for remapping as {@code buckets} grows. That is, {@code consistentHash(h,
544   * n)} equals:
545   *
546   * <ul>
547   *   <li>{@code n - 1}, with approximate probability {@code 1/n}
548   *   <li>{@code consistentHash(h, n - 1)}, otherwise (probability {@code 1 - 1/n})
549   * </ul>
550   *
551   * <p>This method is suitable for the common use case of dividing work among buckets that meet the
552   * following conditions:
553   *
554   * <ul>
555   *   <li>You want to assign the same fraction of inputs to each bucket.
556   *   <li>When you reduce the number of buckets, you can accept that the most recently added
557   *       buckets will be removed first. More concretely, if you are dividing traffic among tasks,
558   *       you can decrease the number of tasks from 15 and 10, killing off the final 5 tasks, and
559   *       {@code consistentHash} will handle it. If, however, you are dividing traffic among
560   *       servers {@code alpha}, {@code bravo}, and {@code charlie} and you occasionally need to
561   *       take each of the servers offline, {@code consistentHash} will be a poor fit: It provides
562   *       no way for you to specify which of the three buckets is disappearing. Thus, if your
563   *       buckets change from {@code [alpha, bravo, charlie]} to {@code [bravo, charlie]}, it will
564   *       assign all the old {@code alpha} traffic to {@code bravo} and all the old {@code bravo}
565   *       traffic to {@code charlie}, rather than letting {@code bravo} keep its traffic.
566   * </ul>
567   *
568   * <p>See the <a href="http://en.wikipedia.org/wiki/Consistent_hashing">Wikipedia article on
569   * consistent hashing</a> for more information.
570   */
571  public static int consistentHash(long input, int buckets) {
572    checkArgument(buckets > 0, "buckets must be positive: %s", buckets);
573    LinearCongruentialGenerator generator = new LinearCongruentialGenerator(input);
574    int candidate = 0;
575    int next;
576
577    // Jump from bucket to bucket until we go out of range
578    while (true) {
579      next = (int) ((candidate + 1) / generator.nextDouble());
580      if (next >= 0 && next < buckets) {
581        candidate = next;
582      } else {
583        return candidate;
584      }
585    }
586  }
587
588  /**
589   * Returns a hash code, having the same bit length as each of the input hash codes, that combines
590   * the information of these hash codes in an ordered fashion. That is, whenever two equal hash
591   * codes are produced by two calls to this method, it is <i>as likely as possible</i> that each
592   * was computed from the <i>same</i> input hash codes in the <i>same</i> order.
593   *
594   * @throws IllegalArgumentException if {@code hashCodes} is empty, or the hash codes do not all
595   *     have the same bit length
596   */
597  public static HashCode combineOrdered(Iterable<HashCode> hashCodes) {
598    Iterator<HashCode> iterator = hashCodes.iterator();
599    checkArgument(iterator.hasNext(), "Must be at least 1 hash code to combine.");
600    int bits = iterator.next().bits();
601    byte[] resultBytes = new byte[bits / 8];
602    for (HashCode hashCode : hashCodes) {
603      byte[] nextBytes = hashCode.asBytes();
604      checkArgument(
605          nextBytes.length == resultBytes.length, "All hashcodes must have the same bit length.");
606      for (int i = 0; i < nextBytes.length; i++) {
607        resultBytes[i] = (byte) (resultBytes[i] * 37 ^ nextBytes[i]);
608      }
609    }
610    return HashCode.fromBytesNoCopy(resultBytes);
611  }
612
613  /**
614   * Returns a hash code, having the same bit length as each of the input hash codes, that combines
615   * the information of these hash codes in an unordered fashion. That is, whenever two equal hash
616   * codes are produced by two calls to this method, it is <i>as likely as possible</i> that each
617   * was computed from the <i>same</i> input hash codes in <i>some</i> order.
618   *
619   * @throws IllegalArgumentException if {@code hashCodes} is empty, or the hash codes do not all
620   *     have the same bit length
621   */
622  public static HashCode combineUnordered(Iterable<HashCode> hashCodes) {
623    Iterator<HashCode> iterator = hashCodes.iterator();
624    checkArgument(iterator.hasNext(), "Must be at least 1 hash code to combine.");
625    byte[] resultBytes = new byte[iterator.next().bits() / 8];
626    for (HashCode hashCode : hashCodes) {
627      byte[] nextBytes = hashCode.asBytes();
628      checkArgument(
629          nextBytes.length == resultBytes.length, "All hashcodes must have the same bit length.");
630      for (int i = 0; i < nextBytes.length; i++) {
631        resultBytes[i] += nextBytes[i];
632      }
633    }
634    return HashCode.fromBytesNoCopy(resultBytes);
635  }
636
637  /** Checks that the passed argument is positive, and ceils it to a multiple of 32. */
638  static int checkPositiveAndMakeMultipleOf32(int bits) {
639    checkArgument(bits > 0, "Number of bits must be positive");
640    return (bits + 31) & ~31;
641  }
642
643  /**
644   * Returns a hash function which computes its hash code by concatenating the hash codes of the
645   * underlying hash functions together. This can be useful if you need to generate hash codes of a
646   * specific length.
647   *
648   * <p>For example, if you need 1024-bit hash codes, you could join two {@link Hashing#sha512} hash
649   * functions together: {@code Hashing.concatenating(Hashing.sha512(), Hashing.sha512())}.
650   *
651   * @since 19.0
652   */
653  public static HashFunction concatenating(
654      HashFunction first, HashFunction second, HashFunction... rest) {
655    // We can't use Lists.asList() here because there's no hash->collect dependency
656    List<HashFunction> list = new ArrayList<>();
657    list.add(first);
658    list.add(second);
659    Collections.addAll(list, rest);
660    return new ConcatenatedHashFunction(list.toArray(new HashFunction[0]));
661  }
662
663  /**
664   * Returns a hash function which computes its hash code by concatenating the hash codes of the
665   * underlying hash functions together. This can be useful if you need to generate hash codes of a
666   * specific length.
667   *
668   * <p>For example, if you need 1024-bit hash codes, you could join two {@link Hashing#sha512} hash
669   * functions together: {@code Hashing.concatenating(Hashing.sha512(), Hashing.sha512())}.
670   *
671   * @since 19.0
672   */
673  public static HashFunction concatenating(Iterable<HashFunction> hashFunctions) {
674    checkNotNull(hashFunctions);
675    // We can't use Iterables.toArray() here because there's no hash->collect dependency
676    List<HashFunction> list = new ArrayList<>();
677    for (HashFunction hashFunction : hashFunctions) {
678      list.add(hashFunction);
679    }
680    checkArgument(!list.isEmpty(), "number of hash functions (%s) must be > 0", list.size());
681    return new ConcatenatedHashFunction(list.toArray(new HashFunction[0]));
682  }
683
684  private static final class ConcatenatedHashFunction extends AbstractCompositeHashFunction {
685
686    private ConcatenatedHashFunction(HashFunction... functions) {
687      super(functions);
688      for (HashFunction function : functions) {
689        checkArgument(
690            function.bits() % 8 == 0,
691            "the number of bits (%s) in hashFunction (%s) must be divisible by 8",
692            function.bits(),
693            function);
694      }
695    }
696
697    @Override
698    HashCode makeHash(Hasher[] hashers) {
699      byte[] bytes = new byte[bits() / 8];
700      int i = 0;
701      for (Hasher hasher : hashers) {
702        HashCode newHash = hasher.hash();
703        i += newHash.writeBytesTo(bytes, i, newHash.bits() / 8);
704      }
705      return HashCode.fromBytesNoCopy(bytes);
706    }
707
708    @Override
709    public int bits() {
710      int bitSum = 0;
711      for (HashFunction function : functions) {
712        bitSum += function.bits();
713      }
714      return bitSum;
715    }
716
717    @Override
718    public boolean equals(@CheckForNull Object object) {
719      if (object instanceof ConcatenatedHashFunction) {
720        ConcatenatedHashFunction other = (ConcatenatedHashFunction) object;
721        return Arrays.equals(functions, other.functions);
722      }
723      return false;
724    }
725
726    @Override
727    public int hashCode() {
728      return Arrays.hashCode(functions);
729    }
730  }
731
732  /**
733   * Linear CongruentialGenerator to use for consistent hashing. See
734   * http://en.wikipedia.org/wiki/Linear_congruential_generator
735   */
736  private static final class LinearCongruentialGenerator {
737    private long state;
738
739    public LinearCongruentialGenerator(long seed) {
740      this.state = seed;
741    }
742
743    public double nextDouble() {
744      state = 2862933555777941757L * state + 1;
745      return ((double) ((int) (state >>> 33) + 1)) / 0x1.0p31;
746    }
747  }
748
749  private Hashing() {}
750}