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 com.google.common.annotations.Beta;
018import com.google.common.primitives.Ints;
019import com.google.errorprone.annotations.Immutable;
020import java.nio.ByteBuffer;
021import java.nio.charset.Charset;
022
023/**
024 * A hash function is a collision-averse pure function that maps an arbitrary block of data to a
025 * number called a <i>hash code</i>.
026 *
027 * <h3>Definition</h3>
028 *
029 * <p>Unpacking this definition:
030 *
031 * <ul>
032 *   <li><b>block of data:</b> the input for a hash function is always, in concept, an ordered byte
033 *       array. This hashing API accepts an arbitrary sequence of byte and multibyte values (via
034 *       {@link Hasher}), but this is merely a convenience; these are always translated into raw
035 *       byte sequences under the covers.
036 *   <li><b>hash code:</b> each hash function always yields hash codes of the same fixed bit length
037 *       (given by {@link #bits}). For example, {@link Hashing#sha1} produces a 160-bit number,
038 *       while {@link Hashing#murmur3_32()} yields only 32 bits. Because a {@code long} value is
039 *       clearly insufficient to hold all hash code values, this API represents a hash code as an
040 *       instance of {@link HashCode}.
041 *   <li><b>pure function:</b> the value produced must depend only on the input bytes, in the order
042 *       they appear. Input data is never modified. {@link HashFunction} instances should always be
043 *       stateless, and therefore thread-safe.
044 *   <li><b>collision-averse:</b> while it can't be helped that a hash function will sometimes
045 *       produce the same hash code for distinct inputs (a "collision"), every hash function strives
046 *       to <i>some</i> degree to make this unlikely. (Without this condition, a function that
047 *       always returns zero could be called a hash function. It is not.)
048 * </ul>
049 *
050 * <p>Summarizing the last two points: "equal yield equal <i>always</i>; unequal yield unequal
051 * <i>often</i>." This is the most important characteristic of all hash functions.
052 *
053 * <h3>Desirable properties</h3>
054 *
055 * <p>A high-quality hash function strives for some subset of the following virtues:
056 *
057 * <ul>
058 *   <li><b>collision-resistant:</b> while the definition above requires making at least <i>some</i>
059 *       token attempt, one measure of the quality of a hash function is <i>how well</i> it succeeds
060 *       at this goal. Important note: it may be easy to achieve the theoretical minimum collision
061 *       rate when using completely <i>random</i> sample input. The true test of a hash function is
062 *       how it performs on representative real-world data, which tends to contain many hidden
063 *       patterns and clumps. The goal of a good hash function is to stamp these patterns out as
064 *       thoroughly as possible.
065 *   <li><b>bit-dispersing:</b> masking out any <i>single bit</i> from a hash code should yield only
066 *       the expected <i>twofold</i> increase to all collision rates. Informally, the "information"
067 *       in the hash code should be as evenly "spread out" through the hash code's bits as possible.
068 *       The result is that, for example, when choosing a bucket in a hash table of size 2^8,
069 *       <i>any</i> eight bits could be consistently used.
070 *   <li><b>cryptographic:</b> certain hash functions such as {@link Hashing#sha512} are designed to
071 *       make it as infeasible as possible to reverse-engineer the input that produced a given hash
072 *       code, or even to discover <i>any</i> two distinct inputs that yield the same result. These
073 *       are called <i>cryptographic hash functions</i>. But, whenever it is learned that either of
074 *       these feats has become computationally feasible, the function is deemed "broken" and should
075 *       no longer be used for secure purposes. (This is the likely eventual fate of <i>all</i>
076 *       cryptographic hashes.)
077 *   <li><b>fast:</b> perhaps self-explanatory, but often the most important consideration. We have
078 *       published <a href="#noWeHaventYet">microbenchmark results</a> for many common hash
079 *       functions.
080 * </ul>
081 *
082 * <h3>Providing input to a hash function</h3>
083 *
084 * <p>The primary way to provide the data that your hash function should act on is via a {@link
085 * Hasher}. Obtain a new hasher from the hash function using {@link #newHasher}, "push" the relevant
086 * data into it using methods like {@link Hasher#putBytes(byte[])}, and finally ask for the {@code
087 * HashCode} when finished using {@link Hasher#hash}. (See an {@linkplain #newHasher example} of
088 * this.)
089 *
090 * <p>If all you want to hash is a single byte array, string or {@code long} value, there are
091 * convenient shortcut methods defined directly on {@link HashFunction} to make this easier.
092 *
093 * <p>Hasher accepts primitive data types, but can also accept any Object of type {@code T} provided
094 * that you implement a {@link Funnel}{@code <T>} to specify how to "feed" data from that object
095 * into the function. (See {@linkplain Hasher#putObject an example} of this.)
096 *
097 * <p><b>Compatibility note:</b> Throughout this API, multibyte values are always interpreted in
098 * <i>little-endian</i> order. That is, hashing the byte array {@code {0x01, 0x02, 0x03, 0x04}} is
099 * equivalent to hashing the {@code int} value {@code 0x04030201}. If this isn't what you need,
100 * methods such as {@link Integer#reverseBytes} and {@link Ints#toByteArray} will help.
101 *
102 * <h3>Relationship to {@link Object#hashCode}</h3>
103 *
104 * <p>Java's baked-in concept of hash codes is constrained to 32 bits, and provides no separation
105 * between hash algorithms and the data they act on, so alternate hash algorithms can't be easily
106 * substituted. Also, implementations of {@code hashCode} tend to be poor-quality, in part because
107 * they end up depending on <i>other</i> existing poor-quality {@code hashCode} implementations,
108 * including those in many JDK classes.
109 *
110 * <p>{@code Object.hashCode} implementations tend to be very fast, but have weak collision
111 * prevention and <i>no</i> expectation of bit dispersion. This leaves them perfectly suitable for
112 * use in hash tables, because extra collisions cause only a slight performance hit, while poor bit
113 * dispersion is easily corrected using a secondary hash function (which all reasonable hash table
114 * implementations in Java use). For the many uses of hash functions beyond data structures,
115 * however, {@code Object.hashCode} almost always falls short -- hence this library.
116 *
117 * @author Kevin Bourrillion
118 * @since 11.0
119 */
120@Beta
121@Immutable
122public interface HashFunction {
123  /**
124   * Begins a new hash code computation by returning an initialized, stateful {@code Hasher}
125   * instance that is ready to receive data. Example:
126   *
127   * <pre>{@code
128   * HashFunction hf = Hashing.md5();
129   * HashCode hc = hf.newHasher()
130   *     .putLong(id)
131   *     .putBoolean(isActive)
132   *     .hash();
133   * }</pre>
134   */
135  Hasher newHasher();
136
137  /**
138   * Begins a new hash code computation as {@link #newHasher()}, but provides a hint of the expected
139   * size of the input (in bytes). This is only important for non-streaming hash functions (hash
140   * functions that need to buffer their whole input before processing any of it).
141   */
142  Hasher newHasher(int expectedInputSize);
143
144  /**
145   * Shortcut for {@code newHasher().putInt(input).hash()}; returns the hash code for the given
146   * {@code int} value, interpreted in little-endian byte order. The implementation <i>might</i>
147   * perform better than its longhand equivalent, but should not perform worse.
148   *
149   * @since 12.0
150   */
151  HashCode hashInt(int input);
152
153  /**
154   * Shortcut for {@code newHasher().putLong(input).hash()}; returns the hash code for the given
155   * {@code long} value, interpreted in little-endian byte order. The implementation <i>might</i>
156   * perform better than its longhand equivalent, but should not perform worse.
157   */
158  HashCode hashLong(long input);
159
160  /**
161   * Shortcut for {@code newHasher().putBytes(input).hash()}. The implementation <i>might</i>
162   * perform better than its longhand equivalent, but should not perform worse.
163   */
164  HashCode hashBytes(byte[] input);
165
166  /**
167   * Shortcut for {@code newHasher().putBytes(input, off, len).hash()}. The implementation
168   * <i>might</i> perform better than its longhand equivalent, but should not perform worse.
169   *
170   * @throws IndexOutOfBoundsException if {@code off < 0} or {@code off + len > bytes.length} or
171   *     {@code len < 0}
172   */
173  HashCode hashBytes(byte[] input, int off, int len);
174
175  /**
176   * Shortcut for {@code newHasher().putBytes(input).hash()}. The implementation <i>might</i>
177   * perform better than its longhand equivalent, but should not perform worse.
178   *
179   * @since 23.0
180   */
181  HashCode hashBytes(ByteBuffer input);
182
183  /**
184   * Shortcut for {@code newHasher().putUnencodedChars(input).hash()}. The implementation
185   * <i>might</i> perform better than its longhand equivalent, but should not perform worse. Note
186   * that no character encoding is performed; the low byte and high byte of each {@code char} are
187   * hashed directly (in that order).
188   *
189   * <p><b>Warning:</b> This method will produce different output than most other languages do when
190   * running the same hash function on the equivalent input. For cross-language compatibility, use
191   * {@link #hashString}, usually with a charset of UTF-8. For other use cases, use {@code
192   * hashUnencodedChars}.
193   *
194   * @since 15.0 (since 11.0 as hashString(CharSequence)).
195   */
196  HashCode hashUnencodedChars(CharSequence input);
197
198  /**
199   * Shortcut for {@code newHasher().putString(input, charset).hash()}. Characters are encoded using
200   * the given {@link Charset}. The implementation <i>might</i> perform better than its longhand
201   * equivalent, but should not perform worse.
202   *
203   * <p><b>Warning:</b> This method, which reencodes the input before hashing it, is useful only for
204   * cross-language compatibility. For other use cases, prefer {@link #hashUnencodedChars}, which is
205   * faster, produces the same output across Java releases, and hashes every {@code char} in the
206   * input, even if some are invalid.
207   */
208  HashCode hashString(CharSequence input, Charset charset);
209
210  /**
211   * Shortcut for {@code newHasher().putObject(instance, funnel).hash()}. The implementation
212   * <i>might</i> perform better than its longhand equivalent, but should not perform worse.
213   *
214   * @since 14.0
215   */
216  <T> HashCode hashObject(T instance, Funnel<? super T> funnel);
217
218  /**
219   * Returns the number of bits (a multiple of 32) that each hash code produced by this hash
220   * function has.
221   */
222  int bits();
223}