1 /*
2 * Copyright (c) 2012, Oracle and/or its affiliates. All rights reserved.
3 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
4 *
5 * This code is free software; you can redistribute it and/or modify it
6 * under the terms of the GNU General Public License version 2 only, as
7 * published by the Free Software Foundation. Oracle designates this
8 * particular file as subject to the "Classpath" exception as provided
9 * by Oracle in the LICENSE file that accompanied this code.
10 *
11 * This code is distributed in the hope that it will be useful, but WITHOUT
12 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
13 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
14 * version 2 for more details (a copy is included in the LICENSE file that
15 * accompanied this code).
16 *
17 * You should have received a copy of the GNU General Public License version
18 * 2 along with this work; if not, write to the Free Software Foundation,
19 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
20 *
21 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
22 * or visit www.oracle.com if you need additional information or have any
23 * questions.
24 */
25 package java.util.stream;
26
27 /**
28 * An operation in a stream pipeline that takes a stream as input and produces
29 * a stream, possibly of a different type, as output. An intermediate operation
30 * has an input type and an output type (and, an associated input shape and
31 * output shape). An intermediate operation also has a set of <em>operation
32 * flags</em> that describes how it transforms characteristics of the stream
33 * (such as sortedness or size; see {@link StreamOpFlag}).
34 *
35 * <p>Intermediate operations are implemented in terms of <em>sink transforms
36 * </em>; given a {@code Sink} for the output type of the operation, produce a
37 * {@code Sink} for the input type of the operation, which, when fed with
38 * values, has the effect of implementing the desired operation on the input
39 * values and feeding them to the output sink.
40 *
41 * <p>Some intermediate operations are <em>stateful</em>. This means that the
42 * sinks they produce as a result of the above wrapping may maintain state from
43 * processing earlier elements. Stateful intermediate operations must implement
44 * the {@link StatefulOp} interface. Statefulness has an effect on how the
45 * operation can be parallelized. Stateless operations parallelize trivially
46 * because they are homomorphisms under concatenation:
47 *
48 * <pre>
49 * statelessOp(a || b) = statelessOp(a) || statelessOp(b)
50 * </pre>
51 *
52 * where {@code ||} denotes concatenation. Stateful operations may still be
53 * parallelizable, but are not amenable to the automatic parallelization of
54 * stateless operations. Accordingly, a stateful operation must provide its own
55 * parallel execution implementation
56 * ({@link StatefulOp#evaluateParallel(PipelineHelper)}).
57 *
58 * @apiNote
59 * As an example, consider the stream pipeline:
60 * <pre>
61 * int oldestBob = people.stream().
62 * filter(p -> p.getFirstName.equals("Bob")).
63 * map(p -> p.getAge()).max();
64 * </pre>
65 *
66 * <p>This pipeline has two intermediate operations, filter and map. The
67 * filtering operation has input and output types of {@code Person} (with input
68 * and output shape of {@code REFERENCE}), and the mapping operation has an
69 * input type of {@code Person} and an output type of {@code Integer} (with
70 * shape {@code INT_VALUE}.) When we construct a sink chain, the mapping
71 * operation will be asked to transform a {@code Sink.OfInt} which computes the
72 * maximum value into a {@code Sink} which accepts {@code Person} objects, and
73 * whose behavior is to take the supplied {@code Person}, call {@code getAge()}
74 * on it, and pass the resulting value to the downstream sink. This sink
75 * transform might be implement as:
76 *
77 * <pre>
78 * new Sink.ChainedReference<U>(sink) {
79 * public void accept(U u) {
80 * downstream.accept(mappingFunction.applyAsInt(u));
81 * }
82 * }
83 * </pre>
84 *
85 * @param <E_IN> Type of input elements to the operation
86 * @param <E_OUT> Type of output elements to the operation
87 * @see TerminalOp
88 * @see StatefulOp
89 * @since 1.8
90 */
91 interface IntermediateOp<E_IN, E_OUT> {
92
93 /**
94 * Gets the shape of the input type of this operation
95 *
96 * @implSpec The default returns {@code StreamShape.REFERENCE}
97 * @return Shape of the input type of this operation
98 */
99 default StreamShape inputShape() { return StreamShape.REFERENCE; }
100
101 /**
102 * Gets the shape of the output type of this operation
103 *
104 * @implSpec The default returns {@code StreamShape.REFERENCE}
105 * @return Shape of the output type of this operation
106 */
107 default StreamShape outputShape() { return StreamShape.REFERENCE; }
108
109 /**
110 * Gets the operation flags of this operation.
111 *
112 * @implSpec The default returns {@code 0}
113 * @return a bitmap describing the operation flags of this operation
114 * @see StreamOpFlag
115 */
116 default int getOpFlags() { return 0; }
117
118 /**
119 * Returns whether this operation is stateful or not. If it is stateful,
120 * then the method {@link #evaluateParallel(PipelineHelper)} must be
121 * overridden.
122 *
123 * @implSpec The default implementation returns {@code false}.
124 * @return {@code true} if this operation is stateful
125 */
126 default boolean isStateful() { return false; }
127
128 /**
129 * Accepts a {@code Sink} which will receive the results of this operation,
130 * and return a {@code Sink} which accepts elements of the input type of
131 * this operation and which performs the operation, passing the results to
132 * the provided {@code Sink}.
133 *
134 * <p>The implementation may use the {@code flags} parameter to optimize the
135 * sink wrapping. For example, if the input is already {@code DISTINCT},
136 * the implementation for the {@code Stream#distinct()} method could just
137 * return the sink it was passed.
138 *
139 * @param flags The combined stream and operation flags up to, but not
140 * including, this operation.
141 * @param sink elements will be sent to this sink after the processing.
142 * @return a sink which will accept elements and perform the operation upon
143 * each element, passing the results (if any) to the provided
144 * {@code Sink}.
145 */
146 Sink<E_IN> wrapSink(int flags, Sink<E_OUT> sink);
147
148 /**
149 * Performs a parallel evaluation of the operation using the specified
150 * {@code PipelineHelper}, which describes the stream source and upstream
151 * intermediate operations. Only called on stateful operations. If
152 * {@link #isStateful()} returns true then implementations must override the
153 * default implementation.
154 *
155 * @implSpec The default implementation throws an
156 * {@link UnsupportedOperationException}
157 *
158 * @param helper the pipeline helper
159 * @param <P_IN> the type of elements in the pipeline source
160 * @return a {@code Node} describing the result of the evaluation
161 */
162 default <P_IN> Node<E_OUT> evaluateParallel(PipelineHelper<P_IN, E_OUT> helper) {
163 throw new UnsupportedOperationException("Parallel evaluation is not supported");
164 }
165 }