问题
I have a question on the intermediate stages sequential state - are the operations from a stage applied to all the input stream (items) or are all the stages / operations applied to each stream item?
I'm aware the question might not be easy to understand, so I'll give an example. On the following stream processing:
List<String> strings = Arrays.asList("Are Java streams intermediate stages sequential?".split(" "));
strings.stream()
.filter(word -> word.length() > 4)
.peek(word -> System.out.println("f: " + word))
.map(word -> word.length())
.peek(length -> System.out.println("m: " + length))
.forEach(length -> System.out.println("-> " + length + "\n"));
My expectation for this code is that it will output:
f: streams
f: intermediate
f: stages
f: sequential?
m: 7
m: 12
m: 6
m: 11
-> 7
-> 12
-> 6
-> 11
Instead, the output is:
f: streams
m: 7
-> 7
f: intermediate
m: 12
-> 12
f: stages
m: 6
-> 6
f: sequential?
m: 11
-> 11
Are the items just displayed for all the stages, due to the console output? Or are they also processed for all the stages, one at a time?
I can further detail the question, if it's not clear enough.
回答1:
This behaviour enables optimisation of the code. If each intermediate operation were to process all elements of a stream before proceeding to the next intermediate operation then there would be no chance of optimisation.
So to answer your question, each element moves along the stream pipeline vertically one at a time (except for some stateful operations discussed later), therefore enabling optimisation where possible.
Explanation
Given the example you've provided, each element will move along the stream pipeline vertically one by one as there is no stateful operation included.
Another example, say you were looking for the first String
whose length is greater than 4
, processing all the elements prior to providing the result is unnecessary and time-consuming.
Consider this simple illustration:
List<String> stringsList = Arrays.asList("1","12","123","1234","12345","123456","1234567");
int result = stringsList.stream()
.filter(s -> s.length() > 4)
.mapToInt(Integer::valueOf)
.findFirst().orElse(0);
The filter
intermediate operation above will not find all the elements whose length is greater than 4
and return a new stream of them but rather what happens is as soon as we find the first element whose length is greater than 4
, that element goes through to the .mapToInt
which then findFirst
says "I've found the first element" and execution stops there. Therefore the result will be 12345
.
Behaviour of stateful and stateless intermediate operations
Note that when a stateful intermediate operation as such of sorted
is included in a stream pipeline then that specific operation will traverse the entire stream. If you think about it, this makes complete sense as in order to sort elements you'll need to see all the elements to determine which elements come first in the sort order.
The distinct
intermediate operation is also a stateful operation, however, as @Holger has mentioned unlike sorted
, it does not require traversing the entire stream as each distinct element can get passed down the pipeline immediately and may fulfil a short-circuiting condition.
stateless intermediate operations such as filter
, map
etc do not have to traverse the entire stream and can freely process one element at a time vertically as mentioned above.
Lastly, but not least it's also important to note that, when the terminal operation is a short-circuiting operation the terminal-short-circuiting methods can finish before traversing all the elements of the underlying stream.
reading: Java 8 stream tutorial
回答2:
Your answer is
loop fusion
. What we see is that the four intermediate operationsfilter() – peek() – map() – peek() – println using forEach() which is a kinda terminal operation
have been logically joined together to constitute a single pass. They are executed in order for each of the individual element. This joining together of operations in a single pass is an optimization technique known asloop fusion
.
More for reading: Source
回答3:
An intermediate operation is always lazily executed. That is to say they are not run until the point a terminal operation is reached. A few of the most popular intermediate operations used in a stream
filter – the filter operation returns a stream of elements that
satisfy the predicate passed in as a parameter to the operation. The
elements themselves before and after the filter will have the same
type, however the number of elements will likely change
map – the map operation returns a stream of elements after they have
been processed by the function passed in as a parameter. The
elements before and after the mapping may have a different type, but
there will be the same total number of elements.
distinct – the distinct operation is a special case of the filter
operation. Distinct returns a stream of elements such that each
element is unique in the stream, based on the equals method of the
elements
.java-8-streams-cheat-sheet
回答4:
Apart from optimisation, the order of processing you'd describe wouldn't work for streams of indeterminate length, like this:
DoubleStream.generate(Math::random).filter(d -> d > 0.9).findFirst();
Admittedly this example doesn't make much sense in practice, but the point is that rather than backed by a fixed-size collection,DoubleStream.generate()
creates a potentially infinite stream. The only way to process this is element by element.
来源:https://stackoverflow.com/questions/44862517/are-java-streams-stages-sequential