Parallel data processing and performance - Java 8 in Action: Lambdas, Streams, and Functional-Style Programming - page 220

This Spliterator is created from the String to be parsed and iterates over its Characters by
holding the index of the one currently being traversed. Let's quickly revisit the methods of the
WordCounterSpliterator implementing the Spliterator interface:
 The tryAdvance method feeds the Consumer with the Character in the String at the current index
position and increments this position. The Consumer passed as argument is an internal Java class
forwarding the consumed Character to the set of functions that have to be applied to it while
traversing the stream, which in this case is only a reducing function, namely, the accumulate method
of the WordCounter class. The tryAdvance method returns true if the new cursor position is less
than the total String length and there are further Character s to be iterated.
 The trySplit method is the most important one in a Spliterator because it's the one defining the
logic used to split the data structure to be iterated. As you did in the compute method of the
RecursiveTask implemented in listing 7.1 (on how to use the fork/join framework), the first thing
you have to do here is set a limit under which you don't want to perform further splits. Here, you use a
very low limit of 10 Character s only to make sure that your program will perform some splits with
the relatively short String you're parsing, but in real-world applications you'll have to use a higher
limit, as you did in the fork/join example, to avoid creating too many tasks. If the number of
remaining Character s to be traversed is under this limit, you return null to signal that no further
split is necessary. Conversely, if you need to perform a split, you set the candidate split position to the
half of the String chunk remaining to be parsed. But you don't use this split position directly because