为什么以可变对象作为身份在并行流中调用reduce方法不会保留结果中的顺序？

有以下看似“正确”的代码：

List<String> list = Arrays.asList("1","2","3","4","5","6",
    "7","8","9","10","11","12");
String result = list.parallelStream()
   .reduce(new StringBuilder(), StringBuilder::append, 
       StringBuilder::append).toString();
System.out.println(result);

The problem of this snippet is that the identity, new StringBuilder(), in the reduce method call is mutable, thus the result is undermined, i.e. the order of the result does not preserve. But I cannot fully understand the reason, thus I am not able to visualize a case of producing the result with in the different order from the original list. So I drew the corresponding map-reduce diagram, which by chance got the result preserving the order:

问题：首先，我想确认一下这张图是否正确。其次，如果这个图是正确的，我想知道这个代码片段并不总是产生的原因在哪里result维持秩序

您的图表不正确 - 您假设每个并行减少都以新的开始StringBuilder。相反，每个并行减少开始具有相同的单位元- 使用相同的 StringBuilder（您创建并作为第一个参数传递给reduce方法）。

每个并行流调用StringBuilder.append在您传递给的（唯一的）StringBuilder 上reduce方法，从而将当前遇到的元素附加到它。

下一步是组合部分结果，也可以通过调用StringBuilder.append在同一个 StringBuilder 上，将 StringBuilder 的内容副本附加到其自身上。

要创建您绘制的图表，您必须传递一个Supplier<StringBuilder>作为归约操作的第一个参数。

实际上，根据 Holger 的评论，使用时这是可能的可变约简.

为此，您不调用reduce方法，而是调用collect method:

List<String> list = Arrays.asList("1","2","3","4","5","6",
    "7","8","9","10","11","12");
String result = list.parallelStream()
   .collect(StringBuilder::new, StringBuilder::append, 
       StringBuilder::append).toString();
System.out.println(result);