stream map with Collectors.toMap vs Accumulator

问题

I have a a set of enums Set<Task> tasks; and process this to an EnumMap:

Version 1 (using supplier, accumulator, combiner):

return task.stream()
           .collect(() -> new EnumMap<>(Task.class),
                        (map, item) -> map.put(item, compute(item)), 
                        Map::putAll);

Version 2 (using Collectors.toMap):

return tasks.stream().collect(
       Collectors.toMap(
              Function.identity(), 
              t -> compute(t), 
              null, 
              () -> new EnumMap<>(Task.class)));

I like version 1 more because it is shorter. However, I must select the fastest version. Will version 1 work good with parallelStream? Unlike version 2 which uses Collectors.toMap has the following API-note:

Impl Note: For parallel stream pipelines, the combiner function operates by merging the keys from one map into another, which can be an expensive operation.

3 Questions:

1. Does this also happen with version 1?

2. Which one should I select for fastest and accurate processing?

3. And is there a way to eliminate the need for materializing it into an EnumMap? I was thinking about returning only a key-value-stream, but it seems not possible with stream.map(..).

来源：https://stackoverflow.com/questions/61479650/stream-map-with-collectors-tomap-vs-accumulator