Java Stream: grouping and counting by multiple fields

I have the following object:

class Event {
private LocalDateTime when;
private String what;

public Event(LocalDateTime when, String what) {
  super();
  this.when = when;
  this.what = what;
}

public LocalDateTime getWhen() {
  return when;
}

public void setWhen(LocalDateTime when) {
  this.when = when;
}

public String getWhat() {
  return what;
}

public void setWhat(String what) {
  this.what = what;
}

      

}

I need to aggregate by year / month (yyyy-mm) and event type and then count. For example, the following list

List<Event> events = Arrays.asList(
  new Event(LocalDateTime.parse("2017-03-03T09:01:16.111"), "EVENT1"),
  new Event(LocalDateTime.parse("2017-03-03T09:02:11.222"), "EVENT1"),
  new Event(LocalDateTime.parse("2017-04-03T09:04:11.333"), "EVENT1"), 
  new Event(LocalDateTime.parse("2017-04-03T09:04:11.333"), "EVENT2"),
  new Event(LocalDateTime.parse("2017-04-03T09:06:16.444"), "EVENT2"),
  new Event(LocalDateTime.parse("2017-05-03T09:01:26.555"), "EVENT3")
);

      

should have the following result:

Year/Month  Type  Count
2017-03     EVENT1    2  
2017-04     EVENT1    1
2017-04     EVENT2    2
2017-04     EVENT3    1

      

Any idea if (and if so, how) I can achieve this using the Streams API?

+3


source to share


5 answers


If you don't want to define your own key, you can groupBy

double. The result is the same, but in a slightly different format:

 System.out.println(events.stream()
            .collect(Collectors.groupingBy(e -> YearMonth.from(e.getWhen()),
                    Collectors.groupingBy(Event::getWhat, Collectors.counting()))));

      



And the result:

 {2017-05={EVENT3=1}, 2017-04={EVENT2=2, EVENT1=1}, 2017-03={EVENT1=2}}

      

+3


source


If you don't want to create a new key class like assylias suggested, you can do double groupingBy

Map<YearMonth,Map<String,Long>> map = 
     events.stream()
           .collect(Collectors.groupingBy(e -> YearMonth.from(e.getWhen()),
                    Collectors.groupingBy(x -> x.getWhat(), Collectors.counting()))
                   );

      

... and then nested print

map.forEach((k,v)-> v.forEach((a,b)-> System.out.println(k + " " +  a + " " + b)));

      

Will print

2017-05 EVENT3 1
2017-04 EVENT2 2
2017-04 EVENT1 1
2017-03 EVENT1 2

      




EDIT: I noticed that the order of the dates was the opposite of the OP's expected solution. Using the 3 parameter version groupingBy

, you can specify an ordered map implementation

Map<YearMonth,Map<String,Long>> map = 
     events.stream()
           .collect(Collectors.groupingBy(e -> YearMonth.from(e.getWhen()), TreeMap::new, 
                    Collectors.groupingBy(x -> x.getWhat(), Collectors.counting()))
                   );

      

The same map.forEach(...)

now prints

2017-03 EVENT1 2
2017-04 EVENT2 2
2017-04 EVENT1 1
2017-05 EVENT3 1

      

+3


source


You can create a "key" class that contains the year / month and event type:

class Group {
  private YearMonth ym;
  private String type;

  public Group(Event e) {
    this.ym = YearMonth.from(e.getWhen());
    this.type = e.getWhat();
  }

  //equals, hashCode, toString etc.
}

      

Then you can use this key to group events:

Map<Group, Long> result = events.stream()
                .collect(Collectors.groupingBy(Group::new, Collectors.counting()));
result.forEach((k, v) -> System.out.println(k + "\t" + v));

      

which outputs:

2017-04 EVENT1  1
2017-03 EVENT1  2
2017-04 EVENT2  2
2017-05 EVENT3  1

      

+2


source


final DateTimeFormatter dateTimeFormatter = DateTimeFormatter.ofPattern("yyyy-MM");
    Stream.of(
            new Event(LocalDateTime.parse("2017-03-03T09:01:16.111"), "EVENT1"),
            new Event(LocalDateTime.parse("2017-03-03T09:02:11.222"), "EVENT1"),
            new Event(LocalDateTime.parse("2017-04-03T09:04:11.333"), "EVENT1"),
            new Event(LocalDateTime.parse("2017-04-03T09:04:11.333"), "EVENT2"),
            new Event(LocalDateTime.parse("2017-04-03T09:06:16.444"), "EVENT2"),
            new Event(LocalDateTime.parse("2017-05-03T09:01:26.555"), "EVENT3")
            ).collect(Collectors.groupingBy(event -> 
               dateTimeFormatter.format(event.getWhen()),
               Collectors.groupingBy(Event::getWhat, counting())))
             .forEach((whenDate,v) -> v.forEach((whatKey,counter) -> 
                System.out.println(whenDate+ " "+ whatKey+" "+counter)));

      

You don't need to use the Arrays.asList () method to navigate to the stream. Use Stream.of () method directly to get a stream.

Output

2017-03 EVENT1 2
2017-04 EVENT2 2
2017-04 EVENT1 1
2017-05 EVENT3 1

      

0


source


We can create a method in POJO that contains a list of fields to use for grouping, like below

public String getWhenAndWhat() {
    return YearMonth.from(when) + ":" + what; //you can use delimiters like ':','-',','
}

      

And the stream code,

System.out.println(events.stream()
            .collect(Collectors.groupingBy(Event::getWhenAndWhat, Collectors.counting())));

      

Output:

{2017-05: EVENT3 = 1, 2017-04: EVENT1 = 1, 2017-04: EVENT2 = 2, 2017-03: EVENT1 = 2}

0


source







All Articles