Flatten a Map<Integer, List<String>> to Map<String, Integer> with stream and lambda

Question

I would like to flatten a Map which associates an Integer key to a list of String, without losing the key mapping. I am curious as though it is possible and useful to do so with stream and lambda.

We start with something like this:

Map<Integer, List<String>> mapFrom = new HashMap<>(); 

Let's assume that mapFrom is populated somewhere, and looks like:

1: a,b,c 2: d,e,f etc. 

Let's also assume that the values in the lists are unique.

Now, I want to "unfold" it to get a second map like:

a: 1 b: 1 c: 1 d: 2 e: 2 f: 2 etc. 

I could do it like this (or very similarly, using foreach):

Map<String, Integer> mapTo = new HashMap<>(); for (Map.Entry<Integer, List<String>> entry: mapFrom.entrySet()) {     for (String s: entry.getValue()) {         mapTo.put(s, entry.getKey());     } } 

Now let's assume that I want to use lambda instead of nested for loops. I would probably do something like this:

Map<String, Integer> mapTo = mapFrom.entrySet().stream().map(e -> {     e.getValue().stream().?     // Here I can iterate on each List,      // but my best try would only give me a flat map for each key,      // that I wouldn't know how to flatten. }).collect(Collectors.toMap(/*A String value*/,/*An Integer key*/)) 

I also gave a try to flatMap, but I don't think that it is the right way to go, because although it helps me get rid of the dimensionality issue, I lose the key in the process.

In a nutshell, my two questions are :

• Is it possible to use streams and lambda to achieve this?
• Is is useful (performance, readability) to do so?

Answer 1

You need to use flatMap to flatten the values into a new stream, but since you still need the original keys for collecting into a Map, you have to map to a temporary object holding key and value, e.g.

Map<String, Integer> mapTo = mapFrom.entrySet().stream()        .flatMap(e->e.getValue().stream()                     .map(v->new AbstractMap.SimpleImmutableEntry<>(e.getKey(), v)))        .collect(Collectors.toMap(Map.Entry::getValue, Map.Entry::getKey)); 

The Map.Entry is a stand-in for the nonexistent tuple type, any other type capable of holding two objects of different type is sufficient.

An alternative not requiring these temporary objects, is a custom collector:

Map<String, Integer> mapTo = mapFrom.entrySet().stream().collect(     HashMap::new, (m,e)->e.getValue().forEach(v->m.put(v, e.getKey())), Map::putAll); 

This differs from toMap in overwriting duplicate keys silently, whereas toMap without a merger function will throw an exception, if there is a duplicate key. Basically, this custom collector is a parallel capable variant of

Map<String, Integer> mapTo = new HashMap<>(); mapFrom.forEach((k, l) -> l.forEach(v -> mapTo.put(v, k))); 

But note that this task wouldn’t benefit from parallel processing, even with a very large input map. Only if there were additional computational intense task within the stream pipeline that could benefit from SMP, there was a chance of getting a benefit from parallel streams. So perhaps, the concise, sequential Collection API solution is preferable.