Mapping a list to a map and vice versa

[filiphr]

The main idea is to offer mapping from a list into a map and vice versa. This could be done by treating the map as an Iterable<Map.Entry<K,V>>.

This would me that we could do something like:

// First example
@Mapper
public interface MyMapper {

    List<FlattenedTarget> map(Map<String, Source> source);

    @Mapping(target = "day", source = "key")
    @Mapping(target = ".", source = "value")
    FlattenedTarget map(Map.Entry<String, Source> entry);
}

// Second example
@Mapper
public interface MyMapper {

    List<FlattenedTarget> map(Map<String, Source> source);

    @Mapping(target = "day", source = "key")
    @Mapping(target = "times", source = "value")
    FlattenedTarget map(Map.Entry<String, Source> entry);
}

and the reverse could look like:

// First example
@Mapper
public interface ReverseMapper {
    Map<String, Target> map(List<FlattenedSource> source);

    @Mapping(target = "key", source = "day")
    @Mapping(target = "value", source = "item")
    Map.Entry<String, Target> map(FlattenedSource item);

    Target map(FlattenedSource item);
}

// Second example
@Mapper
public interface ReverseMapper {
    Map<String, List<String> map(List<FlattenedSource> source);

    @Mapping(target = "key", source = "day")
    @Mapping(target = "value", source = "times")
    Map.Entry<String, List<String> map(FlattenedSource item);
}

To support something like this we would have to provide a custom support for mapping to Map.Entry and to treat Map<K,V> as Iterable<Map.Entry<K,V>> when mapping a map into an iterable.

Discussed in #3263

^{Originally posted by hinrik May 5, 2023}
I sometimes want to flatten a map into a list (and vice versa), e.g. when I'm converting from different representations (e.g. Avro to/from application domain). I have to do the entire mapping with a default method on my mapper interface, duplicating some logic that MapStruct could have taken care of for me.

There are two general use cases I run into.

Flattening a multi-value map

{
  "monday": {
    "times": ["11:00", "16:00"],
    "type": "foo"
  },
  "wednesday": {
    "times": ["14:00", "19:00"],
    "type": "bar"
  }
}

Target:

[
  {"day": "monday", "times": ["11:00", "16:00"], "type": "foo"},
  {"day": "wednesday", "times": ["14:00", "19:00"], "type": "bar"},
]

Here I wish I could do something like @Mapping(target = "schedule", source = "flatSchedule", flatKey = "day"). which would also work if source and target were flipped.

Flattening a single-value map

Source:

{
  "monday": ["11:00", "16:00"],
  "wednesday": ["14:00", "19:00"]
}

Target:

[
  {"day": "monday", "times": ["11:00", "16:00"]},
  {"day": "wednesday", "times": ["14:00", "19:00"]},
]

This example is similar, except here the value in the source map is not a bean, so we need to introduce a key for it in the target map. Here I wish I could do @Mapping(target = "schedule", source = "flatSchedule", flatKey = "day", flatValue = "times").

Would this be worth implementing in MapStruct?