#3306 Support for map keys that contain dots when maps are mapped

[thunderhook]

Tackling #3066 for map keys containing dots.
It was tricky to not reintroduce #3144 again.

[filiphr]

@thunderhook how about something like Map<String, Source> if we use dotted then this would work properly right now, right? However, with this PR this will no longer work, right?

[thunderhook]

I was wondering why there was no test handling this. I just tried the following:

@Mapper
public interface MapToBeanFromMapWithSource {

    MapToBeanFromMapWithSource INSTANCE = Mappers.getMapper( MapToBeanFromMapWithSource.class );

    @Mapping(target = "targetName", source = "name")
    // @Mapping(target = "targetName", source = "source.name")
    // @Mapping(target = "targetName", source = "sourceA.name")
    // @Mapping(target = "targetName", source = "source.sourceA.name")
    Target toTarget(Map<String, Source> source);

    class Source {

        private String name;

        public String getName() {
            return name;
        }

        public void setName(String name) {
            this.name = name;
        }
    }

    class Target {

        String targetName;

        public String getTargetName() {
            return targetName;
        }

        public void setTargetName(String targetName) {
            this.targetName = targetName;
        }
    }

}

    @ProcessorTest
    @WithClasses(MapToBeanFromMapWithSource.class)
    void shouldMap() {
        Map<String, MapToBeanFromMapWithSource.Source> sourceMap = new HashMap<>();
        MapToBeanFromMapWithSource.Source sourceA = new MapToBeanFromMapWithSource.Source();
        sourceA.setName( "value" );
        sourceMap.put( "sourceA", sourceA );
        sourceMap.put( "sourceB", new MapToBeanFromMapWithSource.Source() );

        MapToBeanFromMapWithSource.Target target = MapToBeanFromMapWithSource.INSTANCE.toTarget( sourceMap );

        assertThat( target.getTargetName() ).isEqualTo( "value" );
    }

And no matter what, it always shows:

DiagnosticDescriptor: ERROR MapToBeanFromMapWithSource.java:22
Can't map property "MapToBeanFromMapWithSource.Source name" to "String targetName".
Consider to declare/implement a mapping method: "String map(MapToBeanFromMapWithSource.Source value)

So it seems like this has never worked out of the box. But maybe I was doing something wrong. Do you have a working example in the main branch?

[filiphr]

@thunderhook when I try your test on main it works when one of the following mappings are used

@Mapping(target = "targetName", source = "sourceA.name")
@Mapping(target = "targetName", source = "source.sourceA.name")

The other two

@Mapping(target = "targetName", source = "name")
@Mapping(target = "targetName", source = "source.name")

do not work because when using name the type that you get is Source and thus the compile error to provide a mapping from Source to String. When using source.name it is the same as using name since the source parameter is called source.

[thunderhook]

Thanks for testing. Looks like I made a typo or something.
Yes, you're right. It works on main, but not on my branch.

To be backwards compatible, we would need some sort of control, whether to use a dot as a parameter segment or as a key containing a dot. But that leads to other complexities.
Think of nested maps, where the top level needs to be treated as a parameter segment, and the nested map as a string. An intermediate method would be needed, and nesting like this would be error-prone: sourceA.innerMap.some.key.with.dots.person.firstName, which should be separated like this:
sourceA - innerMap - some.key.with.dots - person - firstName.

But, there may be another way: Using the dot is a common and regular pattern for parameter segments. It is more of an exception to use it as a string like "a.b". So the user may need to escape the dot, if it is meant to be a string, such as: "a\.b".
This is the same with regular expressions where the dot is also meant as a wildcard, and therefore needs to be escaped.

WDYT? Is there already some kind of escaping? I only know of the MappingConstants with their <...> naming syntax. But I don't think a qualifier like this such as <DOT> would be helpful.

[thunderhook]

Here's my take with the escaped character. It is now possible to uses nested mappings with dotted keys as I stated in my previous comment. Please have a look.

Here is an example of a nested mapper: https://github.com/mapstruct/mapstruct/pull/3469/files#diff-c6a0b5742b174a67682dcf0a6384868c29be04169fbabe8c1c7eed7b330084d8R22

[filiphr]

I think having a special escape character like there is in regular expressions is completely fine. I'd try to find some time to review this. I would like to first focus on some of the remaining things for 1.6 before start to add some new functionality as well.