Align AOT and reflective repository behavior. #5038
Conversation
… ones. Also fix projection field name mapping error
Make sure to register evaluation context with extension lookup.
| builder.addStatement("stages.add(new $T($S, $L))", Document.class, "$sort", | ||
| context.localVariable("sortDocument")); | ||
|
|
||
| if (outputType == Document.class || MongoSimpleTypes.HOLDER.isSimpleType(outputType) |
There was a problem hiding this comment.
MongoRepositoryContributor uses a MongoMappingContext. Ideally, we use its SimpleTypeHolder for further introspection to use a consistent configuration source.
|
|
||
| private static Class<?> getOutputType(MongoQueryMethod queryMethod) { | ||
| Class<?> outputType = queryMethod.getReturnedObjectType(); | ||
| if (MongoSimpleTypes.HOLDER.isSimpleType(outputType)) { |
There was a problem hiding this comment.
Let's use SimpleTypeHolder from MongoRepositoryContributor.
| @@ -52,7 +53,7 @@ static Placeholder indexed(int position) { | |||
| * @param type | |||
| * @return | |||
| */ | |||
| public static Shape geoJson(int index, String type) { | |||
| static Shape geoJson(int index, String type) { | |||
There was a problem hiding this comment.
FWIW, I liked to keep public in package-protected types to indicate a method is intended to be used as sort-of a public API (entrypoint) for other callers outside of the class. In any case, just a thought.
There was a problem hiding this comment.
I'd rather have them aligned with the type scope in such cases.
| @@ -157,12 +179,34 @@ static class PlaceholderParameterAccessor implements MongoParameterAccessor { | |||
|
|
|||
| private final List<Object> placeholders; | |||
There was a problem hiding this comment.
FWIW, in JPA we've introduced a concept of ParameterBinding definining BindingIdentifier (named/indexed identifier within the query) and ParameterOrigin (Expression, MethodInvocationArgument, Synthetic).
ParameterBinding can be an In or Like binding to carry around information about post-processing of the binding. I have the impression that MongoDB could benefit from such an arrangement as well.
There was a problem hiding this comment.
currently we're missing the link between the parameter and part tree part, so we'd have to introduce that first. I agree that having it would make things easier to follow. Should we do it in a next step?
There was a problem hiding this comment.
I think so. For now, we can use the current approach.
Closes: #5027