Performance Issues with Lumen/Dingo/Transformers

Hi there,

I'm currently working on an API using Lumen 5.4.*, Dingo 1.0.8 and Fractal Transformers 0.17.0. What I am facing currently are significant performance issues when requesting data that is based purely based on a model (so no expensive calculation or filtering is done on the result set). However it takes almost 6 seconds on average to return the result set consisting of 20 items.

The result is transformed by a transformer that also adds in a transformed representation of the relations of the requested model.

What I found out so far is that when I directly return the result set and omit the transformation, the response takes roughly a second (so improvement x5). Also when I do not return the relations in the transformed result the response is returned way faster (~ x2).

Is there anything I can do about this? How can I get more insights on why the transformers take so long and track down the bottle neck?

Thanks!