Since dozer is ~1000 times slower then mapping by hand and still 10 times slower then doing it by reflection or spring BeanUtils we simply can not use it.
What about generating fast mapping code from the dozer mapping xml file?
Here is a link on the performance http://www.christianschenk.org/blog/java-bean-mapper-performance-tests/
We dicovered the same timings.
this was discussed previously and we are going towards that throw long and painful refactoring. But the light in the tunnel is still far away due to some design decisions taken early in the project.
Nice performance comparison! Could not comment on that in blog since it requires registration. Would be great to see more detailed analysis results.