[Question] Mapper performance recommendations

[rsavuliak-macpaw]

Hey, just a question to clarify, cause didn't find anything meaningful in docs: Can you pls list recommendations regarding best performance of this package? Cause according to profiling it takes huge amount of time to deal with mapping, so we thought maybe there are some hints we can follow to decrease this time?

For example, from what I've noticed:

• I see that after first run it creates proxy files for the objects participating in mapping, extending them. However final classes obviously cannot be extended, that means that metadata for them will not be cached?
• "Warm up proxies on build time from the list of classes provided by the user." - this is not implemented yet, I guess.

Generally it would be great to get some page in the docs, with a recommendations on performance or maybe things that increase/decrease performance. Thx in advance!

[priyadi]

I see that after first run it creates proxy files for the objects participating in mapping, extending them. However final classes obviously cannot be extended, that means that metadata for them will not be cached?

Metadata is always cached, final or not final. Final only means the proxy will not be created.

"Warm up proxies on build time from the list of classes provided by the user." - this is not implemented yet, I guess.

Not yet implemented. But proxy and metadata generation only affects the first mapping of the pair.

Currently, the only available option to speed up things is to disable proxy (but then again proxy might speed things up depending on the use case). Further optimization needs to happen in the codebase.

If you have the profiling result, it might be very useful.