Is there any performance gain if I only select several columns instead of a whole entity

[snoob]

Currently, I am fetching all the columns and then use MapStruct to map it into the DTO, it works fine for me.
It is bothersome to just select several fields in Hibernate/HQL.
For example:

• Use select forum.id, forum.name, forum.lastThread from Forum forum", I need to use setResultTransformer which is deprecated in Hibernate 5.2

• select new (...) from ... will make my query a big mess, and idk whether it allows nested new.

• Spring Data JPA Projection will still fetch all columns of the joined enity (https://stackoverflow.com/questions/40194614/spring-data-jpa-projection-selected-fields-from-the-db#comment89435552_40194614)

So I think I should just stick with the old way (select the whole enity and use MapStruct).

I want to know is there any big gap about the performance between between these two, especially in the case my queries contains several join statement

[vlad]

Yes, it is.

As explained in this article, selecting only a subset of columns is much more efficient and will always perform better than selecting all columns only to discard whatever fields you don’t need.