[int_19h]

The problem is that once you start fetching specific columns in your queries, the resulting objects aren't entities, just records - i.e. it's not really object-oriented, which is the main allure of ORM.

ORMs that are more honest about this, such as SQLAlchemy, are generally better than those that try to pretend that you really are dealing with entities.

[collinmanderson]

I use .values() and .values_list() to fetch raw records instead of objects. It still allows me to construct the query using the ORM which is nice.

[int_19h]

I think that "ORM" is a bit of a misnomer for that. I'd call it "language-integrated query", if not for the potential confusion with LINQ (which stands for exactly that, not coincidentally).

[takeda]

There's a clear distinction between query builder and ORM though you can especially see this with SQLAlchemy.

But anyway even query builder is not that great. Recently used PyCharm's integration with DataGrip. Basically the way it works is that if you configured a database in your project and let PyCharm fetch its schema, suddenly the IDE started recognizing the SQL statements providing autocomplete not only the statements but also table names etc, it also offered refactoring which created migration scripts.

After using that I think that's the proper way of solving the impedance mismatch and at that point you no longer need ORM or query builders. I hope other IDEs will start doing the same thing.

[collinmanderson]

> I think that "ORM" is a bit of a misnomer for that.

Good point: I'm no longer doing object mapping in that case. :) Though I think the most complicated part of the process that the Django "ORM" does for you is build the query. Maybe it should be called Django query generator.