The Python ORM Problem
adam.gomaa.us
adam.gomaa.us
Also, the next Google won't be using SQL Alchemy either. The truth is when you get that big the law of leaky abstractions comes to bite you in the ass. You can only ignore the fundamental issues that these "frameworks" abstract away for so long. This allows you to build prototypes and medium sized applications easily but sooner or later you need to learn the details and most likely build a custom framework around it that solves your problem. As your application becomes larger chances are YOUR problem is significantly different from the AVERAGE problem and as such you can't use average tools.
The dilemma, as I see it, is choosing the appropriate abstractions, and choosing when to abstract yourself. For example, SQLA can be used as an SQL kit + a little personal ORM, or using their ORM, or their ORM with Elixir, or their ORM with your own thing, or some combination of the above. Or if you're the blogger, write your own declarative layer.
Not that its a particular problem in the Python world (indeed afaict Python is ahead of the game in this regard), but it is a significant issue for much Python development, worthy of exploration.
[seems to be a bug, I tried to add this as 'text' for the submission, but it didn't show up]