On Thu, Sep 24, 2009 at 4:22 PM, Jean-Louis Leroy <jl@yorel.be> wrote:
I still say a Boost.RDB should be modeled after Python's SQLAlchemy library, such a wonderful design after using it for a year, and it optimizes your SQL for the underlying DB (so some may use joins, some may use other things, etc... whatever is best for the DB).
After having a look at SQLAlchemy, I have the impression that what I am working on is just a tiny part of it, namely what is described here : http://www.sqlalchemy.org/docs/05/sqlexpression.html
The goal of this lib is not ORM or anything, it merely tries to make it possible to talk to the database with fine control over the sql, and in a typesafe manner at that (something Python doesn't care about). It's just "embedded SQL done in C++". If it makes decisions behing your back, then it's missing its goal.
Later higher-level tools - among them ORMs - can be built upon it. That is my initial goal in fact - build an ORM.
Ah, useful. :) For note, Python is strongly-typed, have to be correct about what you send to the DB with regards to types, if you send an int to a string column, or a string to an int, or whatever, it will throw an exception.