Database Abstraction Layer Generator
Database server support: PostgreSQL
Installation
- Using pip:
pip install dblayer - From source:
python setup.py install
How it works
Abstraction layer
Generates source code of a module implementing row classes with slots and a single database abstraction layer class. These classes can be used as is or extended by inheritance to fit your specific needs. It is also possible to extend the row classes to provide properties or helper methods. The generated code provides Python IDEs as much information as possible to make code completion useful.
Lightweight usage
The generated abstraction layer works like an ORM (Object Relational Mapper), but without a direct mapping of row instances to the corresponding database rows. It makes this approach more lightweight than a full-blown ORM. Database sessions and transactions must be initiated explicitly, but there are context managers defined to simplify your code.
Features
Tables, indexes and constraints are defined in a very clean way by just writing Python classes. Everything is defined explicitly, no automatic discovery of classes. Queries on multiple tables (joins) can also be defined in a clean way. Table aliasing is automatic and natural, since all the participating tables must be given an unique name in order to build the query definition. Highly complex queries can be build up gradually by simple Python class inheritance, which helps reducing code redundancy in real world use cases.
Constraints are enforced on the database server, not by Python code. It is also possible to add record finalization and validation code in Python if needed. There is support for defining efficient full text search indexes in a very simple way, so searching rows based on their textual contents is easy to implement.
Remarks
Performance
Most of the SQL is generated at compile time to reduce the runtime overhead as much as possible. Despite this it is possible to add runtime conditions, even building up complex conditions at runtime, but it will not slow down the simple use cases. Literal values are escaped automatically. The abstraction layer always returns unicode objects. You can pass str objects as literal values, the default encoding is UTF-8 in this case.
Limitations
The only supported database server is PostgreSQL via the psycopg2 extension. The generated code works with gevent-psycopg2 as well.
No support for defining database views, tablespaces and other database servers yet.
No support for inspecting databases (generating the classes defining and existing database) yet.
The unit test coverage is still not 100%.