• Without SQLAlchemy, we’d only use a DBAPI to establish connections and execute SQL statements. Simple, but not scalable as complexity grows.
• SQLAlchemy offers several layers of abstraction and convenient tools for interacting with a database.

SQLAlchemy vs psycopg2

• SQLAlchemy generates SQL statements
• psycopg2 directly sends SQL statements to the database
• SQLAlchemy depends on psycopg2 or other database drivers to communicate with the database under the hood.

SQLAlchemy lets you traverse through all 3 layers of abstraction to interact with your database

• Can stay on the ORM level
• Can dive into database operations to run customized SQL code specific to the database, on the Expressions level.
• Can write raw SQL to execute, when needed, on the Engine level.
  • Can more simply use psycopg2 in this case

Good Design

• Keep your code Pythonic. Work in classes and objects as much as possible.
  • Makes switching to a different backend easy in the future
• Avoid writing raw SQL until absolutely necessary.

Layers of SQLAlchemy

1. DBAPI: Implements protocol for connecting to a database
2. The Dialect: Translates generic SQL instructions to database-specific instructions for the DBAPI + specific backend (MySQL, Postgres, etc.)
3. The Connection Pool: Manages a pool of connections that we reuse
4. The Engine: Connect to the db
5. SQL Expressions: Write SQL using Python expressions
6. SQLAlchemy ORM (optional): Write and query Python objects mapped to db tables, generating SQL expressions for you.

SQLAlchemy layers 的图