sqlalchemy-batch-inserts 0.0.2

Batch inserts for SQLAlchemy on PostgreSQL with psycopg2

Benchling uses SQLAlchemy and psycopg2 to talk to PostgreSQL. To save on round-trip latency, we batch our inserts using this code.

Typically, creating and flushing N models in SQLAlchemy does N roundtrips to the database if the model has an autoincrementing primary key. This module improves creating N models to only require 2 roundtrips, without requiring any other changes to your code.

Is this for me?

You may find use for this module if:

• You are using SQLAlchemy
• You are using Postgres
• You sometimes need to create several models at once and care about performance

Usage

from sqlalchemy import create_engine
from sqlalchemy.orm import sessionmaker
from sqlalchemy_batch_inserts import enable_batch_inserting

engine = create_engine("postgresql+psycopg2://postgres@localhost", executemany_mode="values")  # SQLAlchemy < 1.3.7 needs use_batch_mode=True instead
Session = sessionmaker(bind=engine)
session = Session()
enable_batch_inserting(session)

If you use Flask-SQLALchemy,

from flask_sqlalchemy import SignallingSession
from sqlalchemy_batch_inserts import enable_batch_inserting

# Make sure that you've specified executemany_mode or use_batch_mode when creating your engine! Otherwise
# this library will not have any effect.
enable_batch_inserting(SignallingSession)

Acknowledgements

This is all possible thanks to @dvarrazzo's psycopg2 execute_batch and @zzzeek's SQLAlchemy support for the same and helpful advice on the mailing list.