babar#

Logo

Babar🐘❤️🐍

Turn any Python object into a Postgres extension

Why | How | Installation | Usage | Roadmap

Why#

Postgres has supported Python as a procedural language for years via PL/Python.

Embedding non-trivial Python code in Postgres can get cumbersome and usually involves copy-pasting from existing code bases and/or creating thin wrappers around existing functions.

babar automates this process by allowing you to seamlessly package existing Python components (functions, classes etc.) int a postgres extension.

How#

babar dynamically inspects the definition of Python objects and generates semantically equivalent Postgres definitions along with the necessary extension files (.control, Makefile, .sql)

Installation#

Usage#

pip install pybabar

Let’s create a Postgres extension called pystring which adds two simple Python functions.

from babar import Extension

def pyconcat(x: str, y: str) -> str:
    return x + y

def pyupper(x: str) -> str:
    return x.upper()

if __name__ == "__main__":
    Extension(
        "pystring",
        pyconcat, pyupper,
        comment="this is the pystring extension",
        default_version="0.1.0",
    )

Then, from the command line, you can run:

python pystring.py

That will create the appropriate extension files, which you can then install in the usual Postgres way:

make 
make install

Then you can CREATE the extension and use it

psql -d postgres <<SQL
CREATE EXTENSION pystring

SELECT pyconcat('hello ', 'world');
SELECT pyupper('hello');
SQL

Roadmap#

v0.1.0 (wip)#

  • simple functions (not classes, methods etc)

  • Functions should be 100% self-contained (outer scope is not inspected)

  • primitive types only: support only str/text as arg and return

  • Probably no cli interface / just python main.py

Future#

  • @pg_function decorator for individual functions. Much like click;

  • maybe subclass Extension should make a whole Python class an extension?

  • CLI like babar -m pack.mod.func1 pack2.mod2.func2

  • Handle upgrades

  • pgxn integration

It is lightly inspired by Python-Fire which turns any Python object into a cli interface

Contents#