PG Stream Copy

Fast COPY TO postgresql table directly from python by converting input data to bytes and stream to psycopg2 cursor using COPY <table> FROM STDIN BINARY

Benchmark:

The test with 1 mln rows of different column types on docker environment gave results:

• ~21.5s for pg_stream_copy
• ~54s for psycopg2.extras.execute_values

Usage:

from datetime import date
from psycopg2 import connect
from pg_stream_copy import Schema, WriterEncoder


conn = connect('postgresql://postgres@localhost')
cursor = conn.cursor()
table_name = 'public.example_table'

cursor.execute(f'''
    CREATE TABLE {table_name} (
        _smallint SMALLINT NULL,
        _integer INTEGER NULL,
        _bigint BIGINT NULL,
        _float DOUBLE PRECISION NULL,
        _numeric NUMERIC NULL,
        _character_varying CHARACTER VARYING NULL,
        _date DATE NULL
    );
''')
schema = Schema.load_from_table(cursor, table_name)

with WriterEncoder(cursor, table_name, schema) as writer_encoder:
    writer_encoder.append_tuple((2, 3, 4, 2.34, 'foo bar', date(2019, 2, 1)))
    writer_encoder.append_dict({
        '_smallint': 200,
        '_integer': 300,
        '_bigint': 400,
        '_float': 234,
        '_numeric': Decimal("-12.34")
        '_character_varying': 'bar baz',
        '_date': date(2019, 2, 3),
    })

conn.commit()
conn.close()

Supported PostgreSQL types:

• boolean
• smallint
• integer
• bigint
• double precision
• numeric
• character varying
• text
• date
• timestamp
• timestamp with time zone
• json
• jsonb

Development:

# prepare env
docker compose run py bash
python -m venv venv
pip install -e .[dev,e2e]
# run tests
pytest tests/

If you need to test different PostgreSQL and Python version, you can use env vars:

PYTHON_VERSION=3.10 PG_VERSION=10 docker compose build
PYTHON_VERSION=3.10 PG_VERSION=10 docker compose run py ...

or use CI script:

PG_VERSION=11 PYTHON_VERSION=3.10 ./bin/tests.sh

currently pg_stream_copy is supporting Python 3.9 to 3.13 and PostgreSQL v10 to v16