proton-driver 0.3.0

Python driver with native interface for Proton

Introduction

Timeplus is a unified streaming and historical data processing engine in a single binary.

This project provides python driver to interact with Timeplus Proton or Timeplus Enterprise, the code is based on https://github.com/mymarilyn/clickhouse-driver.

Installation

Timeplus Python Driver currently supports the following versions of Python: 3.8, 3.9, 3.10, 3.11, 3.12, 3.13, and 3.14.

The driver fully supports the free-threaded 3.14t interpreter: all compiled extensions declare free-threading compatibility, so the GIL stays disabled when the driver is imported. cp314t wheels are published alongside the regular cp314 wheels under the same package name — pip on a free-threaded interpreter picks the right wheel automatically.

One caveat: the optional compression extras (proton-driver[lz4] / [zstd]) are not fully free-threaded yet. lz4 4.4.5+ already declares free-threading support, but both extras also require clickhouse-cityhash, which has no free-threaded (cp314t) wheels and whose sdist does not compile on 3.14t — a free-threaded fork is maintained at timeplus-io/proton-cityhash and will replace it in a follow-up release. zstd ships cp314t wheels but does not declare free-threading support, so importing it re-enables the GIL process-wide (everything still works, just without free-threaded parallelism). Uncompressed connections — the default — are unaffected.

Installing with pip We recommend creating a virtual environment when installing Python dependencies. For more information on setting up a virtual environment, see the Python documentation.

pip install proton-driver --extra-index-url https://d.timeplus.com/simple/

Quick Start

1. Run Timeplus Proton with docker. Make sure the port 8463 is exposed.

docker run -d -p 8463:8463 --pull always --name proton d.timeplus.com/timeplus-io/proton:latest

2. Run following python code

from proton_driver import connect
with connect("proton://default:@localhost:8463/default") as conn:
  with conn.cursor() as cursor:
    cursor.execute("select 1")
    print(cursor.fetchone())

above code should return (1,) , which shows that everything is working fine now.

Streaming Query

from proton_driver import client

c = client.Client(host='127.0.0.1', port=8463)

# create a random stream if not exist
c.execute("CREATE RANDOM STREAM IF NOT EXISTS"
          " devices("
          " device string default 'device'||to_string(rand()%4), "
          " temperature float default rand()%1000/10"
          ")")
# query the stream and return in a iterator
rows = c.execute_iter(
    "SELECT device, count(*), min(temperature), max(temperature) "
    "FROM devices GROUP BY device",
)
for row in rows:
    print(row)

the output of the code will be something like following, as for streaming query is unbounded, you can add your flow control to terminate the loop.

('device0', 747, 0.0, 99.5999984741211)
('device1', 723, 0.10000000149011612, 99.30000305175781)
('device3', 768, 0.30000001192092896, 99.9000015258789)
('device2', 762, 0.20000000298023224, 99.80000305175781)
('device0', 1258, 0.0, 99.5999984741211)
('device1', 1216, 0.10000000149011612, 99.69999694824219)
('device3', 1276, 0.30000001192092896, 99.9000015258789)
('device2', 1250, 0.20000000298023224, 99.80000305175781)

Insert Data

from proton_driver import client

c = client.Client(host='127.0.0.1', port=8463)

# create a random stream if not exist
c.execute("INSERT INTO proton_stream (raw) VALUES",rows) #rows is an array of arrays

Pandas DataFrame

Big fan of Pandas? We too! You can mix SQL and Pandas API together. Also you can converting query results to a variety of formats(e.g. Numpy Array, Pandas DataFrame, Polars DataFrame, Arrow Table) by DBAPI.

import pandas as pd
import time

from proton_driver import client

if __name__ == "__main__":
    c = client.Client(host='127.0.0.1', port=8463)

    # setup the test stream
    c.execute("drop stream if exists test")
    c.execute("""create stream test (
                    year int16,
                    first_name string
                )""")
    # add some data
    df = pd.DataFrame.from_records([
        {'year': 1994, 'first_name': 'Vova'},
        {'year': 1995, 'first_name': 'Anja'},
        {'year': 1996, 'first_name': 'Vasja'},
        {'year': 1997, 'first_name': 'Petja'},
    ])
    c.insert_dataframe(
        'INSERT INTO "test" (year, first_name) VALUES',
        df,
        settings=dict(use_numpy=True),
    )
    # or c.execute("INSERT INTO test(year, first_name) VALUES", df.to_dict('records'))
    time.sleep(3) # wait for 3 sec to make sure data available in historical store

    df = c.query_dataframe('SELECT * FROM table(test)')
    print(df)
    print(df.describe())

    # Converting query results to a variety of formats with dbapi
    with connect('proton://localhost') as conn:
        with conn.cursor() as cur:
            cur.execute('SELECT * FROM table(test)')
            print(cur.df()) # Pandas DataFrame

            cur.execute('SELECT * FROM table(test)')
            print(cur.fetchnumpy()) # Numpy Arrays

            cur.execute('SELECT * FROM table(test)')
            print(cur.pl()) # Polars DataFrame

            cur.execute('SELECT * FROM table(test)')
            print(cur.arrow()) # Arrow Table