data-pypeline 0.1.0

Pypeline is a data pipeline builder library.

Pypeline

This is a package for creating iterative data processing pipelines. Note that this is NOT a general purpose stream processing library. It is only designed as being a low overhead and simple-to-setup stream processing library. So for large scale production applications, use something like kafka instead.

Warning

This library is still at an ALPHA stage. So things may not work as intended and the api is not final!

Trivial Example

from pypeline import build_action, Pypeline, ForkingPypelineExecutor, wrap
import asyncio

async def step1():
    results = []
    for i in range(1000):
        results.append(wrap(i))
    return results

async def step2(i):
    return i * 10

async def step3(i):
    return i + 1

async def run_pipeline():
    pypeline = Pypeline()
    # Adding actions to the pipeline
    pypeline.add_action(build_action("Step1", step1)) \ 
            .add_action(build_action("Step2", step2)) \
            .add_action(build_action("Step3", step3, serialize_dir="./example"))  # Serialize results so future runs will skip this step entirely
    results = await pypeline.run(executor=ForkingPypelineExecutor())  # Custom executor that avoids the GIL
    # Results are wrapped in a utility namedtuple, so let's flatten it.
    results = [r.args[0] for r in results]
    return results

results = asyncio.get_event_loop().run_until_complete(run_pipeline())
for result in results:
    print(result)