tanbih-pipeline

Pipeline provides an unified interface to set up data stream processing systems with Kafka, Pulsar, RabbitMQ, Redis and many more. The idea is to free developer from the dynamic change of technology in deployment, so that a docker image released for a certain task can be used with Kafka or Redis through changes of environment variables.

Features

a unified interface from Kakfa to Pulsar, from Redis to MongoDB
components connection controlled via command line, or environment variables
support file and in-memory for testing

Requirements

Python 3.7, 3.8

Installation

$ pip install tanbih-pipeline

You can install the required backend dependencies with:

$ pip install tanbih-pipeline[redis]
$ pip install tanbih-pipeline[kafka]
$ pip install tanbih-pipeline[pulsar]
$ pip install tanbih-pipeline[rabbitmq]
$ pip install tanbih-pipeline[elastic]
$ pip install tanbih-pipeline[mongodb]

If you want to support all backends, you can:

$ pip install tanbih-pipeline[full]

Producer

Producer is to be used when developing a data source in our pipeline. A source will produce output without input. A crawler can be seen as a producer.

>>> from typing import Generator
>>> from pydantic import BaseModel
>>> from pipeline import Producer as Worker, ProducerSettings as Settings
>>>
>>> class Output(BaseModel):
...     key: int
>>>
>>> class MyProducer(Worker):
...     def generate(self) -> Generator[Output, None, None]:
...         for i in range(10):
...             yield Output(key=i)
>>>
>>> settings = Settings(name='producer', version='0.0.0', description='')
>>> producer = MyProducer(settings, output_class=Output)
>>> producer.parse_args("--out-kind MEM --out-topic test".split())
>>> producer.start()
>>> [r.get('key') for r in producer.destination.results]
[0, 1, 2, 3, 4, 5, 6, 7, 8, 9]

Processor

Processor is to be used to process input. Modification will be in-place. A processor can produce one output for each input, or no output.

>>> from pydantic import BaseModel
>>> from pipeline import Processor as Worker, ProcessorSettings as Settings
>>>
>>> class Input(BaseModel):
...     temperature: float
>>>
>>> class Output(BaseModel):
...     is_hot: bool
>>>
>>> class MyProcessor(Worker):
...     def process(self, content, key):
...         is_hot = (content.temperature > 25)
...         return Output(is_hot=is_hot)
>>>
>>> settings = Settings(name='processor', version='0.1.0', description='')
>>> processor = MyProcessor(settings, input_class=Input, output_class=Output)
>>> args = "--in-kind MEM --in-topic test --out-kind MEM --out-topic test".split()
>>> processor.parse_args(args)
>>> processor.start()

Splitter

Splitter is to be used when writing to multiple outputs. It will take a function to generate output topic based on the processing message, and use it when writing output.

>>> from pipeline import Splitter as Worker, SplitterSettings as Settings
>>>
>>> class MySplitter(Worker):
...     def get_topic(self, msg):
...         return '{}-{}'.format(self.destination.topic, msg.get('id'))
>>>
>>> settings = Settings(name='splitter', version='0.1.0', description='')
>>> splitter = MySplitter(settings)
>>> args = "--in-kind MEM --in-topic test --out-kind MEM --out-topic test".split()
>>> splitter.parse_args(args)
>>> splitter.start()

Usage

Choosing backend technology:

kind	description	multi- reader	shared reader	data expire
LREDIS	Redis List	X	X	read
XREDIS	Redis Stream	X	X	limit
KAFKA	Kafka	X	X	read
PULSAR	Pulsar	X	X	ttl
RABBITMQ	RabbitMQ	X		read
ELASTIC	ElasticSearch
MONGODB	MongoDB
FILE*	json,csv
MEM*	memory

FILE accepts jsonl input on stdin and with filename, it also accepts csv file. Both format can be gzipped.
MEM read and write to memory, designed for unit tests.

# check command line arguments for certain input and output
worker.py --in-kind FILE --help
# or
IN_KIND=FILE worker.py
# or
export IN_KIND=FILE
worker.py --help

# process input from file and output to stdout (--in-content-only is
# needed for this version)
worker.py --in-kind FILE --in-filename data.jsonl --in-content-only \
          --out-kind FILE --out-filename -


# read from file and write to KAFKA
worker.py --in-kind FILE --in-filename data.jsonl --in-content-only \
          --out-kind KAFKA --out-namespace test --out-topic articles \
          --out-kafka kafka_url --out-config kafka_config_json

Arguments

common

debug monitoring

kind namespace topic

input:

FILE

Scripts

pipeline-copy is a script to copy data from a source to a destination. It can be used to inject data from a file to a database, or from a database to another database. It is implemented as a Pipeline worker.

Since JSON format does not support datetimes, in order for pipeline-copy to treat datetime field as datetime instead of string, you can provide a model definition via argument –model-definition. An example of such model definition is as following (the class name needs to be Model):

from datetime import datetime
from typing import Optional

from pydantic import BaseModel

class Model(BaseModel):
    hashtag: str
    username: str
    text: str
    tweet_id: str
    location: Optional[str]
    created_at: datetime
    retweet_count: int

Environment Variables

Application accepts following environment variables (Please note, you will need to add prefix IN_, –in- and OUT_, –out- to these variables to indicate the option for input and output). Please refer to backend documentation for available arguments/environment variables.

Customize Settings

class CustomSettings(Settings):
    new_argument: str = Field("", title="a new argument for custom settings")

class CustomProcessor(Processor):
    def __init__(self):
        settings = CustomSettings("worker", "v0.1.0", "custom processor")
        super().__init__(settings, input_class=BaseModel, output_class=BaseModel)

Errors

PipelineError will be raised when error occurs

Contribute

Use pre-commit to run black and flake8

# install pre-commit
pre-commit install

Credits

Yifan Zhang (yzhang at hbku.edu.qa)

Release files for tanbih-pipeline 0.12.18

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for tanbih-pipeline 0.12.18
File	Size	Uploaded
tanbih_pipeline-0.12.18.tar.gz	38.7 kB	Sep 28, 2024	Details

Built distribution (wheel)

Table of built distributions (wheels) for tanbih-pipeline 0.12.18
File	Interpreter	ABI	Platform	Reset
tanbih_pipeline-0.12.18-py3-none-any.whl 38.4 kB Sep 28, 2024	Python 3	none	any	Details

Total release size: 77.0 kB

Release files / tanbih_pipeline-0.12.18.tar.gz

Download URL	tanbih_pipeline-0.12.18.tar.gz
Size	38.7 kB
Tags	Source
SHA-256 checksum How to use checksums	`5ad064ba8fadd3328701ebdda698680aa9c521dabc290be83523dc69a5f84ccd`
BLAKE2b-256 checksum How to use checksums	`f71a43025e300217ff3ad48d066ba0ae6726ab9ccad1669d758128ade7fbefe9`
Upload date	Sep 28, 2024
Uploaded using Trusted Publishing? What is trusted publishing?	No
Uploaded via	`pdm/2.19.1 CPython/3.12.6 Darwin/23.6.0`

Release files / tanbih_pipeline-0.12.18-py3-none-any.whl

Download URL	tanbih_pipeline-0.12.18-py3-none-any.whl
Size	38.4 kB
Tags	Python 3
SHA-256 checksum How to use checksums	`ed53cefce982dcd79645e984d2eff2a6d9513376bf95fe20e006b810dfd225f6`
BLAKE2b-256 checksum How to use checksums	`8cebe1fdeb3768a98d51b7a16fbc22d925244ff7ed53e91d840e05a154a2bb19`
Upload date	Sep 28, 2024
Uploaded using Trusted Publishing? What is trusted publishing?	No
Uploaded via	`pdm/2.19.1 CPython/3.12.6 Darwin/23.6.0`