• Stars
    star
    2,067
  • Rank 22,358 (Top 0.5 %)
  • Language
    Python
  • License
    MIT License
  • Created over 6 years ago
  • Updated 11 months ago

Reviews

There are no reviews yet. Be the first to send feedback to the community and the maintainers!

Repository Details

A lightweight opinionated ETL framework, halfway between plain scripts and Apache Airflow

Mara Pipelines

Build & Test PyPI - License PyPI version Slack Status

This package contains a lightweight data transformation framework with a focus on transparency and complexity reduction. It has a number of baked-in assumptions/ principles:

  • Data integration pipelines as code: pipelines, tasks and commands are created using declarative Python code.

  • PostgreSQL as a data processing engine.

  • Extensive web ui. The web browser as the main tool for inspecting, running and debugging pipelines.

  • GNU make semantics. Nodes depend on the completion of upstream nodes. No data dependencies or data flows.

  • No in-app data processing: command line tools as the main tool for interacting with databases and data.

  • Single machine pipeline execution based on Python's multiprocessing. No need for distributed task queues. Easy debugging and output logging.

  • Cost based priority queues: nodes with higher cost (based on recorded run times) are run first.

Β 

Installation

To use the library directly, use pip:

pip install mara-pipelines

or

pip install git+https://github.com/mara/mara-pipelines.git

For an example of an integration into a flask application, have a look at the mara example project 1 and mara example project 2.

Due to the heavy use of forking, Mara Pipelines does not run natively on Windows. If you want to run it on Windows, then please use Docker or the Windows Subsystem for Linux.

Β 

Example

Here is a pipeline "demo" consisting of three nodes that depend on each other: the task ping_localhost, the pipeline sub_pipeline and the task sleep:

from mara_pipelines.commands.bash import RunBash
from mara_pipelines.pipelines import Pipeline, Task
from mara_pipelines.ui.cli import run_pipeline, run_interactively

pipeline = Pipeline(
    id='demo',
    description='A small pipeline that demonstrates the interplay between pipelines, tasks and commands')

pipeline.add(Task(id='ping_localhost', description='Pings localhost',
                  commands=[RunBash('ping -c 3 localhost')]))

sub_pipeline = Pipeline(id='sub_pipeline', description='Pings a number of hosts')

for host in ['google', 'amazon', 'facebook']:
    sub_pipeline.add(Task(id=f'ping_{host}', description=f'Pings {host}',
                          commands=[RunBash(f'ping -c 3 {host}.com')]))

sub_pipeline.add_dependency('ping_amazon', 'ping_facebook')
sub_pipeline.add(Task(id='ping_foo', description='Pings foo',
                      commands=[RunBash('ping foo')]), ['ping_amazon'])

pipeline.add(sub_pipeline, ['ping_localhost'])

pipeline.add(Task(id='sleep', description='Sleeps for 2 seconds',
                  commands=[RunBash('sleep 2')]), ['sub_pipeline'])

Tasks contain lists of commands, which do the actual work (in this case running bash commands that ping various hosts).

Β 

In order to run the pipeline, a PostgreSQL database is recommended to be configured for storing run-time information, run output and status of incremental processing:

import mara_db.auto_migration
import mara_db.config
import mara_db.dbs

mara_db.config.databases \
    = lambda: {'mara': mara_db.dbs.PostgreSQLDB(host='localhost', user='root', database='example_etl_mara')}

mara_db.auto_migration.auto_discover_models_and_migrate()

Given that PostgresSQL is running and the credentials work, the output looks like this (a database with a number of tables is created):

Created database "postgresql+psycopg2://root@localhost/example_etl_mara"

CREATE TABLE data_integration_file_dependency (
    node_path TEXT[] NOT NULL,
    dependency_type VARCHAR NOT NULL,
    hash VARCHAR,
    timestamp TIMESTAMP WITHOUT TIME ZONE,
    PRIMARY KEY (node_path, dependency_type)
);

.. more tables

CLI UI

This runs a pipeline with output to stdout:

from mara_pipelines.ui.cli import run_pipeline

run_pipeline(pipeline)

Example run cli 1

Β 

And this runs a single node of pipeline sub_pipeline together with all the nodes that it depends on:

run_pipeline(sub_pipeline, nodes=[sub_pipeline.nodes['ping_amazon']], with_upstreams=True)

Example run cli 2

Β 

And finally, there is some sort of menu based on pythondialog that allows to navigate and run pipelines like this:

from mara_pipelines.ui.cli import run_interactively

run_interactively()

Example run cli 3

Web UI

More importantly, this package provides an extensive web interface. It can be easily integrated into any Flask based app and the mara example project demonstrates how to do this using mara-app.

For each pipeline, there is a page that shows

  • a graph of all child nodes and the dependencies between them
  • a chart of the overal run time of the pipeline and it's most expensive nodes over the last 30 days (configurable)
  • a table of all the pipeline's nodes with their average run times and the resulting queuing priority
  • output and timeline for the last runs of the pipeline

Mara pipelines web ui 1

For each task, there is a page showing

  • the upstreams and downstreams of the task in the pipeline
  • the run times of the task in the last 30 days
  • all commands of the task
  • output of the last runs of the task

Mara pipelines web ui 2

Pipelines and tasks can be run from the web ui directly, which is probably one of the main features of this package:

Example run web ui

Β 

Getting started

Documentation is currently work in progress. Please use the mara example project 1 and mara example project 2 as a reference for getting started.

Links

More Repositories

1

mara-example-project-2

An example mini data warehouse for python project stats, template for new projects
Python
176
star
2

mara-schema

Mapping of DWH database tables to business entities, attributes & metrics in Python, with automatic creation of flattened tables
Python
70
star
3

mara-db

Lightweight configuration and access to multiple databases in a single project
Python
38
star
4

mara-etl-tools

Utilities for creating ETL pipelines with mara
PLpgSQL
36
star
5

mara-example-project-1

Runnable e-commerce mini data warehouse based on Python, PostgreSQL & Metabase, template for new projects
Python
27
star
6

mara-metabase

Configuration and schema sync for Metabase from Python
Python
19
star
7

mara-data-explorer

Flask based UI for displaying & segmenting a single database table
JavaScript
15
star
8

mara-app

A framework for distributing flask apps across separate packages with minimal dependencies
Python
15
star
9

mara-mondrian

A python integration for the Saiku ad hoc analysis tool
Python
9
star
10

mara-page

Minimal API for defining pages of Flask-based backends independently from the actual backend infrastructure
Python
4
star
11

mara-olist-ecommerce-data

Stores the Olist E-commerce Dataset in PostgreSQL
Python
4
star
12

mara-storage

Configuration of different storages
Python
3
star
13

mara-google-sheet-downloader

Download Google Sheets to a DB table in a mara ETL
Python
3
star
14

mara-markdown-docs

Display markdown documentation in mara
Python
2
star
15

mara-acl

Default ACL implementation for Mara
Python
2
star
16

mara-superset

Configuring Apache Superset from Mara Schema
Python
2
star
17

mara-dbt

Python
1
star