piperun 0.2.0

A flexible framework for building and executing command pipelines in Python

Piperun - Command Pipeline Framework for Python

piperun is a flexible framework for building and executing command pipelines in Python. It provides tools for running shell commands, creating task pipelines, and executing operations in parallel.

Overview

piperun simplifies the process of building complex command workflows by providing:

• Command construction and execution with detailed control
• Sequential pipeline execution
• Parallel processing capabilities
• Integration with Dask for distributed computation

piperun was designed with the following principles in mind:

1. Simplicity: Easy to use with intuitive interfaces
2. Flexibility: Works with any callable or command
3. Composability: Build complex workflows from simple pieces
4. Performance: Efficient execution with parallel processing capabilities
5. Control: Detailed control over execution flow

Installation

Install piperun using pip:

pip install piperun

Install from source in editable mode:

git clone https://github.com/franioli/piperun.git
cd piperun
pip install -e .

Core Components

Command

The Command class provides an intuitive interface for constructing and executing shell commands:

from piperun import Command

# Create a simple command
cmd = Command("ls -l")
cmd.run()

# Add arguments dynamically
cmd = Command("parallel_stereo")
cmd.extend("image1.tif", "image2.tif", t="rpc", max_level=2)
cmd.run()

Key features:

• Handles both positional and keyword arguments
• Converts Python arguments to command-line format
• Provides execution timing
• Captures command output
• Supports boolean flags and various parameter formats

Pipeline

The Pipeline class enables chaining multiple processing steps:

from piperun import Pipeline, Command

# Create a pipeline
pipeline = Pipeline()

# Add steps
pipeline.add_step(Command("mkdir -p output"))
pipeline.add_step(Command("convert input.jpg output/output.png"))

# Execute all steps
pipeline.run()

Features:

• Sequential execution of steps
• Control for running specific steps or ranges
• Support for any step with a run() method
• Nested pipeline capability

DelayedTask

The DelayedTask class integrates with Dask for delayed execution:

from piperun import DelayedTask

# Create a delayed task
def process_data(x):
    return x * 2

task = DelayedTask(process_data, 10)
result = task.compute()  # Executes when needed

Features:

• Lazy evaluation of tasks
• Execution timing measurement
• Visualization of task graphs

ParallelBlock

The ParallelBlock class enables concurrent execution of multiple steps:

from piperun import ParallelBlock, Command

# Create commands
commands = [
    Command(f"process_file {i}.txt") 
    for i in range(10)
]

# Run in parallel
with ParallelBlock(commands, workers=4) as block:
    block.run()

Features:

• Automatic Dask cluster management
• Configurable worker count
• Support for both Command and DelayedTask objects

Advanced Usage

Pipeline Composition

Pipelines can be composed of various step types, including other pipelines:

# Create nested pipelines
preprocessing = Pipeline([
    Command("clean_data input.csv"),
    Command("validate_data input.csv")
])

processing = Pipeline([
    Command("process_data input.csv output.csv")
])

# Combine pipelines
main_pipeline = Pipeline()
main_pipeline.add_step(preprocessing)
main_pipeline.add_step(processing)
main_pipeline.run()

Parallel Execution with Custom Worker Count

# Run operations in parallel with custom worker count
parallel_tasks = ParallelBlock(workers=8)

for file in input_files:
    parallel_tasks.add_step(Command(f"process {file}"))

parallel_tasks.run()

Flow Control in Pipelines

# Run specific pipeline segments
pipeline = Pipeline([...])

# Run only step 3
pipeline.run_step(3)

# Run from step 2 to the end
pipeline.run_from_step(2)

# Run up to step 4 (not including step 4)
pipeline.run_until_step(4)

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

This project is licensed under the MIT License - see the LICENSE file for details.