• SSC Lunch Time Python
• 2
• 3
• 4
• 5
• 6
• 7
• 8
• 9
• 10
• 11
• 12
• 13

Lunchtime Python #3: Click

Presenter: Dominic Kempf, Scientific Software Center

Click is a Python package for creating beautiful command line interfaces in a composable way with as little code as necessary.

Command Line Interfaces: Why?

Natural evolution of a piece of research software in Python:

• Starts on a Jupyter notebook playground
• At some point freezes into a script for long term use
• (Automated) Application to a wider range of usage scenarios

For the last step, good Command Line Interface (CLI) is very helpful. Important aspects:

• Easy addition into existing code
• Few lines of code to ease maintenance and focus on the scientific part
• Good help text generation

How not to do it

We get access to command line arguments similar to C through sys.argv:

# DON'T DO THIS!
import sys

inputfile = sys.argv[1]
print(f"Calculating statistics from {inputfile}")

Why is this bad?

• No help text generation
• No validation of arguments
• Prone to errors in argument indexing
• Logic for non-string and optional arguments becomes quickly unwieldy

The standard libraries argparse and optparse are better options, but there is even better.

Click: A beautiful, opinionated approach

import click

@click.command()
@click.argument("inputfile", type=click.Path(exists=True))
def stats(inputfile):
    """Read data from the given INPUTFILE and calculate useful statistics"""
    click.echo(f"Calculating statistics from {inputfile}")

# This allows use of this Python file both for imports and for the CLI
if __name__ == "__main__":
    stats()

Arguments vs. Options

Arguments are positional and only too a small extent optional or defaultable. Use them only for absolute essential, self-explanatory input. To customize your script's behavious options are the better choice:

import click

@click.command()
@click.option(
    "--input",
    type=click.Path(exists=True),
    default="input.txt",
    help="The data file to read from",
)
@click.option(
    "--verbose/--no-verbose", type=bool, help="Whether to output intermediate results"
)
def stats(verbose, input):
    """Read data and calculate useful statistics"""
    if verbose:
        click.echo("Started the CLI script")
    click.echo(f"Calculating statistics from {input}")

if __name__ == "__main__":
    stats()

Composability of commands

Add subcommand structure by reusing previously defined commands:

@click.group()
def main():
    pass

@click.command()
def preprocess():
    click.echo("Apply preprocessing")

main.add_command(stats)
main.add_command(preprocess)

if __name__ == "__main__":
    main()

This mechanism is very powerful: arbitrary nesting, runtime extension e.g. through plugins etc.

Setuptools integration

As software becomes more mature, it is also advisable to package and distribute it as a Python package. Click easily integrates with setuptools as well using the entrypoints mechanism:

# In setup.py
setup(entry_points={"console_scripts": ["myscript = mypackage.mymodule:myclifunction"]})

# In setup.cfg
[options.entry_points]
console_scripts =
    myscript = mypackage.mymodule:myclifunction

Further information about click

Today's presentation can be found on the Lunch Time Python website: https://ssciwr.github.io/lunch-time-python/

For further information, see also the (very good) Click documentation

For questions to the Scientific Software Center, please write us to ssc@iwr.uni-heidelberg.de

Lunch Time Python Session #4

Library options, please vote now:

• itertools is a standard library that implements a number of iterator building blocks inspired by functional programming languages.
• pytest The pytest framework makes it easy to write small tests, yet scales to support complex functional testing for applications and libraries.
• matplotlib Matplotlib is a comprehensive library for creating static, animated, and interactive visualizations in Python