Lesson 10: Packages and modules

The Python Standard Library has lots of built-in modules that contain useful functions and data types for doing specific tasks. You can also use modules from outside the standard library. And you will undoubtedly write your own modules!

A module is contained in a file that ends with .py. This file can have classes, functions, and other objects. We will not discuss defining your own classes until much later in the bootcamp, so your modules will essentially just contain functions for now.

A package contains several related modules that are all grouped together under one name. We will extensively use the NumPy, SciPy, Pandas, and Bokeh packages, among others, in the bootcamp, and I’m sure you will also use them beyond. As such, the first module we will consider is NumPy. We will talk a lot more about NumPy later in the bootcamp.

Example: I want to compute the mean and median of a list of numbers

Say I have a list of numbers and I want to compute the mean. This happens all the time; you repeat a measurement multiple times and you want to compute the mean. We could write a function to do this.

[1]:

def mean(values):
    """Compute the mean of a sequence of numbers."""
    return sum(values) / len(values)

And it works as expected.

[2]:

print(mean([1, 2, 3, 4, 5]))
print(mean((4.5, 1.2, -1.6, 9.0)))

3.0
3.275

In addition to the mean, we might also want to compute the median, the standard deviation, etc. These seem like really common tasks. Remember my advice: if you want to do something that seems really common, a good programmer (or a team of them) probably already wrote something to do that. Means, medians, standard deviations, and lots and lots and lots of other numerical things are included in the Numpy module. To get access to it, we have to import it.

[3]:

import numpy

That’s it! We now have the numpy module available for use. Remember, in Python everything is an object, so if we want to access the methods and attributes available in the numpy module, we use dot syntax. In a Jupyter notebook or in the JupyterLab console, you can type

numpy.

(note the dot) and hit tab, and we will see what is available. For Numpy, there is a huge number of options!

So, let’s try to use Numpy’s numpy.mean() function to compute a mean.

[4]:

print(numpy.mean([1, 2, 3, 4, 5]))
print(numpy.mean((4.5, 1.2, -1.6, 9.0)))

3.0
3.275

Great! We get the same values! Now, we can use the numpy.median() function to compute the median.

[5]:

print(numpy.median([1, 2, 3, 4, 5]))
print(numpy.median((4.5, 1.2, -1.6, 9.0)))

3.0
2.85

This is nice. It gives the median, including when we have an even number of elements in the sequence of numbers, in which case it automatically interpolates. It is really important to know that it does this interpolation, since if you are not expecting it, it can give unexpected results. So, here is an important piece of advice:

Always check the doc strings of functions.

We can access the doc string of the numpy.median() function in JupyterLab by typing

numpy.median?

and looking at the output. An important part of that output:

Notes
-----
Given a vector ``V`` of length ``N``, the median of ``V`` is the
middle value of a sorted copy of ``V``, ``V_sorted`` - i
e., ``V_sorted[(N-1)/2]``, when ``N`` is odd, and the average of the
two middle values of ``V_sorted`` when ``N`` is even.

This is where the documentation tells you that the median will be reported as the average of two middle values when the number of elements is even. Note that you could also read the documentation here, which is a bit easier to read.

The as keyword

We use Numpy all the time. Typing numpy over and over again can get annoying. So, it is common practice to use the as keyword to import a module with an alias. Numpy’s alias is traditionally np, and this is the only alias you should ever use for Numpy.

[6]:

import numpy as np

np.median((4.5, 1.2, -1.6, 9.0))

[6]:

2.85

I prefer to do things this way, though some purists differ. We will use traditional aliases for major packages like Numpy and Pandas throughout the bootcamp.

Third party packages

Standard Python installations come with the standard library. Numpy and other useful packages are not in the standard library. Outside of the standard library, there are several packages available. Several. Ha! There are currently (June 8, 2020) about 239,000 packages available through the Python Package Index, PyPI. Usually, you can ask Google about what you are trying to do, and there is often a third party module to help you do it. The most useful (for scientific computing) and thoroughly tested packages and modules are available using conda. Others can be installed using pip, which we will do later in the bootcamp.

Writing your own module

We will skip the rest of this lesson, but it is worth going over as your software projects get more mature and you want to build your own packages.

To write your own module, you need to create a .py file and save it. You can do this using the text editor in JupyterLab. Let’s call our module na_utils, for “nucleic acid utilities.” So, we create a file called na_utils.py. We’ll build this module to have two functions, based on things we’ve already written. We’ll have a function dna_to_rna(), which converts a DNA sequence to an RNA sequence (just changes T to U), and another function reverse_rna_complement(), which returns the reverse RNA complement of a DNA template. The contents of na_utils.py should look as follows.

"""
Utilities for parsing nucleic acid sequences.
"""

def dna_to_rna(seq):
    """
    Convert a DNA sequence to RNA.
    """
    # Determine if original sequence was uppercase
    seq_upper = seq.isupper()

    # Convert to lowercase
    seq = seq.lower()

    # Swap out 't' for 'u'
    seq = seq.replace('t', 'u')

    # Return upper or lower case RNA sequence
    if seq_upper:
        return seq.upper()
    else:
        return seq


def reverse_rna_complement(seq):
    """
    Convert a DNA sequence into its reverse complement as RNA.
    """
    # Determine if original was uppercase
    seq_upper = seq.isupper()

    # Reverse sequence
    seq = seq[::-1]

    # Convert to upper
    seq = seq.upper()

    # Compute complement
    seq = seq.replace('A', 'u')
    seq = seq.replace('T', 'a')
    seq = seq.replace('G', 'c')
    seq = seq.replace('C', 'g')

    # Return result
    if seq_upper:
        return seq.upper()
    else:
        return seq

Note that the file starts with a doc string saying what the module contains.

I then have my two functions, each with doc strings. We will now import the module and then use these functions. In order for the import to work, the file na_utils.py must be in your present working directory, since this is where the Python interpreter will look for your module. In general, if you execute the code

import my_module

the Python interpreter will look first in the pwd to find my_module.py.

[7]:

import na_utils

# Sequence
seq = 'GACGATCTAGGCGACCGACTGGCATCG'

# Convert to RNA
na_utils.dna_to_rna(seq)

[7]:

'GACGAUCUAGGCGACCGACUGGCAUCG'

We can also compute the reverse RNA complement.

[8]:

na_utils.reverse_rna_complement(seq)

[8]:

'CGAUGCCAGUCGGUCGCCUAGAUCGUC'

Wonderful! You now have your own functioning module!

A quick note on error checking

These functions have minimal error checking of the input. For example, the dna_to_rna() function will take gibberish in and give jibberish out.

[9]:

na_utils.dna_to_rna('You can observe a lot by just watching.')

[9]:

'you can observe a lou by jusu wauching.'

In general, checking input and handling errors is an essential part of writing functions, and we will cover that in a later lesson.

Importing modules in your .py files and notebooks

As our first foray into the glory of PEP 8, the Python style guide, we quote:

Imports are always put at the top of the file, just after any module comments and docstrings, and before module globals and constants.

Imports should be grouped in the following order:

  1. standard library imports

  2. related third party imports

  3. local application/library specific imports

You should put a blank line between each group of imports.

You should follow this guide. I generally do it for Jupyter notebooks as well, with my first code cell having all of the imports I need. Therefore, going forward all of our lessons will have all necessary imports at the top of the document. The only exception is when we are explicitly demonstrating a concept that requires an import.

Imports and updates

Once you have imported a module or package, the interpreter stores its contents in memory. You cannot update the contents of the package and expect the interpreter to know about the changes. You will need to restart the kernel and then import the package again in a fresh instance.

This can seem annoying, but it is good design. It ensures that code you are running does not change as you go through executing a notebook. However, when developing modules, it is sometimes convenient to have an imported module be updated as you run through the notebook as you are editing. To enable this, you can use the autoreload extension. To activate it, run the following in a code cell.

%load_ext autoreload
%autoreload 2

Whenever you run a cell, imported packages and modules will be automatically reloaded.

Package management

When we wrote the na_utils module, we stored it in the directory that we were working in, or the pwd. But what if you write a module that you want to use regardless of what directory your are in? To allow this kind of usage, you can use the setuptools module of the standard library to manage your packages. You should read the documentation on Python packages and modules to understand the details of how this is done, but what we present here is sufficient to get simple packages running and installed.

Package architecture

In order for the tools in setuptools to effectively install your modules for widespread use, you need to follow a specific architecture for your package. I made an example jb_bootcamp package that is present in the ~/git/bootcamp/modules folder.

The file structure is of the package is

/jb_bootcamp
  /jb_bootcamp
    __init__.py
    na_utils.py
    bioinfo_dicts.py
    ...
setup.py
README.md

The ellipsis above signifies that there are other files in there that we are not going to use yet. I am trying to keep it simple for now to show how package management works.

To set up the package, you can use your command line skills to make the directories and use the JupyterLab text editor to make the files, in this case, four of them, __init__.py, na_utils.py, bioinfo_dicts.py (the little module we mentioned in the lesson on dictionaries), setup.py, and README.md.

It is essential that the name of the root directory be the name of the package, and that there be a subdirectory with the same name. That subdirectory must contain a file __init__.py. This file contains information about the package and how the modules of the package are imported, but it may be empty for simple modules. In this case, I included a string with the name and version of the package, as well as instructions to import appropriate modules. Here are the contents of __init__.py. The first two lines of code tell the interpreter what to import when running import jb_bootcamp.

"""Top-level package for utilities for bootcamp."""

from .na_utils import *
from .bioinfo_dicts import *

__author__ = 'Justin Bois'
__email__ = 'bois@caltech.edu'
__version__ = '0.0.1'

Also within the subdirectory are the .py files containing the code of the package. In our case, we have, na_utils.py and bioinfo_dicts.py.

It is also good practice to have a README file (which I suggest you write in Markdown) that has information about the package and what it does. Since this little demo package is kind of trivial, the README is quite short. Here are the contents I made for README.md (shown in unrendered raw Markdown).

# jb_bootcamp

Utilities for use in the Introduction to Programming in the Biological Sciences Bootcamp.

Finally, in the main directory, we need to have a file called setup.py, which contains the instructions for setuptools to install the package. We use the setuptools.setup() function to do the installation.

import setuptools

with open("README.md", "r") as f:
    long_description = f.read()

setuptools.setup(
    name='jb_bootcamp',
    version='0.0.1',
    author='Justin Bois',
    author_email='bois@caltech.edu',
    description='Utilities for use in bootcamp.',
    long_description=long_description,
    long_description_content_type='ext/markdown',
    packages=setuptools.find_packages(),
    classifiers=(
        "Programming Language :: Python :: 3",
        "Operating System :: OS Independent",
    ),
)

This is a minimal setup.py function, but will be sufficient for most packages you write for your own use. For your use, you make obvious changes to the name, author, etc., fields.

Installing your package

You will install your package after a future lesson, when we put it under version control.

Computing environment

[10]:

%load_ext watermark
%watermark -v -p numpy,jupyterlab

CPython 3.7.7
IPython 7.13.0

numpy 1.18.1
jupyterlab 1.2.6