sudsoln package in Python

Objectives

• Get familiar with Git
• Get familiar with Python class and unittest
• Polish my Python skills by developing a Python package

Background

So why did I create this package? As Python gets more popular in the field of data science, I felt the need to polish my programming skills in Python. I chose to build a generalized sudoku-solving mechanism since:

1. it seemed challenging enough
2. I couldn’t find any program that solves larger-than-9-by-9 sudoku puzzles
3. there was the note titled Solving Sudoku using simulated annealing in Statistical Computation course (STA410) at the University, and I was inspired by the use of the algorithm to solve a sudoku puzzle in R and wanted to import the same algorithm to Python

I wanted to code in a Pythonic way. Whenever I wrote a code in Python, I found myself thinking how to accomplish a task in R first, and then looking for equivalent ways to solve similar tasks in Python. Looking for online references and spending time to read them were inefficient and time-consuming, and Python codes produced in such a manner were not really utilizing the ability of Python properly. Building this package was one way to learn and extensively use Python features such as list comprehensions, zip, set, dict, filter, and so on.

Also, I wanted to get familiar with how Python packages are developed in general. Back when I was developing funpark package, I got the glimpse of idea of what should be included in a package and what it is like to develop one: development stages, licensing, version control, tests, etc. However, it was the very first package that was not uploaded on CRAN but only on my GitHub repository, and I was not thorough about designing unittests for the later update. As of 2019-08-15, I didn’t create any S3 or R6 methods in the package either. By developing something more complex (not referring to the idea behind functions in a package but rather to a code complexity), I figured that I will learn deeper about the use of user-defined classes and unittests as well as the usefulness of version control system more seriously. I was especially interested in Python class because OOP was the topic that was very briefly covered in the introductory programming course (CSC108) at the University, and was never mentioned as I specialized further in statistics and used R more often than Python.

And to be honest, dynamic badges looked awesome on README.md; I wanted to have use cases on my end.

How it works

• Visit the repository to see the source code
• See vignette for the instructions on how to use solving features of the package
• See analysis to see the performance of solving features

Learning outcomes

• Selecting what needs to be tested in a package
• The convenience and importance of automatic testing
• The use of Travis CI and Codecov.io
• The use of version control system (in particular, git)

Next steps

• Add more solving features
  • make Sudoku.solve_by_pointing_pairs() work properly whenever by = 'row' or by = 'col' in the version >= 0.1.1
  • write Sudoku.solve_by_simulated_annealing()
• Make a progress to Development Status :: 4 - Beta
• Make some attributes private
• Try define subclasses instead of writing separate classes