Crafting Code Podcast

~/podcast

$ cd episodes/021-crafting-katas

~/podcast/episodes/021-crafting-katas $ ls -1a
. .. episode-summary.txt get-mp3.sh
~/podcast/episodes $ cat episode-summary.txt

In this episode, your hosts (finally!) share some actual code examples to illustrate some of our thoughts about crafting code. Follow along with the code at https://github.com/crafting-code-podcast/crafting-katas as we discuss how we implemented Conway's Game of Life in multiple languages. Testing strategies, optimizations, and feedback loops all demonstrate that crafting code is ultimately an iterative process.

Timestamps

[00:00:00] Host introductions. Overview of Conway's Game of Life

[00:03:20] Link to the repo: https://github.com/crafting-code-podcast/crafting-katas Rules in the readme. Separate branches for each implementation we tried.

[00:04:03] The allan-csharp branch. See the commit history for small steps. GoL patterns (e.g. glider or boat). Implementation details altered the tests during a reuse attempt. TDD gives you a context that constrains you.

[00:10:15] Matt liked the patterns & probing of the domain for tests. Discussed performance optimizations. Very deep problem because the game of life is turing complete.

[00:14:09] Nobody remembers how to do a 2D array off-hand. More about optimizations. Tests you start with drive your implementation.

[00:17:34] It is good to add code which makes your life easier during testing / development. Public methods that are not on an interface. Similar strategies in other languages, e.g. Go test packages. Total value being delivered includes the tests.

[00:20:36] Interface vs. implementation concept. Duck typing (a.k.a structural typing).

[00:23:22] The matt-python branch. A 3x3 grid w/ wrapping via REPL-driven development. False start on testing the generate function; fell back to testing neighbors. It is okay to have multiple asserts per test.

[00:27:17] It's common to think the code will go a certain way, but you end up doing it differently as you break it down. Immerse yourself in the problem and let your subconscious help you work it out. Being wrong will get you closer to being right. Feedback. It's not about not making mistakes. Use feedback loops and barriers to guard against mistakes.

[00:33:22] Iterate because you can't guess right. Don't lead with picking a certain pattern. Emergent properties (evolving architecture).

[00:35:58] Overloaded the term "state": test state vs. cell state. Optimized for conceptual simplicity (readability), not performance. Optimization is a trade-off; what happens if the rules change if you optimized? Default grid function was used for convenience in the REPL, then became helpful to use. Classic blunder: mutable data.

[00:42:13] The dave-csharp branch. Table-driven (NUnit TestCase) tests of the rules for quick win. Emergence of "population" driven by testing. Dave avoided the problem Matt fell into because he'd been burned by that same problem before.

[00:50:18] Classic vs. mockist development. Verification of a behavior. A bug in the test. It is easy to accidentally test mocks, but on the other hand, mocking allows you to use different rules w/ the same world engine.

[00:54:41] Acceptance testing. Creating "more code" isn't always bad. Tests should be express clearly, not cleverly. Different types of tests have different purposes.

[01:03:20] The allan-javascript branch. Borrowed ideas from Dave and Matt. If we had been pairing, we could have made it even better. The virtue of a lazy developer.

[01:08:17] The dave-go branch. Dave only did the rules -- that was the only thing the spec asked for! Table-testing and learning enumerations in GoLang.

[01:11:11] If you get the opportunity to go to a code retreat, do it. Crafting code is a process, not a destination.

[01:12:30] Outro

References in this episode: ~/podcast $ cat copyright.txt

Copyright © 2024 - Crafting Code Podcast