The stronger your refactoring skill, the more easily you can use architecture advice as guidelines instead of as rules to enforce. This makes it significantly more likely that you’ll invest wisely in architecture, rather than over- or under-engineer.
Just another little example of two people looking at a situation, one seeing a problem and the other seeing a solution.
What happens when I try to review code in small steps and in public? Let’s find out.
You refactor too much! We’re not refactoring enough! This is a problem that, with a light touch, resolves itself.
The most common definition of “refactoring” suffers from a common weakness that creates big problems for a very specific group of people. I’d like to help.
Refactoring means rework, which means that we didn’t “get the design right the first time”. This can indicate a problem. It can also indicate a natural process of convergence towards a suitable design. Calling it “a code smell” seems to overstate the matter.
Programmers routinely give up on TDD when they try to do it in their toughest, meanest, most-valuable legacy code. I understand their impulse, but I think they’re setting themselves up for failure and ultimately missing out.
Programmers routinely ask me for advice on which kinds of tests they ought to write: unit vs. functional, fast vs. slow, big vs. small. They keep saying “integration test” when they mean “integrated test”. We have made this confusing, so I’d like to take one step towards clarifying it.
Debug with automated tests: it’s systematic, it leaves a record of what we’ve learned, and it’s boring in the best possible ways.
An example of test-first programming, focusing on adding behavior incrementally and removing duplication.
Are you worried that all these little classes and interfaces are going to destroy your system’s performance? Maybe. More often, however, the bottlenecks are caused by bigger classes and fewer interfaces causing duplication in the design.