Kent Beck has a great article about “One Bite At A Time: Partitioning Complexity”.

Here is a summary of the main points:

One problem at a time

• Make it work, make it right, make it fast.
• Work on structure XOR behavior.
• Work on interface XOR implementation.
• Work on test XOR code.

Code structure

• Composed methods provide an outline of what’s happening.
• Mutable core/immutable leaves - not sure how this applies to working with ClojureScript where data is immutable by default.
• Move logic to data so that operations and data are located close together.
• One pile for the messy stuff, rather than having it scattered everywhere.

Workflow

• Start with a constant for complex input data, then compute it in a second step.
• Predict test results and use surprise as a pointer that your understanding is lacking, and you need to re-think the problem.
• Red/red/revert. Go back to the drawing board if the first attempt at fixing it didn’t work.
• Red/red/red/green/revert/red/green/red/green retrace the steps taken to tackle a complex problem for maximum learning.
• Wait let ugly code get more ugly before trying to fix it.
• Stories to separate thoughts of outcome from those of implementation.

A good quote from the end:

Many of the strategies above sacrifice short-term efficiency in order to keep my feet moving. What really slows me down is not programming slowly, it is getting overwhelmed, losing my confidence, and not programming at all.