Testing, Debugging, Optimizing, and Documenting

The recipes in previous chapters focus on writing code to do what you want. This chapter focuses on verifying that your code really works, and on fixing it when it breaks. We start off simple and move to more advanced debugging techniques.

What happens when your program has a bug? The best-case scenario is that you discover the bug before it affects anyone, including other developers. Thats the goal of unit tests (Recipe 17.7). Ruby and the Ruby community promote a philosophy of writing automated tests as (or even before) you write the corresponding functionality. At every stage of development, you know that your program works, and if you make a change that breaks something, you know about it immediately. These tests can replace much boring manual testing and bug hunting.

Suppose a bug slips past your tests, and you only discover it in production. Hows it going to manifest itself? If you e lucky, youll see an exception: a notification from some piece of Ruby code that something is wrong.

Exceptions interrupt the normal flow of execution, and, if not handled, will crash the program. The good news is that they give you a place in the code to start debugging. Its worse if a bug doesn cause an exception, because youll only notice its byproducts: corrupt data or even security violations. We show code for handling exceptions (Recipes 17.3 and 17.4) and for creating your own (Recipe 17.2).

Successful debugging means reproducing the bug in an environment where you can poke at it. This may mean dropping from a running program into an irb session (Recipe 17.10), or it may be as simple as adding diagnostic messages that make the program show its work (Recipe 17.1).

Even a program that has no noticeable bugs may run too slowly or use too many resources. Ruby provides two tools for doing performance optimization: a profiler (Recipe 17.12) and a benchmarking suite (Recipe 17.13). Its easy to create your own analysis tools by writing a trace function that hooks into the Ruby interpreter as it runs. The call graph tracker presented at chapters end (Recipe 17.15) exploits this feature.