It drives me crazy that my most common bug in Ruby is a name error. Maybe I typo a variable, or maybe I rename a class name in every location except for one, or maybe a cosmic ray flips a bit in my source before I commit. These are all mistakes that would be caught in a more static language like C or Java.
As ruby-static-checker shows, there's no inherent reason we can't have a static checker to help catch these sorts of bugs. Because Ruby is so dynamic, it is impossible to write a checker with 100% accuracy (you'll always be able to write byzantine programs that do crazy method redefinitions at runtime). But given a reasonable codebase, you should be able to write a tool to find name errors with a reasonable amount of certainty.
Ruby-static-checker is a static bugfinding tool for Ruby. Ruby-static-checker's goal is to provide a simple set of sanity checks (starting with name error detection) for existing codebases without requiring any additional programmer work. This means you should be able to, without modifying your application, run
And have the static analysis Just Work.
Note that we don't actually completely achieve this goal. Ruby-static-checker does require your codebase to load all required code but not execute the program if loaded as a library (use something like "main if $0 == FILE" to get this property). You can't have everything.
At the moment, ruby-static-checker just searches for name errors in your code. It's possible I'll make it do other things in the future.
Ruby-static-checker can find bugs like in the following (contrived) example:
class A def self.foo baz = 1 # Whoops, typod 'baz'! puts bas end end
The output of this should be (yes, I know, this output could be way better):
gdb@fire-hazard:~$ ruby-static-checker /path/to/a.rb Possible name error while calling A.bas
The current implementation is very barebones. It does have some false positives (and given the dynamic nature of Ruby, there's basically always going to be some false rate in either direction) but I've already used it to find real bugs.
Ruby-static-checker requires the path you give it, thus loading its code into memory. Any load-time class and method generation and any requires are carried out. We then iterate over all newly created classes, analyzing the ASTs of their methods (using the most excellent ParseTree library).
This has the benefit that we don't need to statically determine what the requires graph will look like. Obviously people can always require more files at runtime, but that's relatively rare so we don't try to address that case.
Yes. Some of the cooler ones are druby, reek, and roodi.
Tested on Ruby 1.8. I added handling of AST nodes as I encountered them in actual code, so it's likely there are some more esoteric nodes that I missed.
Glad to hear it! Patches are indeed most welcome. Just open a pull request on Github.