Code Analysis
Code analysis in ruby-lint is done using a set of classes that extend
RubyLint::Analysis::Base. This base class provides various methods for easily
walking over an AST by using callback methods based on the node types. For
example, the callback method on_string
is used before a (string)
node is
processed. For more low level details see the API documentation of
RubyLint::Iterator and RubyLint::Analysis::Base (which extends the former).
Example
For this guide we’ll be creating an analysis class that checks for local variables written in camelCase. Whenever it finds these variables a warning will be added informing the developer that he/she should use snake_case instead.
At the most basic level this class looks like the following:
class CamelCaseVariables < RubyLint::Analysis::Base
end
By extending the base class your own class already comes with a method for walking the AST and calling callbacks as well as a few helper methods for adding errors and the likes.
To use this class you’ll have to create an instance of it and call RubyLint::Iterator#iterate and pass it an AST:
ast = RubyLint::Parser.new.parse('exampleNumber = 10')
iterator = CamelCaseVariables.new
iterator.iterate(ast)
When running the above code you’ll notice that nothing actually happens. This is because no callback methods have been added yet. There are two types of callback methods that can be added:
on_X
after_X
Here X is the name of the node type. Methods that start with on_
will be
executed before any child nodes (of the current node) are processed. Methods
that start with after_
will be executed after the node and its child nodes
have been processed. Each callback method takes a single argument: an instance
of RubyLint::AST::Node containing information about the current node.
In the above example we only need an on_
callback:
class CamelCaseVariables < RubyLint::Analysis::Base
def on_lvar(node)
if node.children[0].to_s =~ /[a-z][A-Z]/
warning('use snake_case for local variables', node)
end
end
end
If you now were to run the above it would technically work but still you won’t see anything. This is because you also need to specify a RubyLint::Report instance to use for storing data such as warnings and error messages. This can be done as following:
report = RubyLint::Report.new
iterator = CamelCaseVariables.new(:report => report)
The last step is to actually display the report using a presenter. For this exercise we’ll use RubyLint::Presenter::Text. Presenters are quite easy to use:
presenter = RubyLint::Presenter::Text.new
presenter.present(report)
The full code of this exercise looks like the following:
class CamelCaseVariables < RubyLint::Analysis::Base
def on_lvar(node)
if node.children[0].to_s =~ /[a-z][A-Z]/
warning('use snake_case for local variables', node)
end
end
end
ast = RubyLint::Parser.new.parse('exampleNumber = 10')
report = RubyLint::Report.new
iterator = CamelCaseVariables.new(:report => report)
presenter = RubyLint::Presenter::Text.new
iterator.iterate(ast)
puts presenter.present(report)
Conditional Analysis
In some cases you want to use a certain analysis class but only enable it if a
certain condition is met. In order to do so a analysis class should define a
class method called analyze?
that returns a boolean that indicates if the
class should be used or not. The basic signature of this method can be seen at
RubyLint::Analysis::Base.analyze?.
For example, if you only want to analyze RSpec files:
class RSpecExample < RubyLint::Analysis::Base
def self.analyze?(ast, vm)
return ast.file =~ /_spec\.rb$/
end
end
By default all analysis classes are enabled.
Registering Analysis Classes
In order for a analysis class to become available in RubyLint::Configuration objects, either via the CLI or via Ruby directly, you must register the analysis class. This can be done by calling the class method RubyLint::Analysis::Base.register:
class CamelCaseVariables < RubyLint::Analysis::Base
register 'camel_case_variables'
end
The value should be a snake_cased
string that matches the class name.