Keep bad code out of your projects. Your idea of bad code, no one elses.
Because good documentation should be executable.
I'm proud to say Outlaw has been featured on Ruby5 and The Ruby Show! I also apoogize for the lack of updates, but I've been sporadically trying to interest other developers in helping to flesh out the use cases for rules they'd like to use in real projects, both ones that work with the existing gem and ones to help prioritize development of the rules engine going forward.
Real world desires and use cases are currently my priority above the planned features below, and some additional plans for a git gem I hope to develop and use with Outlaw to facilitate installing git commit hooks for running outlaw on a project on each commit. Even if you have no plans to contribute to the project directly, please fork and write up some example rules you'd like to see supported, or that you'd like to see adopted into the default rules that ship with outlaw, so I can get a better feel for how it can best serve the community.
NOTE: Outlaw can evaluate any version ruby code, BUT it runs on only 1.9 -- set your system ruby to 1.9 to use
Part of MendicantUniversity.org S10 class, personal project.
The current version of outlaw takes a user provided configuration file (currently named simply '.outlawed' in the project directory or user's home directory) and parses a series of method calls to the outlaw method (defined within the Outlaw module namespace, but module_eval'd so you don't need to namespace the file). You can also define constants in your .outlawed file as collections of strings for use in your rules, but that will be addressed below.
Each call to the outlaw method consists of two string arguments, the first an anti-pattern you wish to prohibit usage of in one or more projects, and the second an explanation to be provided when the anti-pattern is detected.
Syntax for rule creation:
Some examples are included in the .outlawed.example file for reference:
outlaw "@@", "Class variables are evil" outlaw "protected", "use private or public, protected is silly in ruby" outlaw "eval", "never eval, rarely class_eval or instance_eval, but never eval" outlaw "module :token end", "nest modules to avoid empty module declarations" outlaw "class :symbol < :core_class", "core classes implemented in c, can cause bad mojo" outlaw :trailing_whitespace
The first three examples are actual ruby keywords and features being outlawed and may not require much explanation except to indicate that they are detected via regular expression matches constructed from the strings, and attempt to use word boundaries intelligently so that eval is detected but not module_eval.
The bottom two examples above use ruby symbols as standin variable or parameter names for identifiers that are matched at runtime with local variables, instance variables, class names and constants that may appear within the ruby program being analyzed. Here, :symbol can be any ruby symbol if it appears only once, though if used multiple times it will only match the the same identifier on subsequent usage. :core_class as used above is a special case where Outlaw has internally defined a constant called CORE_CLASS as a collection of string objects each containing the name of one of ruby's core classes. You can define your own similar collections in the .outlawed file (to be loaded from an external data file preferably, if more than a few values) and then reference CONST_NAME as :const_name in your outlaw anti- patterns as above. Presently mutliple references to the same collection are independent, but if there is interest special handling could be added to also match specific instances of a collection much like the symbol handling.
The last one is a custom rule that calls a method defined on the Outlaw module called 'trailing_whitespace' which returns a Rule object. Any such new methods may be defined in the .outlawed file (since it is module_eval'd into the outlaw namespace) and then outawed the same way.
Outlaw currently ignores whitespace, parentheses and new lines, though I have ideas to change this behavior dynamically in certain rules if desired.
Execute outlaw on your project from the root directory by simply entering "outlaw" into your shell, or specify another directory to run on with "outlaw /path/to/dir"
Before using outlaw in a project you should create a .outlawed file which Outlaw will read rules from.
It comes with an example file (.outlawed.example) which is included in the gem and will be loaded if no .outlawed file is found in current directory or home directory, and will warn you to provide a real file (and provide location of the sample file in your system from the gem installation).
Planned features (unimplemented):
- Customize sensitivity, for instance whitespace is currently ignored, but could enforce style conventions with some whitespace sensitive rules. Also ignores parens, which might be required or prohibited in some context.
Specify AST-nodes of interest, and within them allow arbitrary amounts of code with a :disjoint_code_seperator token.
This should allow, for instance, something like the following, which is not currently possible to outlaw in a useful way:
outlaw ":conditional_branch unless :disjoint_code_seperator else", "If you write unless else and think it makes sense then you are a cylon"
Integrate Rails Best Practices gem, Reek gem, and perhaps others, so that individual issue detections they provide can be added as rules in the outlawed file while ignoring/not running other detection routines.
Automate optional integration with rake task and/or githooks for enforcement/notification of rules in a project.
Specify classes of rules, such as log, warn and prevent for differing behavior regarding violations at runtime.