BeanCounter is a metrics suite for Cappuccino projects. Currently, BeanCounter has support for calculating the following metrics:
- Lines of Code
- Number of Classes
- Number of Files
- Number of Methods
- Number of Test Cases
- Lines of Code per Class
- Methods per Class
- Test Cases per Method
It is very easy to add a new type of metric to BeanCounter (it will be much easier in the future, possibly using a plug-in architecture). If you wish to create a base metric (a metric that does not depend on any other metric), subclass CKMetric and override: - (void)updateMetricForFile:(id)file line:(CPString)line You can also override some other default behavior for more fine grained control, like: - (id)defaultFileMetric - (id)defaultProjectMetric - (id)calculateTotalMetricWithStartingMetric:(id)startingMetric andNewMetric:(id)newMetric - (id)metricForFile:(id)file - (id)metricForProject See the comments on these methods for descriptions on what they should do. See some of the current metrics in Frameworks/BeanCounter/ for an example.
If you wish to create a composite metric (a metric which depends on one or more metric), subclass CKCompositeMetric and implement: - (id)metricForFile:(id)file - (id)metricForProject In many cases, the default behavior (just dividing two metrics) for a composite metric is good enough. In that case, no subclassing is necessary.
To have your new metric be run with the rest of the metrics, add an @import statement to CKMetricSuite.j and add your new metric to the metrics array in that class. See all the examples. (Again, this is an area that is still very experimental. I want this to be much easier so you won't have to do this in the future. But for now, this is how it needs to be done.)
BeanCounter is a Narwhal package and will soon be added to the tusk package manager. When that happens, you can simply run: tusk update sudo tusk install beancounter Thanks to @hammerdr for helping me set this up.
From the top-level directory of your Cappuccino project, run: env NARWHAL_ENGINE=jsc beancounter
Currently, BeanCounter depends on Narwhal and Cappuccino.
There is a lot left to do before BeanCounter is ready for primetime. Here is what is on the list right now:
- "Smarter" code parsing (currently it is line-by-line and naive of where it is in the file/class).
- Tests to "prove" correctness of metrics.
- A plug-in type architecture for easier integration of 3rd-party metrics.
- More metrics (suggestions?).