Hairball is a plugin-able framework useful for static analysis of Scratch projects.
The paper and presentation slides for Hairball can be found at: http://cs.ucsb.edu/~bboe/p/cv#sigcse13
A Hairball demo web service is running and available at: http://hairball.herokuapp.com
With a proper python environment (one which has pip
available), installation
is as simple as pip install hairball
. easy_install
can also be used via
easy_install hairball
.
To install from source, first checkout this project and then navigate your
command-line interface to the outer hairball directory that contains
setup.py
. Then run python setup.py install
.
Once installed, to see how to use hairball run hairball --help
. That will
produce output similar to the following:
Usage: hairball -p PLUGIN_NAME [options] PATH...
PATH can be either the path to a scratch file, or a directory containing
scratch files. Multiple PATH arguments can be provided.
Options:
--version show program's version number and exit
-h, --help show this help message and exit
-d DIR, --plugin-dir=DIR
Specify the path to a directory containing plugins.
Plugins in this directory take precedence over
similarly named plugins included with Hairball.
-p PLUGIN, --plugin=PLUGIN
Use the named plugin to perform analysis. This option
can be provided multiple times.
-k KURT_PLUGIN, --kurt-plugin=KURT_PLUGIN
Provide either a python import path (e.g, kelp.octopi)
to a package/module, or the path to a python file,
which will be loaded as a Kurt plugin. This option can
be provided multiple times.
Below are a list of available plugins that can be used as the -p PLUGIN_NAME
option:
- blocks.BlockCounts
- blocks.DeadCode
- checks.Animation (not fully tested)
- checks.BroadcastReceive
- checks.SaySoundSync (not fully tested)
- duplicate.DuplicateScripts
- initialization.AttributeInitialization
- initialization.VariableInitialization (not fully tested)
Note: The output for each plugin is not yet completely standardized. Please feel free to file any issues or make improvements and send pull requests.
The python Kurt package unfortunately is pretty slow to parse Scratch 1.4 (and similar) files. To remedy this situation, Hairball has built-in support for caching a serialized version of the Kurt object. On subsequent passes through the same data you should notice a TREMENDOUS speed improvement.
Note: At the moment the cache is unbounded, so keep an eye on your disk space. The cache location can be discovered by running:
python -c "import appdirs; print appdirs.user_cache_dir('Hairball', 'bboe')"