This repository hosts the static analyzer. It analyzes the control flow and data flow, and identifies each ORM APIs that issues DB queries.

This analyzer can help find some anti-patterns mentioned in the paper. For example, by checking how the query result is used in the application, it understands where unnecessary data retrieval happens.

Here's step of how to run it:

0. Run the static analysis:

$cd controller_model_analysis

• run

$ruby main.rb --help

to get the options

• Basically, to get the statistics and program flow of a specific action from an app, you would like to run:

$ruby main.rb -d DIR_TO_APP_FILES -s CONTROLLER_NAME,ACTION_NAME

For example,

$ruby main.rb -s -d ../applications/forem PostsController,index

The instructions in the form of JRuby intermediate representation that will be executed by an action, together with the data dependencies among these instructions, will be printed to the screen.

• If you wish to run all actions from an application, do:

$ruby main.rb -a -d DIR_TO_APP_FILES

Results will be saved to the folder under the app directory. The folder name is defined under global.rb, and by default is /result/.

1. Analysis

The data flow information can be used to find anti-patterns. An example is shown in compute_redundant_field_access.rb, which returns the fields retrieved by a query but not used by the action.

2. Applications:

There are already many applications that are preprocessed and ready to be analyzed by this tool. These apps lie under the applications/ directory.

To analyze your own applications, there are some preprocessing work to do: (not completed yet...)

• copy the ruby files to some directory APPDIR, under the applications/ folder:

$cd applications/
$mkdir APPNAME/
$cd ..

• in your app, run:

$rake routes | tail -n +2 | awk '{ for (i=1;i<=NF;i++) if (match($i, /.#./)) print $i}' | sed -e 's/#/,/g' | sort | uniq

to get the list of actions that can be invoked by the app, copy it to APPDIR/calls.txt

• go to preprocess\_views/ folder, and check the README.md there to extract the code from view.

If there are only .erb files in views, it is probably fine. But if you are using haml, the convertion from haml to erb may run into some trouble.

• copy schema.rb in your app to APPDIR/.

• use JRuby to "compile" your app into JRuby IR. The modified JRuby is also in this repository. Follow the jruby/README.md to deploy jruby.

Then go to the applications folder and run the script:

$cd applications/
$python generate_dataflow.py APPNAME