QegaL

Rule-based fact extraction in a megamodeling context

What is contained:

• The top-level folders correspond to Eclipse projects (that can directly by imported after checkout).

• The top-level project org.softlang.qegal contains the mining related aspects.

• The top-level projects org.softlang.qegal.lang, org.softlang.qegal.lang.ide and org.softlang.qegal.lang.ui contain the Xtext IDE Support.

• The data-set, including the repository lists and all results, are available in the data folder of project org.softlang.qegal (for preserving relative references).

• All available rules for EMF and layout detection are located in the java package org.softlang.qegal.modules as '.qegal' files. The rules can be accessed using a simple text editor or the deployed XText IDE.

• All available (mining) processes are located in the java package org.softlang.qegal.process an can be used to update the data folder.

How to reproduce the inference:

• Import the org.softlang.qegal project into an Eclipse workspace.
• Use Eclipse Maven to automatically build the project.
• Create a file "config.properties" in the project org.softlang.qegal with the following entries:
• temp=<filepath to a temporary folder> (used to check out the target repositories and might require up to 250GB)
• (Optional) login_git=<github-username>
• (Optional) password_git=<github-password> (only needed for the Git API)
• When running Java add the VM Arguments -Xss and -Xmx depending on your system (e.g., -Xss4m -Xmx10000m).
• (Optional) Run the main in QueryGitProcess.java to query recently indexed EMF related files on GitHub. In the data folder, this produces:
  • files_ecore_raw.csv: Query results for .ecore extension.
  • files_eobject_raw.csv: Query results for extends EObject {.
  • files_genmodel_raw.csv: Query results for .genmodel extension.
• (Optional) Run the python script combine_raw.py to combine the data of the previous step and to annotate meta-data to the repositories producing repository_raw.csv.
• (Optional) Run the main in LayoutMiningProcess.java to mine the repository layout detection (rules) on repository_raw.csv to produce repository_layout.csv.
• (Optional) Run the python script filter_vanilla.py to filter the repository_layout.csv and to produce repository_vanilla.csv.
• Run the main in EMFMiningProcess.java to mine the EMF patterns (rules) on repository_vanilla.csv to produce repository_emf.csv (The passed arguments decide on the IO-layer or logging).
• (Optional) Run the python script analyse_emf.py to analyze the repository_emf.csv

How to write and apply a custom mining:

• Create new '.qegal' files in any folder.
• Copy the EMFMiningProcess.java and replace the EMF rules folder by the new folder, if necessary also replace the repository_vanilla.csv by a different selection of repositories. The EMF rules can be extended by combining the rule folders passes as a set of paths.

How to install XText IDE support:

• Import all projects into the workspace.
• Right-click on the workspace and select Export>Plug-in Development>Deployable plug-ins and fragments.
• Check all boxes for the three projects and select Install into host. Repository:.
• Finally install by clicking Finish. Wait until Eclipse finished building. Then, restart (a dialog will ask you to).
• '.qegal' files can now be opened using the IDE Support.