ATL Use Case - Visual Representation for Code Clone Tools

Code Clone, SVG, XML.

A number of code clone detection tools have been developed to aid the detection of code clones in sources files. However, most of the tools ouput the analysis results in different formats, and nearly all of them are based on text, which makes it difficult to compare and incorporate these results. In this work, we first difined specific DSLs for different code clone detection tools (Simian, SimScan, CloneDr) and a generic DSL for code clone. Then, specific DSL models can be transformed to the generic DSL for code clone, which will be continuely transformed to SVG model and finally generate the SVG code.

Code clone detection and SVG(two-dimensional graphics and graphical applications in XML) are the two main domains in this project.

In this use case, AMMA and three of its core DSLs: KM3, ATL, and TCS will be used. AMMA is built on a model-based vision of DSLs. AMMA provides a set of core DSLs that may be used to specify new DSLs.

Code Clone Detection Tools can be used to analyze the source programs and find the same or similar parts in these programs. A text result is often generated after the analysis process, which reports the location of the code clones and other related information such as the length of clones, and similarity, etc. Three tools have been supported by our work. They are Simian, SimScan, and CloneDr. The following is is an excerpt of the result outputed by Simian.

Code clone analysis result by Simian

The output contains some clone groups. In each group, the specific locations of the clones in each file are given. The abstract syntax is specified in KM3. The following figure is an excerpt of the Simian metamodel.

Simian metamodel excerpt

Concrete syntax of Simian has been implemented in TCS. A grammar is automatically derived from both the KM3 metamodel and the TCS model to parse Simian output file into Simian models.

FileInfo DSL is defined to record the total number of lines in each source file. This number is not available in the output result, but necessary for the visual representation. So we defined such a domain and make it as a part of the input together with Simian. As for Simian, the metamodel of FileInfo is defined by KM3, and the concrete syntax is defined by TCS.

FileInfo source excerpt

Generic DSL for Code Clone is the uniform model to which specific DSL models can be transformed. Most of the concepts in this DSL are very similiar with the ones in specific DSLs, because they are in the same domain. The following is an excerpt of the generic DSL metamodel.

Generic DSL for code clone metamodel excerpt

A transformation from specific DSL model to generic DSL model can be implemented by Simian2CodeClone.atl.

Simian to CodeClone transformation excerpt, written in ATL

Then, after defining SVG and XML metamodels in KM3, a second transformation from generic Code Clone DSL to SVG can be completed.

Generic DSL for Code Clone to SVG transformation excerpt, written in ATL

Finally, the SVG model is transformed to XML model by SVG2XML.atl so that the SVG code can be extracted automatically. The following is an excerpt of the generated SVG code.

Generated SVG code excerpt

And two kinds of visual representations are shown in the following.

Visual representation Form 1

Visual representation Form 2

The following picture better demonstrates the full scenario of this project.

Full transformation scenario

None at the current time.