Using the Eclipse C/C++ Development Tooling as a Robust, Fully Functional, Actively Maintained, Open Source C++ Parser

Open Source parsers that support contemporary C/C++, can recover from errors, include a preprocessor, and that are actively maintained, are rare. This work describes how to use the parser contained in the Eclipse C/C++ Development Tooling (CDT) as a Java library. Such parser provides not only the abstract syntax tree of the parsed le but also the semantics, i.e., type information and bindings. The authors used the same approach to obtain Java and JavaScript parsers. Programming language parsers are used by industry and research to create compilers or interpreters, statical code analysis tools, code metrics tools, source code editors with code completion, etc. Parsing C/C++ is particularly tricky (e.g., a construct a * b can be a multiplication or a pointer de nition depending on the type of a). Generic Open Source parser generators do not alleviate this task since the ambiguities cannot be resolved by a parser alone but require type information. We searched for Open Source C++ parsers that include a preprocessor, perform semantic analysis (resolve type information and name bindings), are robust, and support contemporary C/C++ features. Of the found parsers, namely cpp-ripper, Elsa, GCC using the -fdump-translation-unit option, GCC_XML, Clang, and the Eclipse CDT parser only the latter 2 ful lled the requirements. We decided to opt for the Eclipse parser since the approach could also (and did) provide us with parsers for other languages like Java and JavaScript. The actual parser is located in the le org.eclipse.cdt.core_X.jar , in the Eclipse installation folder, where X stands for version of the le. The jar itself is an Eclipse plugin, however, it is possible to use it as a Java library, without initializing the Eclipse platform. The instruction org.eclipse.cdt.core.dom.ast.gnu.cpp.GPPLanguage. getDefault().getASTTranslationUnit(FileContent, IScannerInfo, IncludeFileContentProvider, IIndex, int, IParserLogService) performs the actual parsing, returning an abstract syntax tree (AST). This work deals with an (apparently) simple problem: to nd a working C++ parser . Eclipse CDT contains such parser, but there is no o cial documentation about using it as a library outside of Eclipse. We hope that our poster can ll this gap and be of help.