PDE UI Incubator ApiTools ReferenceExtractor Dev
Reference Extractor Tool – Developer’s docs
The reference extraction tool is there to answer the following three questions:
- Who is using my APIs/non-APIs
- What my code is using
- Identify "internal" methods being used
To clarify this, let’s consider a picture with 3 bundles: MyBundle, UserBundle, and SystemBundle. Let’s say that my code in the MyBundle and I am using the SystemBundle. Somebody else wrote UserBundler that uses both MyBundle and SystemBundle.
To get answers to all the questions above we would run reference extractor on all 3 bundles and store results in the repository. At a later time repository could be queried. Developer working on MyBundle coudl discover where his code is used even if he never heard about UserBundle.
Processing of queries will aggregate results so that question on who uses SystemBundle’s code will contain both MyBundle and UserBundle.
The reference extractor uses custom parser for .class files. Parser extracts all references with possible exception of references by reflection. (Potentially, those could be added with some degree of "fuzziness" by processing strings contained in the .class files.)
The .class files have references in an "unresolved" format. References are lazily resolved when queried; classes in the package “org.eclipse.pde.api.tools.internal.model.ref” used to store and process references.
The data extracted is grouped by elements being referred and stored in the XML file. For instance:
<methodReference methodName="log" parameters="org.eclipse.core.runtime.IStatus" type="void">
The reference extractor itself is working.
However, storage format needs refinement and processing of queries has only the rudimentary code.
Common code base
Some code is shared between API comparison tool and reference extractor. Parser for the .class files; determination of API/non-API status, and repository support are all shared as well as UI implementation details and Java models.
Aside from completing queries and making it all work, the most interesting potential development comes from opening free-form queries to the combined data extracted by API comparison tool and reference extractor. If SQL database is used as a repository, potentially Web access could be allowed to the database with support of user-defined queries. Application could be: determining popularity of APIs, finding code usage patterns, identifying clients.
PDE Incubator - New projects that might be incorporated into PDE in future.