Skip to content
Permalink
Branch: master
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
22 lines (17 sloc) 2.18 KB

Limitations

The long-term goal is for FastR to be a high-performance drop-in replacement for the reference implementation (open source GNU R). However, this is a challenging goal for a language and ecosystem as rich as R's. At the time of writing FastR implements almost the entire language, but some packages do not install or run correctly. For pure R packages this is generally due to incomplete implementations of the R builtin functions; either they are not implemented at all or some argument types are not accepted.

Native Code

GNU R makes extensive use of native code, mostly C, to implement aspects of the default packages and, of course, package developers are encouraged to follow the same strategy to finesse performance problems. The native extensions interface is a difficult fit for a Java implementation such as FastR as it is strongly biased towards the existing GNU R implementation. In contrast Java has a very abstract and portable native code interface, but it is difficult and inefficent to map this to the R interface. Especially problematic are packages that actually depend on the internal details of GNUR, which can be enabled by setting USE_RINTERNALS in native code. FastR cannot support this at all. Where this is being used by a package simply as a performance speedup (some function calls become C macros), the package will still build, but run more slowly. However, where this is genuinely being used to access the GNU R internal data structures, it will fail to build. Note that whereas native C code produces a performance speedup with GNU R it will likely run slower on FastR due to the overhead of the JNI interface. We are actively working on a solution to this problem.

In addition many of the default packages, e.g. stats, graphics contain a lot of C code that may use R internals. FastR has taken a mixed approach to implementing these "package" builtins. Some have been translated to Java, some that do not depend on the GNU R internals are accessed through the native interface, but many are not implemented. FastR is adopting "as-needed" approach to these functions.

You can’t perform that action at this time.