Mirror of forum.nim-lang.org

461 :: Garbage collection for foreign objects?

• Stefan_Salewski (orginal) [2014-06-02T02:45:13+02:00] view original

  After having read the tutorial and part of the nimrod manual, I still wonder if it is possible with nimrod to have garbage collection when interfacing foreign libraries. For GTK we generally have not to care much about memory management, ref-counting works not bad for that case... But of course there exists other libraries, like boost or cgal, which offer its own free(), dispose() or what ever. For Ruby, with its mark- and sweep garbage collector, it is generally possible to implement garbage collection for those objects by writing some glue code. (For a few cases, i.e. when objects are stored hierarchical, like faces in a cgal delaunay triangulation, it is not that easy I guess...) Is there any support for such in nimrod planned? Of course, interfacing with C or C++ libraries is not really nice, but I guess it will take much time to write only a few of CGAL's or Boost's functions like "Constrained Delaunay Triangulation", RTREE, or Fibonacci Queue in Nimrod. I just found an article about Go-langs attempts to do garbage collection for GTK3 -- https://blog.conformal.com/gotk3-gtk3-the-go-way/ -- not read already, sorry.

---

• Jehan (orginal) [2014-06-02T09:05:58+02:00] view original

  It depends on what you need (I'm not sure about your specific requirements), but I think it should be possible for the most part.

  With the reference counting GC, you can use finalizers in conjunction with GC_ref() and GC_unref() to hook into the GC process (cycles can be an issue if you create any). Use new(some_ref_var, finalizer_proc) to allocate a finalized object. I haven't looked at the M&S GC in detail, but my understanding is that the same approach should work there, too (objects with a non-zero reference count due to GC_ref() being treated like roots).

  If your external memory does not contain any references to GCed memory, then finalizers on their own will be sufficient.

  With the Boehm GC, no extra effort should be needed as long as you allocate C data structures with GC_malloc() and friends.

---