[ruby-core:66165] [ruby-trunk - Feature #10195] [Open] Alternative to rb_proc_new that provides access to self, super and passed block

Issue #10195 has been updated by Nobuyoshi Nakada.

Description updated
Status changed from Closed to Open

Passed block is passed as the fifth argument, as the test in r48350.



----------------------------------------
Feature #10195: Alternative to rb_proc_new that provides access to self, super and passed block
https://bugs.ruby-lang.org/issues/10195#change-49864

* Author: Charles Strahan
* Status: Open
* Priority: Normal
* Assignee: 
* Category: 
* Target version: 
----------------------------------------
I'm developing a Haskell binding to the Ruby C API. Unfortunately, `rb_proc_new`
does not provide any means of accessing super, self and/or any passed block. A
better alternative to `rb_proc_new` would make it possible to create procs from
C/Haskell/etc that can access `self`/`super`/&`block`.

I'm currently hacking around this lacking in the C API by creating a closure
from the Ruby side that closes over self et al - a trampoline of sorts:

~~~
  rupee_proc_constructor_with_super = rb_eval_string_protect(
      "Proc.new {|callable|\n"
      "  su = if defined?(super)\n"
      "         Proc.new {|*args,&blk| super(*args,&blk) }\n"
      "       else\n"
      "         nil\n"
      "       end\n"
      "  Proc.new {|*args,&blk|\n"
      "    callable.call(self, su, *args, &blk)\n"
      "  }\n"
      "}",
      &status);
~~~

From the Haskell side, I then pass in the `callable' object, where `callable#call`
is implemented in Haskell. This is pretty clunky/slow, and also creates
unintuitive behavior from the Haskell side. It would be ideal for the C API to
support this natively.



-- 
https://bugs.ruby-lang.org/