Patrick R. Michaud via RT wrote:
On Mon, Oct 27, 2008 at 03:47:50AM -0700, Vasily Chekalkin wrote:
I've (slightly) refactor Junction.pir.
1. Get rid of "junction_comparision_helper". Use "infix_junction_helper"
instead.
2. In "infix" and "unary" junction helpers store original value in
ResultHash to avoid stringification of results.
Okay to apply patch as long as spectest_regression still passes.
Yes, it still passing.
I would prefer to eliminate the 'ResultHash' altogether --
elimination of any duplicate values should be done in the
Junction constructor itself, not in the helper. Please don't
I tried this varian before submitting patch, but hit some parrot's bug
that I didn't manage to fix.
close this ticket until that has been done.
Note that all of this effectively disappears when we have
the dispatcher handling Junctions properly -- i.e., the
specialized operator functions defined in Junction really
should not exist.
Indeed. But current state of MMD doesn't allow to implement
auto-threading. For example Parrot_ex_throw_from_c doesn't specify
continuation, etc.
--
Bacek
