[bitc-dev] Implications of pair
Jonathan S. Shapiro
shap at eros-os.org
Mon Aug 29 14:32:45 EDT 2005
On Sun, 2005-08-28 at 16:01 +0700, Constantine Plotnikov wrote:
> I think that there is an option to have one element tuple and still to
> have a bit more regular processing if we would not try to save on amout
> of types involved at compile time. The approach is less effient at
> compile time as there is always terminator head field of unit type.
Your approach would work, but it is *very* important to reduce the
number types in this area. Any type that we decide to use for argument
formation is necessarily a foundational type for purposes of language
semantics, and the smaller we keep that the sooner we will be able to do
reasoning about programs.
Yesterday we made the decision to eliminate implicit pair construction
and simply let procedures take multiple arguments.
More information about the bitc-dev