In high-level languages, I really like named, default and optional arguments, primarily because it allows one to easily and concisely specify embedded domain specific languages [1]. Default arguments also significantly cut down on the number of overloaded functions needed at complex interface boundaries, and you can ensure forward compatibility by lifting default values to static variables in the module, instead of embedding it at the call site.
Named arguments also allows you to unify activation records and ordinary language records. Not sure whether it's appropriate for BitC, but the feature itself is certainly very nice. I also agree with Eric that all-or-nothing seems like a perfectly acceptable approach to cut down on complexity. Sandro [1] See Ocsigen's defined a statically typed XHTML interface, http://ocsigen.org/docu/1.1.0/Xhtmltypes_duce.html Eric Rannaud wrote: > On Fri, Mar 06, 2009 at 01:41:31PM -0500, Jonathan S. Shapiro wrote: >> We seem to have two broad options for procedure application: I'm not >> entirely convinced that the second one works if we retain n-ary functions >> and we decide to do mixfix: >> >> Option 1: >> >> proc-name( arg ... arg) >> >> Option 2: >> >> proc-name arg ... arg > > I realize you're not polling on this particular question, but this may > have an impact on your decision for procedure application syntax. > > Have you considered supporting keyword arguments (aka. labels, named > arguments)? > > This makes code audit a lot easier and safer: > > set-size :: (box: Box, width: int, height: int) -> void > set-size box: b width: w height: h > set-size b h w // wrong, but hard to catch > > where the argument types are not all different. > > As usual here, there is the question of whether you allow using both > positional arguments and keyword arguments at the same time, and how. > There is the Common Lisp way (positional args appear first, then keyword > args), the OLabl / OCaml way (non labeled arguments are extracted first > as positional arguments, then labeled arguments are extracted; check for > conflicts), the Python way (similar to OCaml, but keyword arguments are > put in a hash at runtime), etc. > > My preference goes toward allowing only two modes: all positional or all > named, no mix. And of course, all of this handled at compile time. > > Another question: can any procedure be called using keywords (Ada, VHDL, > Verilog; my strong preference)? Or does it have to be explicit in the > procedure definition (Ocaml, Lisp, Python)? > > You also have the questions of whether you can have default values, when > to evaluate them, from what kind of expressions, etc. You may want to > allow default values only when all arguments are specified with a > keyword in a call. Default values can be problematic if the default is > changed behind the back of the caller 6 months later. I personally don't > seem them as essential, and somewhat dangerous. > > Thanks. > _______________________________________________ > bitc-dev mailing list > [email protected] > http://www.coyotos.org/mailman/listinfo/bitc-dev _______________________________________________ bitc-dev mailing list [email protected] http://www.coyotos.org/mailman/listinfo/bitc-dev
