On Fri, Mar 6, 2009 at 12:04 PM, Jonathan S. Shapiro <[email protected]> wrote: > - Show quoted text - > On Fri, Mar 6, 2009 at 2:46 PM, Gelf Mrogen <[email protected]> wrote: >> >> > The problem is that in things like >> > f a b + f c d >> > we do not know how many arguments to "consume" for f until f is typed, >> > and we don't have that information at parse time. >> >> Can you explain this more? Why not interpret "f a b + f c d" exactly as >> you'd interpret "(f a b) + (f c d)"? > > Because of my assumption that "+" and "a" are both identifiers. I > didn't know about the ML rule that mixfix operators must be > punctuation. It raises a conundrum, though, because now I need to go > look at the UNICODE standard and see if there is an appropriate > character class for what we want, or if not, then what it would take > to build something of that sort.
If you're going the mixfix route, and will be parsing expressions into lists of tokens that are turned into trees by a separate parsing phase, then I don't think this restriction is necessary. You'll know which identifiers are mixfix operators by the time you parse, which lets you control their precedence relative to function application. Geoffrey _______________________________________________ bitc-dev mailing list [email protected] http://www.coyotos.org/mailman/listinfo/bitc-dev
