Using C, C++, Java, et al syntax, I would prefer if the statements had to always be in brackets as
if (test1) {
if (test2) {
expr1
}
} else {
expr2
}
Doing this, there is no shift-reduce conflict and the code is only very marginally harder to write (a few extra brackets).

Right now, printf does this: “Your name is %s %s”, “First”, “Last”
It should be doing, say, that: “Your name is %1,s %2,s”, “First”, “Last”

Then any language where one usually gives Last Name, First Name, can be trivially accommodated ;) The requisite changes to implement it in printf are minor, it’s more work to get gcc to actually support it w/o barfing, but it’s not astronomical amount of work either.

Qt toolkit has it a tad better: QString(”Your name is %1 %2″).arg(foo).arg(baz). Although all formatting is given to arg(), which makes it non-localizable, but at least the order can be changed at will…

Cheers!

Mike Lin: If this is true, then Ocaml + twt is the first language I’ve heard of that handles this correctly. From experience, Haskell doesn’t- and it’s a royal pain in the tush to track down code that looks like it’s indented correctly, but isn’t.

Greg Moulton: The deal breaker with me and Erlang is the lack of built-in, strong static typing.

John Veer: I’ve fixed the title. And, I’ll comment: I moved 1500 miles to a strange city in order to use Ocaml professionally- what does that say about my opinion of Ocaml?

Hamlet D\’Arcy: It may be related- in either case, the idea has been around for a while. I know that both the HP and IBM C and Fortran compilers have attempts at effectively the same thing. The problem is that it’s damned hard to implement in languages with unrestricted side effects, as proving that the code transformation has the same behavior both before and after is, in the general case, equivalent to the halting problem.

In C, I can use gettext to extract strings and translate them, shuffling the order around as needed to accommodate for different languages’ grammars. This is impossible in C++, because not only is the message to be translated hacked into contextless pieces, it’s also hardcoded into the source, so you have no chance to rearrange it properly. In effect, iostreams fundamentally disallows you to have translated apps. And in this day and era, an IO subsystem that doesn’t allow me to create multilingal versions without shipping multiple binaries is just worthless crap, no matter how grand anything else about it might be.

Erlang doesn’t have mutable data, and has only a limited number of ways to perform effects, but since any call to another module is potentially unsafe (even, due to dynamic code loading, it isn’t at compile time), you’d only ever be able to apply deforrestation on code that completely resides in one module, and is completely free of dangerous constructs (side effects).