• Eldar Insafutdinov <e.insafutdinov gmail.com> Nov 06 2009

Eldar Insafutdinov <e.insafutdinov gmail.com> writes:

Bill Baxter Wrote:

 On Fri, Nov 6, 2009 at 3:24 AM, Eldar Insafutdinov
 <e.insafutdinov gmail.com> wrote:
 I've just written a simple replicate function (the one from std.range does not
work in CTFE):
 T[] replicate(T)(int n, T value)
 {
    T[] ret;
    if (n > 0)
    {
        for(int i = 0; i < n; i++)
            ret ~= value;
    }
    return ret;
 }
 it's not the best implementation, but that's not the thing I want to say
today. When I call it

 string str = replicate(5, 'a');

 I get

 moc.d(370): Error: cannot implicitly convert expression (replicate(5,',')) of
type char[] to string

 That means that char literals are "char", but not "immutable(char)". I
remember the whole conversation about array literals, but char literals are
closer to those of the string. What's the reasoning behind leaving them as
mutable char?


 In Phobos they use std.contracts.assumeUnique to turn char[] to string
 at the end of string processing functions.  It's actually just a cast
 to immutable(char)[].  But the idea is you're asserting that you know
 the char[] data is not referenced by anyone else (your reference is
 unique) so it's safe to make it immutable.  So it documents the intent
 better than a straight cast.  It's also better because if you later
 change the underlying type say from char[] to dchar[], assumeUnique
 will still do the right thing (become a cast to immutable(dchar)[]),
 but cast(string) would silently become bogus.
 
 --bb


Yes you are right, I used cast(string) eventually and made it non-template.
There were other bugs that are specific to CTFE that I could not track down
though.