I'd like to be able to use the reduce template on a range of strings. The following:

int main(char[][] args)
{
    string[] tstarr = ["one", "two", "three"];
    writefln(reduce!("a ~ b")("", tstarr));
}

gives the following error message:

c:\dmd\bin\..\src\phobos\std\algorithm.d(289): Error: functions cannot return static array invariant(char[0u])

Is there another way to do this? I've tried both the ~ operator and the + operator.

Brian

Brian Myers:
> I'd like to be able to use the reduce template on a range of strings. The following:
> int main(char[][] args)
> {
>     string[] tstarr = ["one", "two", "three"];
>     writefln(reduce!("a ~ b")("", tstarr));
> }

Don't do that, probably it's a slow way to join an array of strings.
In the std.string module you can use the join("") function, or you can use joinArr() function from the d.string module of mine.
(I have like a dejavu, isn't this discussion happened already?)

Bye,
bearophile

Well yes and no. I have posted about the map and reduce templates before a few days ago, but this is with a different function -- string concatenation.

I'm just exploring how well D supports the functional paradigm. 2.0 is bringing in a lot of functional features, and I'm trying to get a feel for the boundaries.

Brian

bearophile Wrote:

> Brian Myers:
> > I'd like to be able to use the reduce template on a range of strings. The following:
> > int main(char[][] args)
> > {
> >     string[] tstarr = ["one", "two", "three"];
> >     writefln(reduce!("a ~ b")("", tstarr));
> > }
> 
> Don't do that, probably it's a slow way to join an array of strings.
> In the std.string module you can use the join("") function, or you can use joinArr() function from the d.string module of mine.
> (I have like a dejavu, isn't this discussion happened already?)
> 
> Bye,
> bearophile

Brian Myers wrote:
> The following:
> 
> int main(char[][] args)
> {
>     string[] tstarr = ["one", "two", "three"];
>     writefln(reduce!("a ~ b")("", tstarr));
> }
> 
> gives the following error message:
> 
> c:\dmd\bin\..\src\phobos\std\algorithm.d(289): Error: functions cannot
> return static array invariant(char[0u])

This is just the old problem that string literals are static arrays. Reduce expects its return type to have the same type as the first argument, which is invariant(char[0]) in your case. Replace "" by ""[] and the code compiles.

"Christian Kamm" <kamm-incasoftware@removethis.de> wrote in message news:g27rd9$1l7d$1@digitalmars.com...
> Brian Myers wrote:
>> The following:
>>
>> int main(char[][] args)
>> {
>>     string[] tstarr = ["one", "two", "three"];
>>     writefln(reduce!("a ~ b")("", tstarr));
>> }
>>
>> gives the following error message:
>>
>> c:\dmd\bin\..\src\phobos\std\algorithm.d(289): Error: functions cannot
>> return static array invariant(char[0u])
>
> This is just the old problem that string literals are static arrays.
> Reduce
> expects its return type to have the same type as the first argument, which
> is invariant(char[0]) in your case. Replace "" by ""[] and the code
> compiles.
>

1. Why does reduce need the seed to be manually specified anyway? Why not just take the first element of the first range as the seed?

2. If it did that, that still wouldn't solve the problem from the original post, would it? I assume you'd need to do something weird like "auto tstarr = ["one"[], "two"[], "three"[]];"?

3. I'm still getting my head around D2's strings so this'll probably sound stupid, but why can't "" be implicitly converted to ""[]? (and how exactly does that solve the original problem anyway)?

BTW, ""[] means "a full start-to-end slice of the empty string literal", right? Also, I know arrays and array slices are distict types, but what exactly is the difference? My intuition would tell me that they're both a pointer and a length, and an array slice is simply one that just happens to point to a portion of another array, but lately I've been getting the impression that there's something more to it than that?

"Nick Sabalausky" <a@a.a> wrote in message news:g371nu$275q$1@digitalmars.com...
>
> 1. Why does reduce need the seed to be manually specified anyway? Why not just take the first element of the first range as the seed?

It could.  Some languages/libraries do.

> 2. If it did that, that still wouldn't solve the problem from the original post, would it? I assume you'd need to do something weird like "auto tstarr = ["one"[], "two"[], "three"[]];"?

You don't need to put the slice on every element, just the first.  That makes the compiler infer the type of the array literal as char[][] instead of char[3][].

> 3. I'm still getting my head around D2's strings so this'll probably sound stupid, but why can't "" be implicitly converted to ""[]? (and how exactly does that solve the original problem anyway)?

As said above, the compiler is too dumb to figure out that ["one", "two", "three"] is a char[][] and not a char[3][].  If string literals were dynamic arrays instead of fixed-size, we wouldn't need the [] at the end of the first element.

> BTW, ""[] means "a full start-to-end slice of the empty string literal", right? Also, I know arrays and array slices are distict types, but what exactly is the difference? My intuition would tell me that they're both a pointer and a length, and an array slice is simply one that just happens to point to a portion of another array, but lately I've been getting the impression that there's something more to it than that?

No, they're not different types.  At least not yet.  Walter was throwing around the idea but it hasn't been implemented yet.  There is some ickiness with representing both as the same type.

Getting a full slice of a fixed-size array, such as a string literal, gives you a dynamic array reference.  Hence why it fixes the array literal.