std.conv.to of lazy range to numbers array

[dlangBugzillaToGithub]

bearophile_hugs reported this on 2013-06-18T10:47:23Z

Transfered from https://issues.dlang.org/show_bug.cgi?id=10407

CC List

• hsteoh

Description

import std.regex: splitter, regex;
import std.conv: to;
import std.array: array;
void main() {
    auto s = splitter("10 30 20", r"[\s,]+".regex);
    auto r1 = s.array.to!(real[]); // OK
    auto r2 = s.to!(real[]); // Error
}


DMD 2.063.2 gives:

...\dmd2\src\phobos\std\conv.d(276): Error: template std.conv.toImpl does not match any function template declaration. Candidates are:
...\dmd2\src\phobos\std\conv.d(332):        std.conv.toImpl(T, S)(S value) if (isImplicitlyConvertible!(S, T) && !isEnumStrToStr!(S, T) && !isNullToStr!(S, T))
...\dmd2\src\phobos\std\conv.d(445):        std.conv.toImpl(T, S)(ref S s) if (isRawStaticArray!S)
...\dmd2\src\phobos\std\conv.d(461):        std.conv.toImpl(T, S)(S value) if (is(typeof(S.init.opCast!T()) : T) && !isExactSomeString!T)
...\dmd2\src\phobos\std\conv.d(492):        std.conv.toImpl(T, S)(S value) if (!isImplicitlyConvertible!(S, T) && is(T == struct) && is(typeof(T(value))))
...\dmd2\src\phobos\std\conv.d(542):        std.conv.toImpl(T, S)(S value) if (!isImplicitlyConvertible!(S, T) && is(T == class) && is(typeof(new T(value))))
...\dmd2\src\phobos\std\conv.d(276):        ... (16 more, -v to show) ...
...\dmd2\src\phobos\std\conv.d(332): Error: template std.conv.toImpl cannot deduce template function from argument types !(real[])(Splitter!(string, Regex!char))
...\dmd2\src\phobos\std\conv.d(276): Error: template instance toImpl!(real[]) errors instantiating template
test.d(7): Error: template instance std.conv.to!(real[]).to!(Splitter!(string, Regex!char)) error instantiating

[dlangBugzillaToGithub]

hsteoh commented on 2015-09-08T17:11:58Z

Is this a real bug? Couldn't you just do `s.map!(a => a.to!real).array` to do what you want?

[dlangBugzillaToGithub]

bearophile_hugs commented on 2015-09-18T08:55:47Z

(In reply to hsteoh from comment #1)
> Is this a real bug?

No, it's an enhancement request...


> Couldn't you just do `s.map!(a => a.to!real).array` to do what you want?

Yes, but I think converting lazy ranges is a basic and common need, so it should be supported by "to".