I'm trying to figure out how to use reduce with multiply funs.

Here's my try on minmaxElement:


import std.typecons: tuple;

/** Returns: Tuple of Minmum and Maximum Element in X. */
auto minmaxElement(alias F = min, alias G = max, R)(in R range)
    @safe pure nothrow if (isInputRange!R)
{
    return reduce!(F, G)(tuple(ElementType!R.max,
                               ElementType!R.min), range);
}

unittest { assert([1, 2, 3].minmaxElement == tuple(1, 3)); }


which errors as

/home/per/opt/x86_64-unknown-linux-gnu/dmd/bin/../include/d2/std/algorithm.d(774,29): Error: can only initialize const member _expand_field_0 inside constructor
/home/per/opt/x86_64-unknown-linux-gnu/dmd/bin/../include/d2/std/algorithm.d(774,29): Error: can only initialize const member _expand_field_1 inside constructor
algorithm_ex.d(157,25): Error: template instance std.algorithm.reduce!(min, max).reduce!(Tuple!(const(int), const(int)), const(int)[]) error instantiating
algorithm_ex.d(160,28):        instantiated from here: minmaxElement!(min, max, const(int)[])


If I replace ElementType!R.min and ElementType!R.max with a values, say 0 and 1 it compiles. What is going on here?

This seems related:
http://forum.dlang.org/thread/bug-10408-3@http.d.puremagic.com/issues/

Nordlöw:

> I'm trying to figure out how to use reduce with multiply funs.

import std.algorithm: min, max, reduce;
import std.typecons: tuple, Unqual;
import std.range: isInputRange, ElementType;

/// Returns: Tuple of Minmum and Maximum Element in X.
auto minMax(alias F = min, alias G = max, R)(in R range)
@safe pure nothrow if (isInputRange!R) {
    return reduce!(F, G)(tuple(Unqual!(ElementType!R).max,
                               Unqual!(ElementType!R).min), range);
}

unittest {
    assert([1, 2, 3].minMax == tuple(1, 3));
}

void main() {}


> If I replace ElementType!R.min and ElementType!R.max with a values, say 0 and 1 it compiles. What is going on here?
>
> This seems related:
> http://forum.dlang.org/thread/bug-10408-3@http.d.puremagic.com/issues/

You can't use reduce with a const seed.

Bye,
bearophile

On 03/31/2014 03:13 PM, bearophile wrote:
> Nordlöw:
>
>> I'm trying to figure out how to use reduce with multiply funs.

Could someone please check the following. It looks like a compiler bug on git head (DMD64 D Compiler v2.066-devel-75159e4):

import std.algorithm;

int foo(int value)
{
    return value;
}

void main()
{
    reduce!(foo, foo)(tuple(0, 0), [ 1 ]);
}

statement.c:274: ErrorStatement::ErrorStatement(): Assertion `global.gaggedErrors || global.errors' failed.

Ali

> You can't use reduce with a const seed.
>

This, surely, must be a compiler bug right?

/Per

Ali Çehreli:

> statement.c:274: ErrorStatement::ErrorStatement(): Assertion `global.gaggedErrors || global.errors' failed.

It's a little compiler bug, that should go in Bugzilla (even if perhaps it was recently added).

Bye,
bearophile

On 03/31/2014 04:01 PM, bearophile wrote:
> Ali Çehreli:
>
>> statement.c:274: ErrorStatement::ErrorStatement(): Assertion
>> `global.gaggedErrors || global.errors' failed.
>
> It's a little compiler bug, that should go in Bugzilla (even if perhaps
> it was recently added).

  https://d.puremagic.com/issues/show_bug.cgi?id=12501

Ali

On Monday, 31 March 2014 at 22:25:40 UTC, Nordlöw wrote:
>> You can't use reduce with a const seed.
>>
>
> This, surely, must be a compiler bug right?
>
> /Per

Arguably, it's a user bug ;) The user should have provided a non-const seed.

*But*, there have been many cases of reduce being made to accept const seeds before, via unqualified copy. This case must have just been a missed one.

In any case, I had submitted a re-write for reduce, and it just so happens to support this. So I added your code to the test cases:
https://github.com/D-Programming-Language/phobos/pull/2060

> so happens to support this. So I added your code to the test cases:

Great!

BTW: Why is static qualifier needed on definition of minmaxElement() in the unittest?

On Wednesday, 2 April 2014 at 09:25:53 UTC, Nordlöw wrote:
>> so happens to support this. So I added your code to the test cases:
>
> Great!
>
> BTW: Why is static qualifier needed on definition of minmaxElement() in the unittest?

Whenever you declare something in a nested context, it may or may not have a hidden context pointer, depending on the type, and the implementation.

"static", in this context, ensures this does not happen. It's not actually *needed* in this context though. It's more of a matter of style.

In your example, the function was declared in global context, so the "static" would have been gratuitous.