www.digitalmars.com         C & C++   DMDScript  

digitalmars.D.learn - makeIndex not working

reply Johann MacDonagh <johann.macdonagh..no spam..gmail.com> writes:
I'm confused, what am I doing wrong here? This example came from the 
std.algorithm documentation.

import std.algorithm;

void main()
{
     immutable int[] arr = [ 2, 3, 1, 5, 0 ]; // index using pointers
     auto index1 = new immutable(int)*[arr.length];
     makeIndex!("a < b")(arr, index1);
}

Error is:

main.d(7): Error: template std.algorithm.makeIndex(alias less = "a < 
b",SwapStrategy ss = SwapStrategy.unstable,Range,RangeIndex) if 
(isForwardRange!(Range) && isRandomAccessRange!(RangeIndex) && 
is(ElementType!(RangeIndex) : ElementType!(Range)*)) does not match any 
function template declaration
main.d(7): Error: template std.algorithm.makeIndex(alias less = "a < 
b",SwapStrategy ss = SwapStrategy.unstable,Range,RangeIndex) if 
(isForwardRange!(Range) && isRandomAccessRange!(RangeIndex) && 
is(ElementType!(RangeIndex) : ElementType!(Range)*)) cannot deduce 
template function from argument types !("a < 
b")(immutable(int[]),immutable(int)*[])
main.d(7): Error: template instance errors instantiating template
Jul 02 2011
parent reply bearophile <bearophileHUGS lycos.com> writes:
Johann MacDonagh:


 I'm confused, what am I doing wrong here?


I think that Phobos needs way more&better unittests.

I think the error you receive reduces to this:

import std.range: isForwardRange;

void foo(R)(R r) if (isForwardRange!R) {}

void main() {
     immutable arr = [1, 2];
     foo(arr);
}

Bye,
bearophile
Jul 02 2011
parent reply Johann MacDonagh <johann.macdonagh..no spam..gmail.com> writes:
On 7/2/2011 9:00 PM, bearophile wrote:

 Johann MacDonagh:


 I'm confused, what am I doing wrong here?


 I think that Phobos needs way more&better unittests.

 I think the error you receive reduces to this:

 import std.range: isForwardRange;

 void foo(R)(R r) if (isForwardRange!R) {}

 void main() {
       immutable arr = [1, 2];
       foo(arr);
 }

 Bye,
 bearophile


Strangely enough there is a unit test for this...
https://github.com/D-Programming-Language/phobos/blob/phobos-2.053/std/algorithm.d#L6742

Does this code fail to compile for you too? I want to make sure I didn't 
mess up my config here.
Jul 02 2011
parent reply Johann MacDonagh <johann.macdonagh..no spam..gmail.com> writes:
On 7/2/2011 9:00 PM, Johann MacDonagh wrote:

 On 7/2/2011 9:00 PM, bearophile wrote:

 Johann MacDonagh:


 I'm confused, what am I doing wrong here?


 I think that Phobos needs way more&better unittests.

 I think the error you receive reduces to this:

 import std.range: isForwardRange;

 void foo(R)(R r) if (isForwardRange!R) {}

 void main() {
 immutable arr = [1, 2];
 foo(arr);
 }

 Bye,
 bearophile


 Strangely enough there is a unit test for this...
 https://github.com/D-Programming-Language/phobos/blob/phobos-2.053/std/algorithm.d#L6742


 Does this code fail to compile for you too? I want to make sure I didn't
 mess up my config here.


Ah, here we go:

http://d.puremagic.com/issues/show_bug.cgi?id=6148

Replacing immutable with auto works. I'm still confused how the phobos 
unit tests pass though...
Jul 02 2011
parent reply Jonathan M Davis <jmdavisProg gmx.com> writes:
On 2011-07-02 18:14, Johann MacDonagh wrote:

 On 7/2/2011 9:00 PM, Johann MacDonagh wrote:

 On 7/2/2011 9:00 PM, bearophile wrote:

 Johann MacDonagh:

 I'm confused, what am I doing wrong here?


 
 I think that Phobos needs way more&better unittests.
 
 I think the error you receive reduces to this:
 
 import std.range: isForwardRange;
 
 void foo(R)(R r) if (isForwardRange!R) {}
 
 void main() {
 immutable arr = [1, 2];
 foo(arr);
 }
 
 Bye,
 bearophile


 
 Strangely enough there is a unit test for this...
 https://github.com/D-Programming-Language/phobos/blob/phobos-2.053/std/al
 gorithm.d#L6742
 
 
 Does this code fail to compile for you too? I want to make sure I didn't
 mess up my config here.


 
 Ah, here we go:
 
 http://d.puremagic.com/issues/show_bug.cgi?id=6148
 
 Replacing immutable with auto works. I'm still confused how the phobos
 unit tests pass though...


That's because they're using immutable(int)[], not immutable int[]. The 
elements in the range are immutable, but the range is not. Whereas with 
immutable int[], the whole thing is immutable, so it doesn't work. The range 
has to be mutable.

- Jonathan M Davis
Jul 02 2011
next sibling parent reply Adam D. Ruppe <destructionator gmail.com> writes:
Jonathan M Davis wrote:

 The range has to be mutable.


Is there any easy way around this, aside from casting away
the outer immutable?

It's a very annoying limitation that doesn't really have to be
there - I believe the language itself would let you pass immutable
int[] to a regular function that expects immutable(int)[], since the
outermost reference is passed by value anyway.
Jul 02 2011
next sibling parent reply Johann MacDonagh <johann.macdonagh..no spam..gmail.com> writes:
On 7/2/2011 10:42 PM, Adam D. Ruppe wrote:

 Jonathan M Davis wrote:

 The range has to be mutable.


 Is there any easy way around this, aside from casting away
 the outer immutable?

 It's a very annoying limitation that doesn't really have to be
 there - I believe the language itself would let you pass immutable
 int[] to a regular function that expects immutable(int)[], since the
 outermost reference is passed by value anyway.


Just the pointer is passed by value, referencing elements in the array 
mutate the source.
Jul 02 2011
parent Adam D. Ruppe <destructionator gmail.com> writes:
Johann MacDonagh wrote:

 Just the pointer is passed by value, referencing elements in the
 array mutate the source.


That's true, but when just moving through a range, the pointer
is the only thing that changes. The stuff it points to can still
be immutable.
Jul 02 2011
prev sibling next sibling parent Johann MacDonagh <johann.macdonagh..no spam..gmail.com> writes:
On 7/2/2011 10:42 PM, Adam D. Ruppe wrote:

 Jonathan M Davis wrote:

 The range has to be mutable.


 Is there any easy way around this, aside from casting away
 the outer immutable?

 It's a very annoying limitation that doesn't really have to be
 there - I believe the language itself would let you pass immutable
 int[] to a regular function that expects immutable(int)[], since the
 outermost reference is passed by value anyway.


Oh, I misread your comment. I agree, but I bet you run into issues when 
it comes to appending data to your newly casted immutable(int)[] 
reference (any appending would require a relocation). I should probably 
re-read the D-slices article.
Jul 02 2011
prev sibling parent Jonathan M Davis <jmdavisProg gmx.com> writes:
On 2011-07-02 19:42, Adam D. Ruppe wrote:

 Jonathan M Davis wrote:

 The range has to be mutable.


 
 Is there any easy way around this, aside from casting away
 the outer immutable?
 
 It's a very annoying limitation that doesn't really have to be
 there - I believe the language itself would let you pass immutable
 int[] to a regular function that expects immutable(int)[], since the
 outermost reference is passed by value anyway.


The problem is how the type is determined with templates. The _exact_ type is 
used. So, instead of generating a function which takes immutable(int)[], the 
compiler generates a function which takes an immutable int[], because you 
passed it an immutable int[], not an immutable(int)[]. And it doesn't work to 
iterate over an immutable int[], so the function fails to compile. It's the 
same reason why you can't use static arrays with range-based functions unless 
you slice them. They don't work when the type is a static array, since you 
can't pop elements off of a static array.

If the compiler were smart enough to use immutable(int)[] instead of immutable 
int[], and to use a dynamic array instead of a static one when choosing the 
type for the template, then the problem would be solved. But it isn't that 
smart. There are at least a couple of enhancement requests which relate to the 
issue, and it may be resolved at some point, but as long as the compiler uses 
the _exact_ type given for the template even when a more relaxed type would 
work, you're not going to be able to pass immutable arrays or static arrays to 
templated functions.

- Jonathan M Davis
Jul 03 2011
prev sibling parent Johann MacDonagh <johann.macdonagh..no spam..gmail.com> writes:
On 7/2/2011 10:22 PM, Jonathan M Davis wrote:

 On 2011-07-02 18:14, Johann MacDonagh wrote:

 On 7/2/2011 9:00 PM, Johann MacDonagh wrote:

 On 7/2/2011 9:00 PM, bearophile wrote:

 Johann MacDonagh:

 I'm confused, what am I doing wrong here?


 I think that Phobos needs way more&better unittests.

 I think the error you receive reduces to this:

 import std.range: isForwardRange;

 void foo(R)(R r) if (isForwardRange!R) {}

 void main() {
 immutable arr = [1, 2];
 foo(arr);
 }

 Bye,
 bearophile


 Strangely enough there is a unit test for this...
 https://github.com/D-Programming-Language/phobos/blob/phobos-2.053/std/al
 gorithm.d#L6742


 Does this code fail to compile for you too? I want to make sure I didn't
 mess up my config here.


 Ah, here we go:

 http://d.puremagic.com/issues/show_bug.cgi?id=6148

 Replacing immutable with auto works. I'm still confused how the phobos
 unit tests pass though...


 That's because they're using immutable(int)[], not immutable int[]. The
 elements in the range are immutable, but the range is not. Whereas with
 immutable int[], the whole thing is immutable, so it doesn't work. The range
 has to be mutable.

 - Jonathan M Davis


Ah! How could I miss that? Some of the examples in std.algorithm should 
be updated to reflect this at least until that bug is fixed.
Jul 02 2011