Hey folks!

Please enlight me with that prefix notation of 2-dimensional arrays! I prepared a snippet giving me headaches:

auto int[2][3] test = [[11,12],[21,22],[31,32]];
foreach (x, row; test)
{
	Stdout.format("x={}: ", x+1);

	foreach (y, cell; row)
	{
		Stdout.format("{}:({}) ", y+1, test[x][y]);
//		Stdout.format("{}:({}) ", y+1, test[y][x]); // I expected this here
according the declariation!
	}
	Stdout.newline;
}

According to the documentation, the declaration of test should declare 3 arrays of two ints. The initialization works fine so that's ok for me.

But why do I have to access it with test[x][y] to produce this result?

x=1: y=1:(11) y=2:(12)
x=2: y=1:(21) y=2:(22)
x=3: y=1:(31) y=2:(32)

This literally drives me crazy! I really want to understand that!

Your help is highly appreciated!

Thanks a lot!

Andreas

On Mon, 1 Nov 2010 07:50:12 +0000 (UTC)
Andreas Kaempf <andreas.kaempf@web.de> wrote:

> According to the documentation, the declaration of test should declare 3 arrays of two ints. The initialization works fine so that's ok for me.
> 
> But why do I have to access it with test[x][y] to produce this result?

Yes, this is a bit mind disturbing. If you consider each syntax independantly, they are both completely logical:

* type def: since T[2] defines an array of 2 T's, then T[2][3] well defines an array of 3 arrays of 2 T's.

* element access: as your literal [[11,12],[21,22],[31,32]] well shows, the 2-element arrays are the nested ones; so that, to access an element, one must first access a 2-element array -->
	arr[2-element-array-index][element-index]
which maps to
	arr[row-index][element-index]
as one expects.

But the confrontation of both logics is somewhat troubling ;-) To solve it, one would need to reverse the
array definition format: [n]T means array of n T's, [m][n]T means array of m arrays of n T's.
Simply forget about it and use each syntax according to its own point if view.


Denis
-- -- -- -- -- -- --
vit esse estrany ☣

spir.wikidot.com

On Mon, 01 Nov 2010 03:50:12 -0400, Andreas Kaempf <andreas.kaempf@web.de> wrote:

> Hey folks!
>
> Please enlight me with that prefix notation of 2-dimensional arrays! I
> prepared a snippet giving me headaches:
>
> auto int[2][3] test = [[11,12],[21,22],[31,32]];
> foreach (x, row; test)
> {
> 	Stdout.format("x={}: ", x+1);
>
> 	foreach (y, cell; row)
> 	{
> 		Stdout.format("{}:({}) ", y+1, test[x][y]);
> //		Stdout.format("{}:({}) ", y+1, test[y][x]); // I expected this here
> according the declariation!
> 	}
> 	Stdout.newline;
> }
>
> According to the documentation, the declaration of test should declare 3
> arrays of two ints. The initialization works fine so that's ok for me.
>
> But why do I have to access it with test[x][y] to produce this result?
>
> x=1: y=1:(11) y=2:(12)
> x=2: y=1:(21) y=2:(22)
> x=3: y=1:(31) y=2:(32)
>
> This literally drives me crazy! I really want to understand that!

test is a 3-element array of 2-element arrays.  When you index test, you get a 2-element array as its element.  When you index that, you get the individual ints.

Think about it in terms of a single array.  If you have an array of type T[3], it's element type is T.  So if T is int[2], then it appears perfectly consistent:

int[2][3] t; // T is int[2]

auto x = t[0]; // T x is of type int[2]

As an aside, your foreach code is mighty inefficient.  Each time through the array, row is a copy of the original element, which means you are moving a lot of data without ever using it.

I'd recommend using ref, like this:

foreach(x, ref row; test)

Unfortunately, you can't use the 0..test.length syntax since you are using D1.  You could use a straight for-loop.

-Steve