Using std.algorithm how to uniq and sort a multidimensional array?

e.g. the uniq function takes a function as a predicate but i'm confused how to handle the multiple dimensions.

    string[][] tags;

    tags = tags.uniq!("a[0][1] == b[1][1]").array;

The above is my attempt and it failed.

On Tuesday, 28 January 2014 at 14:46:48 UTC, Gary Willoughby wrote:
> Using std.algorithm how to uniq and sort a multidimensional array?
>
> e.g. the uniq function takes a function as a predicate but i'm confused how to handle the multiple dimensions.
>
>     string[][] tags;
>
>     tags = tags.uniq!("a[0][1] == b[1][1]").array;
>
> The above is my attempt and it failed.

Gah, it was simpler than i thought, of course.

	tags = tags.sort!("a[1] < b[1]").uniq!("a[1] == b[1]").array;

Gary Willoughby:

>     string[][] tags;
>
>     tags = tags.uniq!("a[0][1] == b[1][1]").array;

I think the right abstraction for your use case is:

auto tags = tags.flatten.uniq.array;

Where a std.range.flatten should take a range, range of ranges, range of range of ranges, etc, and yield the items scanning them by rows (flatten takes a template argument enumeration to change the scanning pattern: by rows, by columns, bidirectional rows, zig-zag, Hamming curve, Z curve, 2D rectangular blocks).

Probably flatten should have a compile-time argument to tell it what a iterable range means. And an optional run-time argument "maxDepth".

Ideally flatten should yield references of the items, so you can increase all matrix items by 1 with code like:

int[][][] matrix = ...;
foreach (ref item; mat.flatten)
    item++;

Bye,
bearophile

On Tuesday, 28 January 2014 at 15:07:26 UTC, bearophile wrote:
> Gary Willoughby:
>
>>    string[][] tags;
>>
>>    tags = tags.uniq!("a[0][1] == b[1][1]").array;
>
> I think the right abstraction for your use case is:
>
> auto tags = tags.flatten.uniq.array;

auto tags = tags.flatMap!uniq.array ?

Meta:

> auto tags = tags.flatMap!uniq.array ?

That's another solution. In Phobos a flatMap can be useful (just as a zip overload that accepts a function to apply on the pairs, named zipWith in haskell: http://zvon.org/other/haskell/Outputprelude/zipWith_f.html ).

Bye,
bearophile