The following works great. It sorts an AA by value:

1. by frequency of word
2. by alphabetic order (if two or more words have the same value)

import std.stdio : writefln;
import std.algorithm.sorting : multiSort;

void main() {
  size_t[string] wcount = [
    "hamster":5,
    "zorro":80,
    "troll":90,
    "algorithm":80,
    "beer":80
  ];
  struct Word { string word; size_t count; }
  Word[] sorter;
  foreach (ref k, ref v; wcount)
    sorter ~= Word(k, v);
  assert(wcount.length == sorter.length);
  sorter.multiSort!("a.count > b.count", "a.word < b.word");
  assert(sorter[2].word == "beer");
  foreach (ref it; sorter)
    writefln("%s : %d", it.word, it.count);
}

I'm happy with it, but maybe there is a more concise implementation?

Chris:

> I'm happy with it, but maybe there is a more concise implementation?

This is a bit shorter and a bit better (writefln is not yet able to format tuples nicely):


void main() {
    import std.stdio: writeln;
    import std.algorithm.sorting: multiSort;
    import std.array: array;

    const size_t[string] wCount = [
        "hamster": 5,
        "zorro": 80,
        "troll": 90,
        "algorithm": 80,
        "beer": 80
    ];

    auto pairs = wCount.byKeyValue.array;
    assert(wCount.length == pairs.length);
    pairs.multiSort!(q{a.value > b.value}, q{a.key < b.key});
    assert(pairs[2].key == "beer");
    foreach (const ref it; pairs)
        writeln(it.key, ": ", it.value);
}


Bye,
bearophile

On Tuesday, 21 April 2015 at 11:46:24 UTC, bearophile wrote:
> Chris:
>
>> I'm happy with it, but maybe there is a more concise implementation?
>
> This is a bit shorter and a bit better (writefln is not yet able to format tuples nicely):
>
>
> void main() {
>     import std.stdio: writeln;
>     import std.algorithm.sorting: multiSort;
>     import std.array: array;
>
>     const size_t[string] wCount = [
>         "hamster": 5,
>         "zorro": 80,
>         "troll": 90,
>         "algorithm": 80,
>         "beer": 80
>     ];
>
>     auto pairs = wCount.byKeyValue.array;
>     assert(wCount.length == pairs.length);
>     pairs.multiSort!(q{a.value > b.value}, q{a.key < b.key});
>     assert(pairs[2].key == "beer");
>     foreach (const ref it; pairs)
>         writeln(it.key, ": ", it.value);
> }
>
>
> Bye,
> bearophile

Nice!

Maybe something like bearophile's example should go into the documentation of std.algorithm.sorting.multiSort. It's a common enough thing in programming.