Code golf: find all anagrams

Question

A word is an anagram if the letters in that word can be re-arranged to form a different word.

Task:

• The shortest source code by character count to find

Answer 1

Powershell, 104 97 91 86 83 chars

$k=@{};$input|%{$k["$([char[]]$_|%{$_+0}|sort)"]+=@($_)}
$k.Values|?{$_[1]}|%{"$_"}

Update for the new requirement (+8 chars):

To exclude the words that only differ in capitalization, we could just remove the duplicates (case-insensitvely) from the input list, i.e. $input|sort -u where -u stands for -unique. sort is case-insenstive by default:

$k=@{};$input|sort -u|%{$k["$([char[]]$_|%{$_+0}|sort)"]+=@($_)} 
$k.Values|?{$_[1]}|%{"$_"} 

Explanation of the [char[]]$_|%{$_+0}|sort -part

It's a key for the hashtable entry under which anagrams of a word are stored. My initial solution was: $_.ToLower().ToCharArray()|sort. Then I discovered I didn't need ToLower() for the key, as hashtable lookups are case-insensitive.

[char[]]$_|sort would be ideal, but sorting of the chars for the key needs to be case-insensitive (otherwise Cab and abc would be stored under different keys). Unfortunately, sort is not case-insenstive for chars (only for strings).

What we need is [string[]][char[]]$_|sort, but I found a shorter way of converting each char to string, which is to concat something else to it, in this case an integer 0, hence [char[]]$_|%{$_+0}|sort. This doesn't affect the sorting order, and the actual key ends up being something like: d0 o0 r0 w0. It's not pretty, but it does the job :)

Answer 2

Haskell, 147 chars

prior sizes: 150 159 chars

import Char
import List
x=sort.map toLower
g&a=g(x a).x
main=interact$unlines.map unwords.filter((>1).length).groupBy((==)&).sortBy(compare&).lines

---

This version, at 165 chars satisifies the new, clarified rules:

import Char
import List
y=map toLower
x=sort.y
g&f=(.f).g.f
w[_]="";w a=show a++"\n"
main=interact$concatMap(w.nubBy((==)&y)).groupBy((==)&x).sortBy(compare&x).lines

This version handles:

1. Words in the input that differ only by case should only count as one word
2. The output needs to be one anagram set per line, but extra punctuation is acceptable