Kind of meta: get the longest of the shortest answers

Perl, 195 bytes

while(<>){/(\S+)\t(\d+)\t(.+)/;push@{$a{$1}},$3if$2==$l{$1};$l{$1}=$2,$a{$1}=[$3]if $2<($l{$1}//65536)}$m=(sort{$b<=>$a}values%l)[0];map{$l=$_;map{print"$l\t$m\t$_\n"if$l{$l}==$m}@{$a{$l}}}keys%l

Input is expected in STDIN, result is written to STDOUT:

C#      208     http://short.url/answer/poi
C#      208     http://short.url/answer/yac
Java    208     http://short.url/answer/jav

Ungolfed version

#!/usr/bin/env perl
use strict;
$^W=1;

# hash %language remembers the minimum count for a language
# %language: <language> => <minimum count>
my %language;
# hash %array remembers the URLs for the minimum count of the language
# %array: <language> => [<url>, <url>, ....]
my %array;

while(<>){
    # parse input line (no error checking)
    /(\S+)\t(\d+)\t(.+)/;
    my ($lang, $count, $url) = ($1, $2, $3);
    # add URL, if the count is the current minimum for the language
    if ($count == ($language{$lang}//0)) {
    # better, but longer version:
    # if (defined $language{$lang} and $count == $language{$lang}) {
        push @{$array{$lang}}, $url;
    }
    # create a new entry for the language, if there is a new minimum
    if ($count < ($language{$lang}//65536)) {
    # better, but longer version:
    # if (not defined $language{$lang} or $count < $language{$lang}) {
        $language{$lang} = $count;
        $array{$lang} = [$url];   
    }
}

# Sort the minimal values in numerical descending order and
# get the first entry as maximum.
my $maximum = (sort { $b <=> $a } values %language)[0];

# Loop over all URLs of minimal answers for the language,
# but print only the entries for the languages with the largest
# minima.
foreach my $lang (keys %language) {
    foreach my $url (@{$array{$lang}}) {
        if ($language{$lang} == $maximum) {
            print "$lang\t$maximum\t$url\n";
        }
    }
}
__END__

Java - 556

import java.util.*;class G{public static void main(String[]x){TreeMap<?,TreeMap>m=new TreeMap();try{Scanner s=new Scanner(System.in);for(;;){String[]a=s.nextLine().split("\t");a(new Long(a[1]),a(a[0],m)).put(a[2],a);}}catch(Exception e){}TreeMap<?,Map<?,String[]>>n=new TreeMap();for(TreeMap o:m.values())a(o.firstEntry().getKey(),n).putAll((Map)o.firstEntry().getValue());for(String[]o:n.lastEntry().getValue().values())System.out.println(o[0]+"\t"+o[1]+"\t"+o[2]);}static<T>Map a(T t,Map m){if(m.get(t)==null)m.put(t,new TreeMap());return(Map)m.get(t);}}

Program will read from STDIN.

import java.util.Map;
import java.util.Scanner;
import java.util.TreeMap;

class G {
    public static void main(String[] x) {
        TreeMap<?, TreeMap> m = new TreeMap();
        try {
            Scanner s = new Scanner(System.in);
            for (; ; ) {
                String[] a = s.nextLine().split("\t");
                a(new Long(a[1]), a(a[0], m)).put(a[2], a);
            }
        } catch (Exception e) {
        }
        TreeMap<?, Map<?, String[]>> n = new TreeMap();
        for (TreeMap o : m.values())
            a(o.firstEntry().getKey(), n).putAll((Map) o.firstEntry().getValue());
        for (String[] o : n.lastEntry().getValue().values())
            System.out.println(o[0] + "\t" + o[1] + "\t" + o[2]);
    }

    static <T> Map a(T t, Map m) {
        if (m.get(t) == null)
            m.put(t, new TreeMap());
        return (Map) m.get(t);
    }
}
  1. Programm will read line by line until an exception occours (either ArrayIndexOutOfBoundsException when a blank line is encountered or NoSuchElementException if input ends without trailing new line). Each line read is added to the TreeMap m, which could have been defined as TreeMap<String, TreeMap<Long, TreeMap<String,String[]>>> (left-to-right: language, code-size, URL, input).
  2. Then a result-TreeSet<Long, TreeSet<String, String[]>> n (left-to-right: code-size, URL, input) is build where the contents of every languages firstEntry() are aggregated.
  3. lastEntry() of the aggregated TreeMap contains our result - we only need to print it.

Try on ideone.com (switched last two lines of input to show that all lines are read)


Python 378 377 372

import sys
d=__import__("collections").defaultdict(list)
o={}
x=int
n="\n"
for i,l,u in[a.split()for a in sys.stdin.read().strip().split(n)]:d[i]+=[(l,u)]
for e,b in d.items():o[e]=[i for i in b if i[0]==str(min([x(i[0])for i in b]))]
print("".join(n.join("\t".join([u,s[0],s[1]])for s in y if x(s[0])==max(x(i[0][0])for i in o.values()))+n for u,y in o.items()).strip())

Input on the stdin:

C:\Users\gcq\Documents\python>type m.txt | python test.py
C#      208     http://short.url/answer/poi
C#      208     http://short.url/answer/yac
Java    208     http://short.url/answer/jav

And this is what i had before starting to compress it down, at 551 chars:

from collections import defaultdict
import sys
d = defaultdict(list)

for language, length, url in [a.split() for a in sys.stdin.read().strip().split("\n")]:
    d[language].append((length, url))

o = {}
for language, data in d.items():
    winval = data[0][0]
    for i in data:
        if int(i[0]) < int(winval):
            winval = i[0]
    o[language] = [i for i in data if i[0] == winval]

maxlen = max(int(i[0][0]) for i in o.values())

for language, dataa in o.items():
    for data in dataa:
        if int(data[0]) == maxlen:
            print("\t".join([language, data[0], data[1]]))