ETAOIN SHRDLU golf

Question

Short and sweet description of the challenge:
Based off ETAOIN SHRDLU, your challenge is to write the shortest program or function in any language that outputs the 26 letters of the English alphabet based on their frequency in the input.

Really long, dry and thorough specification:

• Your program/function will receive as input a string of text, which will contain one or more uppercase and/or lowercase letters and may also contain punctuation, numerals, symbols, and other non-alphabetic characters.
• The program/function must output only the 26 UPPERCASE letters of the English alphabet, including those that do not appear in the input, ordered from most to least frequent based on how many times they appear in the input.
• Edit: The frequency is calculated case-insensitively, but the output must be in uppercase.
• If two or more letters have the same frequency, they may be in any order.
• No other output, such as whitespace, is allowed.
• Edit 7/1/2014: Based on feedback, I am amending this rule. The only other output that is allowed is optional leading and/or trailing whitespace, such as a trailing newline. No other output is allowed.
• Undefined behavior is allowed for input that does not contain any letters.

The winner will be picked 7 days from now, so get those fingers typing!

---

Example input:

Lorem ipsum dolor sit amet, consectetur adipiscing elit. Praesent vitae erat velit. Mauris gravida euismod libero ut tincidunt. Phasellus elit dui, consectetur et egestas in, aliquam vitae diam. Donec eget varius ante. Vestibulum cursus diam aliquet, egestas orci quis, placerat dolor. Proin vel nisi lectus. Class aptent taciti sociosqu ad litora torquent per conubia nostra, per inceptos himenaeos. Aliquam erat volutpat. Etiam libero tortor, ornare id dui eget, posuere dignissim libero. Pellentesque commodo consequat volutpat. Integer hendrerit sapien libero, vel viverra augue facilisis sit amet. Quisque consectetur eget nisl quis dignissim. Ut lacinia pretium quam a placerat.
Morbi sed interdum risus, nec pretium lectus. Morbi imperdiet est id accumsan molestie. Duis sed fermentum nisl. Nunc vitae augue mattis, dictum lectus vel, accumsan nisl. Sed ultricies adipiscing rhoncus. Vivamus eu lacus a enim venenatis eleifend. Praesent consectetur tortor non eleifend ultricies. Mauris et odio posuere, auctor erat at, fringilla est. Proin in vestibulum erat. Maecenas congue commodo ante vel varius. Sed tempus mi ut metus gravida, nec dictum libero dapibus. Morbi quis viverra elit. Ut pharetra neque eget lacus tincidunt dictum. Fusce scelerisque viverra tellus et pretium.
Fusce varius adipiscing odio. Nulla imperdiet faucibus sem, at rhoncus ipsum adipiscing vitae. Phasellus imperdiet congue lacus et mollis. Nullam egestas mauris magna, et mollis lectus varius ut. Sed sollicitudin adipiscing dolor, vel elementum elit laoreet molestie. Aliquam nec nulla vel sem ultrices ullamcorper. Nullam nec felis magna. Duis sodales orci non justo aliquam tempus. Integer mi diam, tempor sed vulputate et, varius et nunc. Vestibulum sodales ipsum id mi pharetra, ut convallis mi accumsan. Sed dictum volutpat vestibulum.
Quisque ac dolor sagittis, aliquam libero at, euismod enim. Nulla ullamcorper posuere nulla vitae varius. Nam at dolor non libero elementum pellentesque in in lorem. Fusce porttitor turpis in quam placerat varius. Donec lorem orci, condimentum eu sapien sit amet, aliquet commodo magna. Quisque sed lectus sit amet arcu euismod accumsan et non nunc. Phasellus placerat congue metus, feugiat posuere leo dictum quis. Sed ultricies feugiat eros dignissim bibendum.
Mauris scelerisque consectetur libero eget varius. Aenean neque nunc, ullamcorper vitae orci in, auctor ornare sapien. Nam lacinia molestie imperdiet. Nam vitae mattis nibh. Vestibulum consequat tellus ac nisi sagittis pulvinar. Nullam mollis ornare quam, et venenatis leo porttitor sit amet. Nulla urna neque, dignissim non orci ut, volutpat ultrices erat. Cum sociis natoque penatibus et magnis dis parturient montes, nascetur ridiculus mus. Pellentesque vestibulum tellus nec eros faucibus porta.

Example output:

EITUSALNROMCDPVGQBFHJKWXYZ

Note: there is a 5 way tie between KWXYZ for that input.

Edit:

The competition is over! Thanks to everyone who participated. And now for the winner(s!): Both Dennis' CJam and isaacg's Pyth answers came in at a whopping 19 characters. (Sorry, but I'm not going to accept either of the answers because I think it would be unfair to the other.) Edit: Taking Dennis' advice, I'm going to mark his answer as accepted because his was the first to reach 19 characters. Honorable mentions go to Ilmari Karonen's third-place Golfscript answer at 22 chars as well as undergroundmonorail's 75-char Python answer that got the most upvotes. Once again, thanks to everyone who participated!

Answer 1

CJam, 21 19 bytes

qeu:A;'[,65>{A\-,}$

Try it online.

Example

$ cjam etaoin.cjam <<< "~XyxY YyxZ"
YXZABCDEFGHIJKLMNOPQRSTUVW

(no newline)

How it works

qeu:A; " Read from STDIN, convert to uppercase, save in the variable “A” and discard, ";
'[,    " Push an array of all ASCII characters before “[” (NUL to “Z”).               ";
65>    " Remove the first 64 characters (NUL to “@”).                                 ";
{      " Sort the array of characters by the following mapping:                       ";
  A\   " Swap the character with the string saved in variable “A”.                    ";
  -    " Remove all occurrences of the character from the string.                     ";
  ,    " Push the length of the string.                                               ";
}$     "                                                                              ";

More occurrences means more characters get removed, so the most frequent characters appear at the beginning of the array.

Answer 2

Python 2 or 3 - 77 75 bytes

f=lambda s:''.join(sorted(map(chr,range(65,91)),key=s.upper().count))[::-1]

I had an answer before that grabbed input from STDIN, but I realized it was technically invalid. I used input() which gets only a single line, but the question's example input implies that it should handle multiple lines at once. To meet spec, I turned my answer into a function that takes a string argument. To my surprise, it was two bytes smaller! It didn't occur to me that print(...) and input() were longer than f=lambda s: and s.

This also makes the answer compatible with both Python 2 and Python 3. Originally it was only Python 3, because it used input() (which was called raw_input() in 2). Now that it's a function, it works in both.

Explained

                                  range(65,91)                              # The numbers 65 to 90
                          map(chr,range(65,91))                             # Convert to ASCII

                                                    s                       # The input string
                                                    s.upper()               # Convert to uppercase
                                                    s.upper().count         # Function literal for 'how many times the argument appears in the string'

                   sorted(map(chr,range(65,91)),key=s.upper().count)        # Sort by that function
           ''.join(sorted(map(chr,range(65,91)),key=s.upper().count))       # Concatenate to string
           ''.join(sorted(map(chr,range(65,91)),key=s.upper().count))[::-1] # Step through by -1 (i.e. reverse string)

  lambda s:''.join(sorted(map(chr,range(65,91)),key=s.upper().count))[::-1] # Make it a function (`return` is implicit for lambdas)
f=lambda s:''.join(sorted(map(chr,range(65,91)),key=s.upper().count))[::-1] # Give it a name

Answer 3

Bash, 65 bytes

(tr a-z A-Z;echo {A..Z})|fold -1|sort|uniq -c|sort -nr|tr -dc A-Z

Example

$ bash etaoin.sh <<< "~AbaB BbaC"
BACZYXWVUTSRQPONMLKJIHGFED

How it works

(              #
  tr a-z A-Z   # Turn lowercase into uppercase letters.
  echo {A..Z}  # Print all uppercase letters.
) |            #
fold -1 |      # Split into lines of length 1.
sort |         # Sort those lines (required for piping to uniq).
uniq -c |      # Print the frequencies of all lines.
sort -nr |     # Sort by frequency (reversed).
tr -dc A-Z     # Remove everything that's not an uppercase letter.

Answer 4

Pyth 1.0.2, 19 20

=ZUwsVm;dSm[cZkk)UG

Try it here: http://ideone.com/fork/YlWpEJ

Learn more about Pyth here: http://esolangs.org/wiki/Pyth

Example:

Based off ETAOIN SHRDLU, your challenge is to write the shortest program or function in any language that outputs the 26 letters of the English alphabet based on their frequency in the input.

Gives:

TENOHARSIULGFPYDCBWQMZXVKJ

Explanation:

=ZUw: Convert input to uppercase and store in Z.

sV: Print the sum of the reverse of

m;d: The last entries of

S: Sorted by their first entry, in increasing order

m[cZkk): The lists [count of k in Z, k]

UG: For k in the uppercase letters.

Rough Python equivalent:

G='abcdefghijklmnopqrstuvwxyz'
Z=copy(upper(input()))
print(_sum(rev(_map(lambda d:d.pop(),sorted(_map(lambda k:_list(count(Z,k),k),upper(G)))))))

---

This is not entry, I just thought people might like to see it. In Pyth 1.0.4, the following program is a solution in 10 characters:

JUwo_cJNUG

Explanation:

JUw: Convert input to uppercase and store in J.

o: (Print) Sort by

_cJN: -1*(count of N in J)

UG: Over N in the uppercase letters.

It is not a legal solution because several of the changes from Pyth 1.0.2 to 1.0.4, including the addition of the o, sort by, function, were in response to this problem.

Answer 5

Javascript (ES6) 119 117

Edit: (-2) Removed the need for toUpperCase using a case-insensitive RegEx in the split call.

a=prompt(f=x=>a.split(RegExp(x,'i')).length)
alert([...'ABCDEFGHIJKLMNOPQRSTUVWXYZ'].sort((b,c)=>f(c)-f(b)).join(''))

Alternative (same length): Condensed the sort and character counting into a single function.

a=prompt()
alert([...'ABCDEFGHIJKLMNOPQRSTUVWXYZ'].sort(f=(b,c)=>c?f(c)-f(b):a.split(RegExp(b,'i')).length).join(''))

As a function: 105 104

Edit: (-1) Condensed the sort and character counting into a single function.

F=a=>[...'ABCDEFGHIJKLMNOPQRSTUVWXYZ'].sort(f=(b,c)=>c?f(c)-f(b):a.split(RegExp(b,'i')).length).join('')

Answer 6

GolfScript, 22 chars

:?91,+-26>{.32+]?\-,}$

Try it online.

Explanation:

• :? assigns the input string to the symbol ?. (I use a punctuation symbol so that the following number 91 is not parsed as part of the symbol name.)
• 91, constructs a list of the numbers from 0 to 90 (ASCII code of Z).
• + appends this list to the input string, thereby converting it from an array of ASCII codes to a string (and conveniently also removing the input string from the stack).
• -26> takes the last 26 characters of this string, producing a string containing the uppercase ASCII letters from A to Z.
• The { }$ applies the code block to all the characters in the input string, and sorts those characters according to the result.
• Inside the code block, . duplicates the character and 32+ converts the copy from uppercase to lowercase. The ] collects these two characters into an array, ?\- takes the input string stored in ? and removes all occurrences of the characters in the array from it, and , counts the length of the remaining string, which will be the sort key. The characters will be sorted in ascending order by this key, and thus in descending order by number of occurrences.

Answer 7

Haskell, 110 bytes

import Data.List
import Data.Char
a%f=compare(f a).f
f t=sortBy(% \c->length$filter((/=c).toUpper)t)['A'..'Z']

Example usage:

λ> f "Based off ETAOIN SHRDLU, your challenge is to write the shortest program or function in any language that outputs the 26 letters of the English alphabet based on their frequency in the input."
"ETNAHORISULFGPBCDYMQWJKVXZ"

Answer 8

Ruby 2.0, 53 characters

EDIT: Fixed to work correctly with multiline strings, thanks @durron597!

f=->s{$><<(?A..?Z).sort_by{|c|-s.upcase.count(c)}*''}

Creates a function called f which can be used as follows:

f['jackdaws love my big sphinx of quartzzz']

Prints to STDOUT:

AZOSICGHEJKLMBFPQRDTUVWXYN

Answer 9

Perl, 54 46 bytes

UPDATE: after further optimizations it could be squeezed to 46 bytes: (thx dennis for -n/{} hack; chinese perl goth for <=>->- hack)

s/./$h{uc$&}++/eg}{say sort{$h{$b}-$h{$a}}A..Z

It is to be run with run with perl -nE

Original solution (does not need special Perl options):

s/./$h{uc$&}++/egfor<>;print sort{$h{$b}<=>$h{$a}}A..Z

Verified in Perl 5.8.3, 5.14.2

If you get a warning, separate eg and for with a space (+1 char), if you mind

example usage:

$ python -c 'import this' | perl -le 's/./$h{uc$&}++/egfor<>;print sort{$h{$b}<=>$h{$a}}A..Z' 2>/dev/null
ETAISONLRHPBUCDYMFGXVWKZJQ

EXPLANATION: On each character (.) of each input line (for<>), apply a substitution "pattern", which in fact is evaluated as an expression (e flag of the s///), that increments an upcased (uc) character (./$& is shorter than more obvious (.)/$1) count in the (uninitialized) hash (%h). Then, the letter frequency hash is used in a sort comparison function to print out upper-case alphabet in the right order.

Answer 10

R, 123 bytes

The code is improved thanks to the suggestions by @RichieCotton.

text <- "Based off ETAOIN SHRDLU, your challenge is to write the shortest program or function in any language that outputs the 26 letters of the English alphabet based on their frequency in the input."

f=function(x){b=plyr::count(toupper(strsplit(x,"")[[1]]));c=merge(LETTERS,b,all.x=T);paste(c[order(-c$freq),1],collapse="")}

f(text)

Output:

> f(text)
[1] "ETNAHORISULFGPBCDYMQWJKVXZ"

Answer 11

VBScript 181 109

Updated to use a completely different algorithm. Beats JavaScript!

Pretty:

dim b(99):i=ucase(inputbox(k))
for y=65to 90
    c=chr(y)
    a=len(replace(i,c,k))
    b(a)=c+b(a)
next
msgbox join(b,k)

Golfed:

dim b(99):i=ucase(inputbox(k)):for y=65to 90:c=chr(y):a=len(replace(i,c,k)):b(a)=c+b(a):next:msgbox join(b,k)

Answer 12

J 41 35 Bytes

(u:65+i.26)([\:[#/.~@,e.~#])toupper

Demo:

i=: 'This is a test to see whether this is still working'
(u:65+i.26)([\:[#/.~@,e.~#])toupper i
STIEHLORWAGKNBCDFJMPQUVXYZ

Explanation:

(u:65+i.26) & ( [ \: [ #/.~@,e.~#]) toupper) )
ABCDE...          |    |    |   |      uppercase the right argument
                  |    |    |   \copy from right only member from left
                  |    |     \append the left argument
                  |    \ Afterwards Count apperances of each letter
                  \ Sort the left according to the appearances

The key is to append the left array, such that all letters are available, and already in order. A funny consequence of using a noun as third tine to the fork is that it works as verb as well as as phrase.

Answer 13

SAS - 217 (I think)

Input should be placed on the line after the cards4 statement, or on the appropriate punch cards for your system. I think this approach saves a few characters vs. attempting to quote the input.

data a;
input;
S = upcase(compress(_INFILE_,,'ak'));
do i=1 to length(S);
l=substr(S,i,1);
output;
end;
cards4;
;;;;
run;
proc sql;
select l into :o separated by '' from
(select l, 1/count(l) as f from a group by l) order by f;
quit;

I'm aware that this doesn't meet the full spec, as it only returns characters that appear in the input string. I might need to rethink this a bit.

Answer 14

C++, 185 183 179 177 bytes

Not expected to win, of course (can C++ ever win?) but a fun exercise nonetheless.

#include <algorithm>
#include <stdio.h>
int f[256],p;main(){for(p=65;p<91;p++)f[p]=p;while(~(p=getchar()))f[p&95]+=256;p=256;std::sort(f,f+p);while(p--)f[p]&95&&putchar(f[p]);}

Explanation:

#include <algorithm>         // for std::sort
#include <stdio.h>           // for getchar, putchar
int f[256],p;                // declare an array of count-prefixed chars, and a counter
main(){
    for(p=65;p<91;p++)       // 65 == 'A', 91 == the character after 'Z'
        f[p]=p;              // set the character for the slot
    while(~(p=getchar()))    // read characters until EOF
        f[p&95]+=256;        // increment the packed count for the character stripped of the 'lowercase bit'
    p=256;                   // start a countdown
    std::sort(f,f+p);        // sort the array
    while(p--)               // do the countdown
        f[p]&95 &&           // if the masked-off character is set...
          putchar(f[p]);     // print it
}

Answer 15

Groovy - 130 123 115 112 98 92

As per @cfrick's advice (twice!):

f={('A'..'Z').collectEntries{c->[c,it.grep(~/(?i)$c/).size()]}.sort{-it.value}*.key.join()}

A small test (shamelessly stolen from @jpjacobs):

assert f('This is a test to see whether this is still working') == 
    'STIEHLORWAGKNBCDFJMPQUVXYZ'

And the proposed test is also passing

Answer 16

AppleScript, 278

I noticed that "a" = "A" is true in AppleScript. I can use this in code golf, but the rest of the script is too wordy. I used AppleScript 1.8.3.

This defines a function f. If you add f("a string") at the bottom of the script and run it in the Script Editor, it will show the result.

on c(n)
ASCII character(64+n)
end
on f(s)
set{a,r}to{{},""}
repeat with i from 1 to 26
set j to 0
repeat with b in s
if b&""=c(i)then set j to j+1
end
set a to a&j
end
repeat with j from 0 to(count s)
repeat with i from 1 to 26
if a's item i=j then set r to c(i)&r
end
end
r
end

Formatted and commented:

-- Returns nth letter of alphabet.
on c(n)
    ASCII character (64 + n)
end c

-- Returns letters in s sorted by frequency.
on f(s)
    -- a: list of letter counts
    -- r: resulting string
    set {a, r} to {{}, ""}

    -- For each letter from A to Z,
    -- count letters in string s.
    repeat with i from 1 to 26
        set j to 0
        repeat with b in s
            -- Can't use b = c(i), because
            -- b is a reference to a string
            -- and = never dereferences its
            -- operands. Get contents of b,
            -- here by coercing b to string.
            if b & "" = c(i) then set j to j + 1
        end repeat
        -- Set item i of a to count j.
        set a to a & j
    end repeat

    -- Sort letters by frequency.  Do a counting sort
    -- because AppleScript lacks a sort command.
    repeat with j from 0 to (count s)
        repeat with i from 1 to 26
            if a's item i = j then set r to c(i) & r
        end repeat
    end repeat
    r
end f

-- Example call:
f("Now is the time for all good men to come to the aid of their country.")
-- Result: "OTEIRNMHLFDCAYWUSGZXVQPKJB"

Answer 17

Jelly, 9 bytes (non-competing)

ØAŒuċ¥@ÞU

Explanation

ØAŒuċ¥@ÞU  Main Link
       Þ   Sort
ØA         The uppercase alphabet by
  Œuċ¥@    number of occurrences in the input:
  Œu       Uppercase
    ċ      Count occurrences
     ¥     Grammar: Last two links as a dyad
      @    Swap arguments
        U  Reverse (because sort sorts up)

This reads as "sort the uppercase alphabet by the number of occurrences in the input uppercased, reversed", which is a rather literal translation of the challenge :P

Try it online!

This challenge was linked to Jelly HyperTraining where we solved the challenge. I posted this because I was the first to reach 10 bytes.

-1 byte thanks to Erik the Outgolfer (JHT teacher)

Answer 18

J - 38 35 char

A function taking input on the right as a string. Not a winner, but it was fun to write.

(u:65+i.26)([\:[#/.~@,e.~#])toupper

Explained:

• toupper is a verb in the standard library that upcases a string. That then becomes the right argument of the verb, while the left argument is the alphabet: ASCII codepoints 65 to 90.

• [ and ,e.~#]) selects (#) those letters in the right arg (]) that are elements of (e.~) the left, and then prepends (,) the left arg ([). To wit, we keep only the uppercase chars, and add a single copy of the alphabet to the end, to make sure we catch them all.

• #/.~@ then gives the frequencies of each character. It so happens that this is given in alphabetical order, so right afterwards we can downsort (\:) the alphabet (the left argument [).

A quick lazy example below. Feel free to try it for yourself at tryj.tk.

   (u:65+i.26)([\:[#/.~@,e.~#])toupper 'Based off ETAOIN SHRDLU, your challenge is to write the shortest program or function in any language that outputs the 26 letters of the English alphabet based on their frequency in the input.'
ETNAHORISULFGPBCDYMQWJKVXZ

Answer 19

VBScript 157 156 bytes

Edit: changed msgbox(p) to msgbox p

More readable:

s=ucase(InputBox(z))    'z is empty.
L=len(s)
Dim a(255)
for i=1to L
    x=asc(mid(s,i))
    a(x)=a(x)+1
next
for t=0to L
    For i=65To 90
        If a(i)=t then p=chr(i)&p
    next
next
msgbox p

Golfed: (155 characters + 1 carriage return)

s=ucase(InputBox(z)):L=len(s):Dim a(255):for i=1to L:x=asc(mid(s,i)):a(x)=a(x)+1:next:for t=0to L:For i=65To 90:If a(i)=t then p=chr(i)&p
next:next:msgbox p

I had it at 171 earlier with code that I found more interesting, but comfortablydrei's sorting method is shorter and requires len(s), which makes a "for" shorter than the "while" for the first loop. (yawn)

's=UCase(InputBox(Z))&8 'just need any extra character.  0-7 don't work because &7 is octal

s=UCase(InputBox(Z)) 'nevermind
Dim a(999)
While Len(s)
    x=Asc(s) 'returns ascii of first char
    a(x)=a(x)-1 'going negative saves a character later...
    s=Mid(s,2) 'doesn't care if you run out of string
Wend
for j=1 to 26 'this used to be   While Len(p)<26
    For i=65To 90
        If a(i)<a(y) Then y=i 'it is barely not worth it to do a(i)+a(i+32)>a(y) here to skip the ucase() above
    Next
    p=p&Chr(y)
    a(y)=1 'if I didn't go negative this would have to be -1.  arrays default to 0.
Next
MsgBox(p)

Answer 20

Perl, 78 bytes

undef$/;$i=<>;$r{$i=~s/$_//gi}.=$_ for A..Z;print$r{$_}for sort{$b<=>$a}keys%r

• Only the 26 uppercase ASCII letters without any white space are output in frequency order.
• Tied characters are given in alphabetic order.

Result for the example in the question:

EITUSALNROMCDPVGQBFHJKWXYZ

Ungolfed:

# read input
# ----------
undef $/; # disable input separator
$i = <>;  # $i holds the complete input as one string

# analyze
# -------
# For each uppercase letter (A upto Z) its occurences are counted
# via the number of substitutions made by s/$_//gi. The lowercase
# letter is included via modifier "i".
# 
# The occurrence count is then used as key for hash %r.
# The uppercase letter is appended to the value of that hash entry.
$r{$i =~ s/$_//gi} .= $_ for A..Z;

# output
# ------
# The hash keys are sorted numerically in reverse order by
# the specified sort function.
print $r{$_} for sort {$b<=>$a} keys %r

Answer 21

T-SQL 178

Basically, this is my VBScript solution but implemented in SQL.

This is XML output abuse to concatenate a column. In practical use, it can be joined to an outer table to emulate GROUP_CONCAT functions in MySQL and others.

Declaring the @ variable:

DECLARE @ CHAR(1024)= 'enter your text here';

Code:

with y AS(
    SELECT UPPER(@)i,0l,91y
    UNION ALL
    SELECT i,len(replace(i,char(y-1),'')),y-1
    FROM y
    WHERE y>65
)
SELECT LTRIM(
(
    SELECT char(y)
    FROM y
    WHERE y<91
    ORDER BY l
    FOR XML PATH(''))
)

Answer 22

C# (and LINQ) 255 226 210 Bytes

Using Patrick Huizinga's advice, the query syntax is now shorter:

namespace System.Linq{class P{static void Main(string[]a){Console.Write((from c in(a[0]+"ABCDEFGHIJKLMNOPQRSTUVWXYZ").ToUpper()where c>'@'&&c<'['group c by c into g orderby-g.Count()select g.Key).ToArray());}}}

Explanation:

Console.Write(
    (from c //declare our range variable
       in (a[0] + "ABCDEFGHIJKLMNOPQRSTUVWXYZ").ToUpper() //declare the datasource
     where c > '@' && c < '[' //include only letters
     group c by c into g //run of the mill group by
     orderby -g.Count() //order by descending
     select g.Key //we only want the actual letters
     ).ToArray() //mash it all into an array
  );

Equivalent method syntax (217):

namespace System.Linq{class P{static void Main(string[]a){Console.Write((a[0]+"ABCDEFGHIJKLMNOPQRSTUVWXYZ").ToUpper().GroupBy(c=>c).OrderBy(c=>-c.Count()).Where(c=>c.Key>'@'&&c.Key<'[').Select(c=>c.Key).ToArray());}}}

Original post:

namespace System.Linq{class P{static void Main(string[]a){(a[0]+"ABCDEFGHIJKLMNOPQRSTUVWXYZ").ToUpper().GroupBy(c=>c).OrderByDescending(c=>c.Count()).Where(c=>c.Key>'@'&&c.Key<'[').ToList().ForEach(c=>Console.Write(c.Key));}}}

This is my first ever submission and I should be doing things at work but this just looked like so much fun because I felt like I could actually participate for once.

Explanation:

(a[0] + "ABCDEFGHIJKLMNOPQRSTUVWXYZ") //ensure each character appears once
  .ToUpper()
  .GroupBy(c => c) //get access to .Count()
  .OrderByDescending(c => c.Count())
  .Where(c => c.Key > '@' && c.Key < '[') //exclude anything other than letters
  .ToList() //Only lists have a .ForEach() :(
  .ForEach(c => Console.Write(c.Key)); //print output

I never use the method syntax for LINQ so this was a learning experience for me :) also thinking about it now I could save 2 bytes by replacing the character literals with their integer counterparts, but, meh.

Shortened thanks to tips from ProgramFOX and Num Lock :)

The equivalent query syntax (slightly longer):

(from c in (a[0]+"ABCDEFGHIJKLMNOPQRSTUVWXYZ").ToUpper() where c>'@'&&c<'[' group c by c into g orderby g.Count() descending select g.Key).ToList().ForEach(c=>Console.Write(c));

Answer 23

PHP - 105 bytes

<?preg_filter(~‹§æ“Ö¢‹ö,'$f[$0&fl]++',join('',range(a,z)).$argv[1]);arsort($f);foreach($f as$l=>$F)echo$l;

Here's a hexdump, cause of the special characters:

0000000 3c 3f 70 72 65 67 5f 66 69 6c 74 65 72 28 7e dc
0000010 a4 be d2 85 a2 dc 9a 2c 27 24 66 5b 24 30 26 df
0000020 5d 2b 2b 27 2c 6a 6f 69 6e 28 27 27 2c 72 61 6e
0000030 67 65 28 61 2c 7a 29 29 2e 24 61 72 67 76 5b 31
0000040 5d 29 3b 61 72 73 6f 72 74 28 24 66 29 3b 66 6f
0000050 72 65 61 63 68 28 24 66 20 61 73 24 6c 3d 3e 24
0000060 46 29 65 63 68 6f 24 6c 3b                     
0000069

And a slightly less golfed version:

<?
preg_filter(           // regular expression
  "#[A-z]#e",          // matches every letter + 'eval' flag
  '$f[$0&fl]++',        // so this code runs for every letter
                       // $f is an array whose indices are uppercase letters
                       //   and whose values represent the number of occurences
                       // lowercase is converted to uc with the bitwise and
                       //   fl is 11011111 in binary, every bit except for 32's is set
  join('', range(a,z)) // adding abcdefghijklmnopqrstuvwxyz to the input
    .$argv[1]);        //   because not all letters have to appear in the input
arsort($f);            // sort $f in reverse, maintaining indices
foreach($f as$l=>$F)   //
  echo$l;              // print each index in order

Example:

 $ php etaoin_shrdlu.php "This function sorts an array such that array indices maintain their correlation with the array elements they are associated with."
 ATIRESHNOCYUWMDLFXZBVGPQKJ

Answer 24

C# in LINQPad - 203 Bytes

I took a different approach to the answer of Logan Dam. I made sure first that every character in the input string is sorted by its appearance and only exists once in the output string. After that I added every missing character from the alphabet to the output string.

void e(string i){var a="";foreach(var d in i.ToUpper().GroupBy(x=>x).OrderByDescending(u=>u.Count()))if(d.Key<91&&d.Key>64){a+=d.Key;}for(int x=65;x<91;x++)if(!a.Contains((char)x)){a+=(char)x;}a.Dump();}

Sadly it wouldn't beat Logan Dam's answer if I would have made in Visual Studio.

More readable version:

void e(string i)
    {
        var a = "";
        foreach (var d in i.ToUpper().GroupBy(x => x).OrderByDescending(u => u.Count()))
        {
            if (d.Key < 91 && d.Key > 64)
            {
                a += d.Key;
            }
        }
        for (int x = 65; x < 91; x++)
        {
            if (!a.Contains((char)x))
            {
                a += (char)x;
            }
        }
        a.Dump();
    }

Answer 25

C++ 701 322 232 bytes

First version 701 bytes (idiomatic STL usage)

#define _HAS_TRADITIONAL_STL 1
#include <numeric>
#include <iostream>
#include <iterator>
#include <string>
#include <algorithm>
#include <functional>
#include <map>
#include <set>
#define ALL(x) x.begin(), x.end()
using namespace std;
typedef istream_iterator<char> iic;typedef pair<int, char> pic;map<char, int> c;set<pic> d;
void f1(char x) {c[x]--;}
void f2(const pic &p) {d.insert(make_pair(p.second, p.first));}
int main(){string s(26, 0);stdext::iota(ALL(s), 65);copy(ALL(s), ostream_iterator<char>(cout));transform(iic(cin), iic(), back_inserter(s), toupper);for_each(ALL(s), f1);for_each(ALL(c), f2);transform(ALL(c2), ostream_iterator<char>(cout), select2nd<pic>());}

Expanded clean version :

#define _HAS_TRADITIONAL_STL 1
#include <numeric>
#include <iostream>
#include <iterator>
#include <string>
#include <algorithm>
#include <functional>
#include <map>
#include <set>
using namespace std;

typedef istream_iterator<char> iic;
map<char, int> counts;
set<pair<int, char> > counts2;

void docount(char ch) { counts[ch]--; }
void toCounts2(const pair<char, int> &p) { counts2.insert(make_pair(p.second, p.first)); }

int main()
{
    string s(26, 0);
    stdext::iota(s.begin(), s.end(), 65);
    transform(iic(cin), iic(), back_inserter(s), toupper);
    for_each(s.begin(), s.end(), docount);
    for_each(counts.begin(), counts.end(), toCounts2);
    transform(counts2.begin(), counts2.end(), ostream_iterator<char>(cout), select2nd< pair<int, char> >());
}

The idea is to demonstrate a "proper" C++ program without any hacks. Ignore the boilerplate and the fact that this only compiles on VC++

Explanation :

We fill A to Z into a string with iota(), this ensures that when we count the occurrences, each character appears even if it's not in the input.

transform() copies character by character from standard input and puts it into the end of s after calling toupper() on each one

The count of each character is decremented in the map (by maintaining negative counts we can have descending sort without extra code)

The counts map entries are copied to a set of pairs, swapping (char, count) to (count, char). Since sets are ordered, we get them sorted by decreasing frequency count

Finally we copy the set's content to standard out, using transform, and using select2nd() to pick out only the second member of the pair.

The code is fairly readable. A C++11 solution would look much prettier, since we can use lambdas

C++11 version - no need of lambdas, but auto and range based for makes things very clean (come to think of it you can do very similar with regular C++98)

#include<iostream>
#include<iterator>
#include<map>
#include<set>
using namespace std;int main(){istream_iterator<char> b(cin),e;map<char,int> c;set<pair<int,char>> d;for(char i='A';i<='Z';++i){--c[i];}for(auto i=b;i!=e;++i){c[toupper(*i)]--;}for(auto p:c){d.insert(make_pair(p.second,p.first));}for(auto p:d){cout<<p.second;}}

Expanded version :

#include <iostream>
#include <iterator>
#include <map>
#include <set>
using namespace std;
int main()
{
    istream_iterator<char> b(cin), e;
    map<char, int> c;
    set<pair<int, char>> d;
    for(char i = 'A'; i <= 'Z'; ++i) {--c[i];}
    for(auto i = b; i != e; ++i) {c[toupper(*i)]--;}
    for(auto p : c) { d.insert(make_pair(p.second, p.first)); }
    for(auto p : d) { cout << p.second; }
}

Next iteration (why read from stdin when we have argv):

#include <set>
#include <iostream>
int c[256];int main(int n, char **s){std::set<std::pair<int,char>> d;while(*s[1]){c[toupper(*s[1]++)]--;}for(n=65;n<92;++n){d.insert(std::make_pair(--c[n],n));}for(auto p:d){std::cout<<p.second;}}

Expanded version:

#include <set>
#include <iostream>
int c[256];
int main(int n, char **s)
{
    std::set<std::pair<int, char>> d;
    while (*s[1])
    {
        c[toupper(*s[1]++)]--;
    }
    for (n = 65; n < 92; n++)
    {
        d.insert(std::make_pair(--c[n], n));
    }
    for (auto p : d)
    {
        std::cout << p.second;
    }
}

Answer 26

Java 8, 171 169 154 139 138 bytes

s->{int x=26,a[]=new int[96],i,l;for(int c:s)a[c&~32]++;for(;x-->0;System.out.printf("%c",l),a[l]=-1)for(i=l=65;i<91;i++)l=a[i]>a[l]?i:l;}

Surprised there wasn't a Java answer yet.

-27 bytes thanks to @OlivierGrégoire.
-6 bytes thanks to @Nevay.

Explanation:

Try it here.

s->{               // Method with character-array parameter and no return-type
  int x=26,        //  Counter-integer, starting at 26
  a[]=new int[96], //  Integer array with 96x 0
  i,l;             //  Index integers
  for(int c:s)     //  Loop (1) over the input-array
    a[c&~32]       //   Convert the character to uppercase, and use it's index in the array
             ++;   //   and increase the integer at that current index by 1
                   //  End of loop (1) (implicit / single-line body)
  for(;x-->0       //  Loop (2) from 26 down to 0 (exclusive)
      ;            //    After every iteration:
       System.out.printf("%c",l+65),
                   //     Print the highest frequency uppercase letter
       a[l]=-1)    //     And then change it's frequency to -1, since we're done with it
    for(i=l=65;    //   Reset both `i` and `l` to 65
        i<91;i++)  //   Inner loop (3) from 65 to 91 (exclusive)
      l=a[i]>a[l]? //    If the current frequency is larger than the `l`'th frequency:
         i         //     Change `l` to the current index `i`
        :          //    Else:
         l;        //     Leave `l` the same
                   //   End of inner loop (3) (implicit / single-line body)
                   //  End of loop (2) (implicit / single-line body)
}                  // End of method

Answer 27

APL, 26 20 characters

⎕a[⍒+/⎕a∘.=('\w'⎕r'\u0')⍞]

⎕a[⍒+/⎕a∘.=1(819⌶)⍞]

-6 thanks to Adam.

Answer 28

C++ 377

Implements qsort using letter counts in array n to sort alphabet in array A. Run via command line: golf.exe < in.txt

int n[26],c,k,N;
char A[26];
int C(const void*a,const void*b)
{
int i=(int)(*(char*)a -'A');
int j=(int)(*(char*)b -'A');
return n[j]-n[i];
}
int main()
{
for(;k<26;k++)
{
A[k]=k+'A';
}
N=sizeof(A);
c=getchar();
while(c>0)
{
c=toupper(c);
c=c-'A';
if(c>=0&&c<26)n[c]++;
c=getchar();
}
qsort(A,N,1,C);
for(k=0;k<N;k++)
{
putchar(A[k]);
}
return 0;
}

Answer 29

C, 117(119) bytes

x[256];m=1;char c;main(){while(c=getchar()+1)++x[c-1&95];for(;m=x[++c]<x[m]?m:c;x[m<65|m>90||c?m*!c:putchar(m)]=-1);}

• Some inputs containing ASCII codes >= 128 will incorrectly increase letter frequencies. To fix this, replace the constant 95 with 223, at a cost of 1 extra byte.
• This will terminate early on inputs containing the character with ASCII code 255. To fix this at the cost of 1 extra byte, change char c; to just c; and ++c to c=c+1%255.

Answer 30

PowerShell - 139 chars

First, I'm not an PowerShell expert. Pretty sure there are shorter than this. But was happy with it and decided to share.

$a = Read-host
$b = ($a.ToUpper() -replace '[^A-Z]','').ToCharArray() + (65..90|%{[char[]]$_})|Group|sort Count -desc|%{$_.Name}
-join $b

---

How it works:

$a = Read-host            # read from stdin and save into a string var $a
$a.ToUpper()              # Convert the string to UPPERCASE
-replace'[^A-Z]',''       # Remove all non A-Z characters from the str
(...).ToCharArray()       # Convert the inner object (string) to a Char Array
+  (65..90|%{[char[]]$_}) # Create another char array with A-Z chars expanded, 
                          #  and append it to the previous one.
|Group                    # Group the char array by value for each element, 
                          #  consolidates them and count each char occurrence. Example:
                          #  Count | Name
                          #  ----- | -----
                          #      4 | B
                          #      1 | F
                          #      2 | C 
                          #     .. | ..
                          # 
|sort Count -desc         # Sorts the previous hash-table by the 'Count' column 
                          #   in desc ordering
|%{$_.Name}               # Grab only the 'Name' column from the previous sorted hash-table. 
                          # The retuslt obj will be a simple char array again, 
                          #   with the letters in the desired order
$b = (...)                # Saves the resulting char array into a new variable $b
-join $b                  # join the resulting char array elements into a single 
                          #   string, and print it to stdout. 

Answer 31

Ceylon, 98 bytes

String f(String s)=>String(('A':26).sort(byDecreasing((Object c)=>s.uppercased.count(c.equals))));

Answer 32

K4, 16 bytes

Solution:

.Q.A@>+/.Q.a=/:_

Try it online! (Note this is an oK port/demonstration as there is no TIO for K4).

Example:

Note that the input string is stored as variable 'x' for ease:

q)\
  .Q.A@>+/.Q.a=/:_x
"EITUSALNROMCDPVGQBFHJKWXYZ"

Explanation:

.Q.A@>+/.Q.a=/:_  / the solution
               _  / convert to lowercase
            =/:   / equals each right (/:)
        .Q.a      / lowercase alphabet "a..z"
      +/          / sum
     >            / descending sorted indices
    @             / index into
.Q.A              / uppercase alphabet "A..Z"

Notes:

In q/kdb+ it would be 27 bytes:

.Q.A idesc sum .Q.a=/:lower

Answer 33

Scala 184 170 136 132

 def c(s:String){print((('A'to'Z').mkString+s).filter(_.isLetter).groupBy(x=>x.toUpper).toSeq.sortBy(-_._2.size).map(_._1).mkString)}

Full Version

  def c(s: String) {
    print((('A' to 'Z').mkString + s).filter(_.isLetter).groupBy(x => x.toUpper).toSeq.sortBy(-_._2.size).map(_._1).mkString)
  }

Answer 34

R, 111 characters

Another solution in R (using only base R):

f=function(x)cat(names(sort(table(factor(strsplit(gsub("[^A-Z]","",toupper(x)),"")[[1]],LETTERS)),d=T)),sep="")

Explanation:

toupper(x) makes the input upper case.
gsub("[^A-Z]","",...) gets rid of everything non alphabetical.
strsplit(...,"")[[1]] splits the character string into a vector of single characters.
factor(...,LETTERS) makes a factor out of it with uppercase letters as possible levels.
table(...) build a frequency table of each factor levels.
sort(...,d=T) sorts it according to its values, in decreasing order.
names(...) takes the names of the elements (aka the upper case letters).
cat(...,sep="") prints the result to stdout with no separator.
And everything wrapped into a function to make it directly comparable with djurio solution.

Usage:

> f=function(x)cat(names(sort(table(factor(strsplit(gsub("[^A-Z]","",toupper(x)),"")[[1]],LETTERS)),d=T)),sep="")
> f("Lorem ipsum dolor sit amet, consectetur adipiscing elit. Praesent vitae erat velit. Mauris gravida euismod libero ut tincidunt. Phasellus elit dui, consectetur et egestas in, aliquam vitae diam. Donec eget varius ante. Vestibulum cursus diam aliquet, egestas orci quis, placerat dolor. Proin vel nisi lectus. Class aptent taciti sociosqu ad litora torquent per conubia nostra, per inceptos himenaeos. Aliquam erat volutpat. Etiam libero tortor, ornare id dui eget, posuere dignissim libero. Pellentesque commodo consequat volutpat. Integer hendrerit sapien libero, vel viverra augue facilisis sit amet. Quisque consectetur eget nisl quis dignissim. Ut lacinia pretium quam a placerat.
+ Morbi sed interdum risus, nec pretium lectus. Morbi imperdiet est id accumsan molestie. Duis sed fermentum nisl. Nunc vitae augue mattis, dictum lectus vel, accumsan nisl. Sed ultricies adipiscing rhoncus. Vivamus eu lacus a enim venenatis eleifend. Praesent consectetur tortor non eleifend ultricies. Mauris et odio posuere, auctor erat at, fringilla est. Proin in vestibulum erat. Maecenas congue commodo ante vel varius. Sed tempus mi ut metus gravida, nec dictum libero dapibus. Morbi quis viverra elit. Ut pharetra neque eget lacus tincidunt dictum. Fusce scelerisque viverra tellus et pretium.
+ Fusce varius adipiscing odio. Nulla imperdiet faucibus sem, at rhoncus ipsum adipiscing vitae. Phasellus imperdiet congue lacus et mollis. Nullam egestas mauris magna, et mollis lectus varius ut. Sed sollicitudin adipiscing dolor, vel elementum elit laoreet molestie. Aliquam nec nulla vel sem ultrices ullamcorper. Nullam nec felis magna. Duis sodales orci non justo aliquam tempus. Integer mi diam, tempor sed vulputate et, varius et nunc. Vestibulum sodales ipsum id mi pharetra, ut convallis mi accumsan. Sed dictum volutpat vestibulum.
+ Quisque ac dolor sagittis, aliquam libero at, euismod enim. Nulla ullamcorper posuere nulla vitae varius. Nam at dolor non libero elementum pellentesque in in lorem. Fusce porttitor turpis in quam placerat varius. Donec lorem orci, condimentum eu sapien sit amet, aliquet commodo magna. Quisque sed lectus sit amet arcu euismod accumsan et non nunc. Phasellus placerat congue metus, feugiat posuere leo dictum quis. Sed ultricies feugiat eros dignissim bibendum.
+ Mauris scelerisque consectetur libero eget varius. Aenean neque nunc, ullamcorper vitae orci in, auctor ornare sapien. Nam lacinia molestie imperdiet. Nam vitae mattis nibh. Vestibulum consequat tellus ac nisi sagittis pulvinar. Nullam mollis ornare quam, et venenatis leo porttitor sit amet. Nulla urna neque, dignissim non orci ut, volutpat ultrices erat. Cum sociis natoque penatibus et magnis dis parturient montes, nascetur ridiculus mus. Pellentesque vestibulum tellus nec eros faucibus porta.")
EITUSALNROMCDPVGQBFHJKWXYZ
> f("Based off ETAOIN SHRDLU, your challenge is to write the shortest program or function in any language that outputs the 26 letters of the English alphabet based on their frequency in the input.")
ETNAHORISULFGPBCDYMQWJKVXZ

Answer 35

Retina, 81 75 bytes (not competing)

The language (especially the Sort mode) is newer than the challenge.

;
{T`;L`L`.$
}`[^Z]$
$0$0
T`l`L
Os`.
[^A-Z]

(.)\1*
$0¶
O#$^`.*
$.0
¶

D`.

Try it online

Explanation:

;               # Add the entire alphabet plus a bunch of semicolons into the input
{T`;L`L`.$          # (so that all letters are included in the input at least once)
}`[^Z]$             #
$0$0                #
T`l`L           # Translate all lowercase into uppercase
Os`.            # Sort all characters by code point
[^A-Z]          # Remove non-letters

(.)\1*          # Separate same characters by newlines
$0¶
O#$^`.*         # Sort backwards numerically by the length
$.0
¶               # Remove newlines

D`.             # Remove excess letters

Answer 36

PHP, 92+1 bytes (late entry)

$r=count_chars(strtoupper($argn));arsort($r);foreach($r as$c=>$n)$c>64&$c<91&&print chr($c);

Run as pipe with -nR or try it online.

Answer 37

05AB1E, 9 bytes (Non-competing)

lAΣ¢}ÙRáu

Try it online!

---

More interesting 10-byte version:

05AB1E, 10 bytes

l{γéJAìÚRá

Try it online!

Answer 38

Perl, 43 bytes

adding 1 for -n, as usual.

I know variable variables are generally a bad idea and shouldn't be used, but hey it's code golf :)

#!perl -n
${uc$_}++for/\w/g}{print+sort{$$b-$$a}A..Z

Answer 39

k4, 21 bytes

  {.Q.A@<<>#:'.Q.a#=_x}
  {.Q.A@<<>#:'.Q.a#=_x}@,/0:`:e
"EITUSALNROMCDPVGQBFHJKWXYZ"

This approach uses the builtin alphabetic constants .Q.a (the lowercase letters) and .Q.A (the uppercase) to do the filtering and translation back to upper. The process:

• _: convert to lowercase (leaves non-letters untouched)
• =: group into associative array of characters and their indices
• .Q.a#: take only lowercase letters (includes empty lists for characters that did not occur)
• #:': count each value in the array
• >: get the keys of the array ordered by its values, descending
• <<: find the position where each item would appear if the list were sorted
• .Q.A@: apply that list of positions back to the list of uppercase letters

The last two steps are conceptually a bit tricky; they basically allow you to permute one list according to another. See the q documentation for rank for more details.

(Note that if we stop before the translation back to uppercase, we have a correct result, but in lowercase:

  {>#:'.Q.a#=_x}@,/0:`:e
"eitusalnromcdpvgqbfhjkwxyz"

and a code length of 14!)

Answer 40

PHP 135 Bytes

I know I'm late but I wanted to post anyway:

function a($b){
  $b=strtoupper($b);
  foreach(range('A','Z') as $c){
   $d[$c]=substr_count($b, $c);
  }  
  arsort($d);
  return implode(array_keys($d));
}
echo a($z); //let $z be the supplied string.

golfed:

function a($b){$b=strtoupper($b);foreach(range('A','Z') as $c){$d[$c]=substr_count($b, $c);}arsort($d);return implode(array_keys($d));}echo a($z); 

Answer 41

Factor, 327 261 248 bytes

:o

CONSTANT: a "ABCDEFGHIJKLMNOPQRSTUWXYZ"
: s ( s -- s ) >array [ 1string ] map ;
[ >upper s dup a s [ [ = ] curry ] map [ [ count ] curry ] map [ call( x -- x ) ] with map [ a s union natural-sort ] dip zip [ last ] sort-with [ first ] map "" join ]

Writing this took a while since I'm still learning the language, but it was fun.

Run the above code, then do "string" swap call to call the lambda. Less golfed:

USING: arrays ascii assocs combinators kernel sequences
  sequences.generalizations sets sorting strings ;
IN: input-freq

CONSTANT: alph "ABCDEFGHIJKLMNOPQRSTUWXYZ"

: alphabet-counter ( -- alph-quot )
  alph string>array [
    [ = ] curry
  ] map [
    [ count ] curry
  ] map ;

: string>array ( str -- array )
  >array [ 1string ] map ;

: freq-find ( str -- freqmap )
  >upper string>array dup alphabet-counter
  [ call( x -- x ) ] with map
  [ alph string>array union natural-sort ] dip zip
  [ last ] sort-with [ first ] map "" join ;

Answer 42

Lua, 226 Bytes

I'm coming in long after this contest has ended, but I though it would be fun to do :). Also, it may be one of the longest submission down here, but actually took me some time to find a clever way to do it. It may still be golfable I'll explain it below, but right now, here's the code:

f,t={},{}(...):upper():gsub("%u",function(c)c=c:byte()-64 f[c]=f[c]and f[c]+1or 1 end)for i=1,26
do c=(i+64 ..''):char()f[i]=f[i]and f[i]or 0 x=f[i]t[x]=t[x]and c..t[x]or c end
for i=#...,0,-1 do io.write(t[i]and t[i] or'')end

It's a full program, taking its input via command-line argument as a string.

Ungolfed and explanations

f={}                       -- f will be our array for counting iteration
t={}                       -- t will be the sorted array
(...):upper()              -- convert the input in all uppercase
     :gsub("%u",function(c)-- then iterate over each uppercase letter in it
  c=c:byte()-64            -- set c at its position in the alphabet
  f[c]=f[c]                -- if f[c] is set (letter already encountered)
       and f[c]+1          -- increment it
       or 1                -- else, we set it to 1
end)
   -- At that point, wev'e counted each character, let's sort based on occurence!
for i=1,26                 -- iterate on the size of the alphabet
do
  c=(i+64 ..''):char()     -- c is the character corresponding to the current i
  f[i]=f[i]and f[i]or 0    -- if c wasn't in the input, initialise a counter at 0
  x=f[i]                   -- set x to the number of occurence of c
  t[x]=t[x]                -- if t[x] is set, two or more letter have the same count
       and c..t[x]         -- so concatenate the string in t[x] with the new letter
       or c                -- else, just set it
end

for i=#...,0,-1            -- iterate from the size of input to 0
do
    io.write(t[i]          -- if we have i occurence of letter(s)
         and t[i]          -- output them
         or'')             -- else, output an empty string
end

This algorythm is really inefficient. It works because arrays don't exist in lua, we only use tables, which are associative arrays. Which means I can map t[123456]='B' even if t[1] doesn't exist. It makes sorting the number of occurence really fast, but if you want to find them, the time will depend on the size of the input you gave to the program.

Answer 43

Mathcad, 102 "bytes"

Mathcad doesn't index strings, so the function f converts the string to a vector of its equivalent character codes (subtracting A's code to aid indexing). f then counts each occurence of the letters A..Z (taking 32 of the values of codes >= "a" to implicitly convert case).

Then there's a verbose sequence of nested functions that sort the character codes by count in ascending order, extract the sorted codes in reverse order and add the "A" offset back, and convert the resulting character code vector back into a string.

I interpret Mathcad "bytes" as the number of keyboard / menu operations necessary to create a Mathcad symbol.

• Identifier names, eg submatrix, are typed in letter-by-letter, hence the byte count = number of characters in the name.
• Type ' to add a balanced pair of parentheses
• Array indices are entered by typing the array name and then the character '[' followed by the indices. For c₂₅, this would be the 4 bytes c [ 2 5.
• Type ctl-shft-# A to create a "for loop"; this results in a for programming construct that automatically displays the for keyword and the element-of symbol, plus three placeholders. The placeholders are for the iteration variable, it's range and the main body of the loop, respectively.
• Type ] to add a new programming line. See the image below ...

Answer 44

Floroid - 37 bytes

Ba:''.y(t(R(P,w(65,91)),fi=a.fk.fm))D

A little late to the party, but here it is. An Floroid port of @undergroundmonorails answer.

This language was made some time after this challenge, therefore this isn't competing.

Answer 45

VB.net

Dim r = String.Join("",( From a In "ABCDEFGHIJKLMNOPQRSTUVWXYZ"
                         Group Join c In t On a Equals c Into Group
                         Order By Group.Count Descending
                       ).Select(Function(g) g.a))