AWK, 61 bytes

$0=$0($0<9||$0>14?$NF~1?"st":$NF~2?"nd":$NF~3?"rd":"th":"th")

Attempt This Online!

Tcl, 74 bytes

proc o n {puts $n[lindex {th st nd rd} [expr $n%100-11&&$n%10<4?$n%10:0]]}

Try it online!

J -  44  41 char

Nothing in J? This is an outrage!

(":,th`st`nd`rd{::~4|4<.10(|*|~:-~)100&|)

Explained (note that 1 is boolean true in J and 0 is false):

• 100&| - First, reduce the input modulo 100.
• 10(|*|~:-~) - In this subexpression, | is the input mod 10 and -~ is the input (mod 100) minus 10. You don't get many trains as clean as this one in J, so it's a treat to see here!
  • ~: is not-equals, so |~:-~ is false if the tens digit is 1 and true otherwise.
  • * is just multiplication, so we're taking the input mod 10 and multiplying by a boolean, which will zero it out if false. The result of this is "input mod 10, except 0 on the tens".
• 4|4<. - Min (<.) the above with 4, to clamp down larger values, and then reduce modulo 4 so that they all go to zero.
• th`st`nd`rd{::~ - Use that as an index into the list of suiffixes. Numbers ending in X1 or X2 or X3, but not in 1X, will find their suffix, and everything else will take the 0th suffix th.
• ":, - Finally, take the original input, convert it to a string (":), and append the suffix.

Usage is obvious, though as-is the verb can only take one ordinal, not a list.

   (":,th`st`nd`rd{::~4|4<.10(|*|~:-~)100&|) 112         NB. single use
112th
   (":,th`st`nd`rd{::~4|4<.10(|*|~:-~)100&|) 1 2 3 4 5   NB. doing it wrong
|length error
|       (":,th`st`nd`rd{::~4|4<.10(|*|~:-~)100&|)1 2 3 4 5
   NB. i.5 10   makes a 5x10 grid of increasing integers
   NB. &.>      to operate on each integer separately, and box the result after
   (":,th`st`nd`rd{::~4|4<.10(|*|~:-~)100&|)&.> i.5 10   NB. all better
+----+----+----+----+----+----+----+----+----+----+
|0th |1st |2nd |3rd |4th |5th |6th |7th |8th |9th |
+----+----+----+----+----+----+----+----+----+----+
|10th|11th|12th|13th|14th|15th|16th|17th|18th|19th|
+----+----+----+----+----+----+----+----+----+----+
|20th|21st|22nd|23rd|24th|25th|26th|27th|28th|29th|
+----+----+----+----+----+----+----+----+----+----+
|30th|31st|32nd|33rd|34th|35th|36th|37th|38th|39th|
+----+----+----+----+----+----+----+----+----+----+
|40th|41st|42nd|43rd|44th|45th|46th|47th|48th|49th|
+----+----+----+----+----+----+----+----+----+----+

Alternative 41 char solution showing off a couple logical variants:

(":;@;st`nd`rd`th{~3<.10(|+3*|=-~)100|<:)

Previously I had an overlong explanation of this 44 char hunk of junk. 10 10#: takes the last two decimal digits and /@ puts logic between them.

(":,th`st`nd`rd{::~10 10(]*[(~:*])4>])/@#:])

Go, 173 110 bytes

import."fmt"
func o(n int){s,k:="th",n%10
if k>0&&k<4&&n/10%10!=1{s=[]string{"st","nd","rd"}[k-1]}
Print(n,s)}

Attempt This Online!

• -47 bytes from using slice indexing (by @Steffan)
• -14 bytes from outputting to STDOUT (by @Steffan)
• -2 bytes from removing duplicate n%10.

Old answer, 173 bytes

import."fmt"
func o(n int)string{s:=Sprint(n)
switch n%100{case 11,12,13:s+="th"
default:switch n%10{case 1:s+="st"
case 2:s+="nd"
case 3:s+="rd"
default:s+="th"}}
return s}

Attempt This Online!

Vyxal, 6 bytes

∆o2NȯJ

Try it Online!

Javascript (ES6) 50 44 Bytes

a=>a+=[,"st","nd","rd"][a.match`1?.$`]||"th"

Notes

• takes imput as a string
• Removed 6 bytes, thanks @user81655

Python 2, 49 bytes

lambda n:`n`+'tsnrhtdd'[n%5*(n%100^15>4>n%10)::4]

An anonymous function. A full program would be counted at 55 bytes.

'tsnrhtdd'[i::4] encodes the suffixes th st nd rd for values of i from 0 to 3. Given this, all we need is a way to map the values of n to the index of the corresponding suffix, i. A straightforward expression that works is (n%10)*(n%10<4 and not 10<n%100<14). We can easily shorten this by dropping the first set of parentheses and observing that n%5 gives the same results as n%10 for the values of n with the special suffixes. With a bit of trial and error, one may also shorten not 10<n%100<14 to n%100^15>4, which can be chained with the other conditional to save even more bytes.

brainfuck, 313 bytes

This was painful. Setup could probably be golfed better. May write a half-decent explanation later.

++++[>++++++<-]>-[<+++++>-]<[->+>+>+>+>+>+>+>+<<<<<<<<]>->--------------->----->--------------->>+>+>----------->>>>+>+>+>,[.>,]<<<[-]>>>-[>+<-----]>---[-<<-<->>>]<<<-[[-]>[->+>++<<]>+<+++++[->-[>]<<<]>[<<<]>[>>[-<<<<<[<]<<<<<[>]<[[->+<]<]>>>>>>>>>[-<<<<<<<<+>>>>>>>>]>>>>>[>]>>>>]<<<]]<<<[<]<<[<]<<<[<]>>>>>>>.>.

Try it online!

Vyxal, 14 bytes

ÞoÞ∞ZƛhṘ2ẎṘntp

Try it Online!

ÞoÞ∞Z          # Zip ordinals with positive integers
     ƛ         # Over each
      hṘ2ẎṘ    # Get the last two characters of the ordinal
           ntp # Append the number

Scratch, idk bytes

I tried to convert it to sb syntax, but it did not work

Try it!

Please comment if you know a way to golf, or byte-count this

Vyxal, 23 bytes

₀%:4<?₀ḭċ∧*4«∨¦İk√«/$i+

Try it Online!

Explanation:

₀%:4<?₀ḭċ∧*4«∨¦İk√«/$i+    # full program

₀%:                         # push input mod 10 twice
   4<                       # less than 4
     ?₀ḭ                    # input floor div 10
        ċ                   # != 1
         ∧*                 # and both bools, then multiply by initial mod 10
            4«∨¦İk√«/       # "thstndrd" split into 4 parts
                      $i    # indexed by input mod 10
                        +   # concat

Bash, 58 bytes

Just a modification of @stephanmg's post in Bash. Output is on STDERR as in "command not found".

case $1 in
*1?|*[04-9])$1th;;*1)$1st;;*2)$1nd;;*)$1rd
esac

Try it online!

This program completely relies on the limitation that the input is a non-negative decimal integer.

C, 87 bytes (correct teen handling)

This C solution is longer than the current shortest 83 byte C solution posted, but as of writing, that 83 byte solution is incorrect and prints numbers like 11st, 12nd, and 13rd.

Here is my solution, which correctly handles teen numbers:

main(n){scanf("%d",&n);printf("%d%s",n,"th\0st\0nd\0rd"+(n%10<4&&n/10%10-1?n%10*3:0));}

Degolfed:

main(n) {
    scanf("%d", &n);
    printf("%d%s", n, "th\0st\0nd\0rd" + (n%10 < 4 && n/10%10 - 1 ? n%10*3 : 0));
}

FEU, 41 bytes

m/1.$/\0th/1$/1st/2$/2nd/3$/3rd/\d$/\0th/

Try it online!

APL(NARS), 38 chars, 76 bytes

(⍕,{((⍳3)⍳10∣⍵×3≠⌊√⍵)⊃'snrt',¨'tddh'})

test:

  h←(⍕,{((⍳3)⍳10∣⍵×3≠⌊√⍵)⊃'snrt',¨'tddh'})
  h¨0..19
 0th 1st 2nd 3rd 4th 5th 6th 7th 8th 9th 10th 11th 12th 13th 14th 15th 16th 17th 18th 19th 
  h¨20..31
 20th 21st 22nd 23rd 24th 25th 26th 27th 28th 29th 30th 31st 
  ⎕fmt h¨100..104
┌5───────────────────────────────────────────┐
│┌5─────┐ ┌5─────┐ ┌5─────┐ ┌5─────┐ ┌5─────┐│
││ 100th│ │ 101st│ │ 102nd│ │ 103rd│ │ 104th││
│└──────┘ └──────┘ └──────┘ └──────┘ └──────┘2
└∊───────────────────────────────────────────┘

Bash (Cardinal to ordinal) 91 Bytes:

function c(){ case "$1" in *1[0-9]|*[04-9])"$1"th;;*1)"$1"st;;*2)"$1"nd;;*3)"$1"rd;;esac }

05AB1E, 21 bytes

т%T‰ć≠*4ªß…thŠØ3ôs訫

Try it online or verify a few more test cases at once.

Explanation:

Mostly ported from a sub-section of this answer of mine, except with additional leading т%.

т%                     # Take modulo-100 on the (implicit) input-integer
  T‰                   # Take the divmod-10 ([n//10,n%10]) on that
    ć                  # Extract the head: n//10
     ≠                 # Check that it's NOT equal to 1 (0 if 1; 1 otherwise)
      *                # Multiply the [n%10] by this
       4ª              # Append a 4 to this list: [n%10*(n//10!=1),4]
         ß             # Pop and push the smallest number in this list
          …thŠØ        # Push dictionary string "th standards"
               3ô      # Split it into parts of size 3: ["th ","sta","nda","rds"]
                 sè    # Index the calculated minimum into this list
                       # (modular 0-based, so both 0 and 4 will index into "th ")
                   ¨   # Remove the last character
                    «  # And append it to the (implicit) input-integer
                       # (after which the result is output implicitly)

See this 05AB1E tip of mine (section How to use the dictionary?) to understand why …thŠØ is "th standards".

Haskell, 95 100 chars

h=foldr g"th".show
g '1'"th"="1st"
g '2'"th"="2nd"
g '3'"th"="3rd"
g '1'[x,_,_]='1':x:"th"
g a s=a:s

Testing:

*Main> map h [1..40]
["1st","2nd","3rd","4th","5th","6th","7th","8th","9th","10th","11th","12th","13t
h","14th","15th","16th","17th","18th","19th","20th","21st","22nd","23rd","24th",
"25th","26th","27th","28th","29th","30th","31st","32nd","33rd","34th","35th","36
th","37th","38th","39th","40th"]

Must be loaded with -XNoMonomorphismRestriction.

Javascript, 75 60

With the new arrow notation:

o=(s)=>s+((0|s/10%10)==1?"th":[,"st","nd","rd"][s%10]||"th")

Old version, without arrow notation (75 chars):

function o(s){return s+((0|s/10%10)==1?"th":[,"st","nd","rd"][s%10]||"th")}

APL (Dyalog Unicode), 38 36 bytes

Thanks to ngn for fixing a bug while maintaining byte count.

Anonymous tacit prefix function. Requires ⎕IO (Index Origin) set to 0, which is default on many systems. Even works for 0!

⍕,{2↑'thstndrd'↓⍨2×⊃⍵⌽∊1 0 8\⊂10↑⍳4}

Try it online!

{…} anonymous lambda; ⍵ is argument:

 ⍳4 first four ɩndices; [0,1,2,3]

 10↑ take first ten elements from that, padding with zeros: [0,1,2,3,0,0,0,0,0,0]

 ⊂ enclose to treat as single element; [[0,1,2,3,0,0,0,0,0,0]]

 1 0 8\ expand to one copy, a prototypical copy (all-zero), eight copies;
  [[0,1,2,3,0,0,0,0,0,0],
   [0,0,0,0,0,0,0,0,0,0],
   [0,1,2,3,0,0,0,0,0,0],
   [0,1,2,3,0,0,0,0,0,0],
   ⋮ (5 more)
   [0,1,2,3,0,0,0,0,0,0]]

 ∊ ϵnlist (flatten);
  [0,1,2,3,0,0,0,0,0,0,
   0,0,0,0,0,0,0,0,0,0,
   0,1,2,3,0,0,0,0,0,0,
   0,1,2,3,0,0,0,0,0,0,
   ⋮ (50 more)
   0,1,2,3,0,0,0,0,0,0]

 ⍵⌽ cyclically rotate left as many steps as indicated by the argument

 ⊃ pick the first number (i.e. the argument-mod-100'th number)

 2× multiply two by that (gives 0, 2, 4, or 6)

 'thstndrd'↓⍨drop that many characters from this string

 2↑ take the first two of the remaining characters

⍕, concatenate the stringified argument to that

R, 79 76 bytes

Since there is no R solution yet... no tricks here, basic vector indexing, golfed down 3 chars thanks to Giuseppe. Previously tried index: [1+(x%%10)-(x%%100==11)] and [1+(x%%10)*(x%%100!=11)].

function(x)paste0(x,c("th","st","nd","rd",rep("th",6))[1+x%%10*!x%%100==11])

Try it online!

With substr, 79 bytes:

function(x,y=1+2*min(x%%10-(x%%100==11),4))paste0(x,substr("thstndrdth",y,y+1))

Try it online!

Julia 0.6, 68 bytes

f(x,m=x%10,p=x%100-m)="$x$(["th","st","nd","rd"][m>3||p==10?1:m+1])"

Try it online!

Straightforward remainder rules in Julia.

JavaScript (Node.js), 51 bytes

credit to @KevinCruijssen for improving the answer

n=>n+=[,'st','nd','rd'][~~(n/10%10)-1?n%10:0]||'th'

Try it online!

Explanation :

n =>                      // input
    n+=[,'st','nd','rd']     // the array of suffices add everything to n
        [~~(n/10%10)-1?n%10:0] // the full magic 
    || 'th'               // works for everything except 1 , 2 , 3 just like what you want

Procedural Footnote Language, 39 bytes

[1]
[PFL1.0]
[1] [ORD:[INPUT]]
[PFLEND]

The BODY section of the document contains a reference to footnote [1]; footnote [1] contains the [ORD]inal representation of the [INPUT].

JavaScript, 64 characters (ES3) or 47 characters (ES6)

ES3 (64 characters):

function(n){return n+=[,'st','nd','rd'][n%100>>3^1&&n%10]||'th'}

ES6 (47 characters):

n=>n+=[,'st','nd','rd'][n%100>>3^1&&n%10]||'th'

Explanation

The expression n % 100 >> 3 ^ 1 evaluates to 0 for any positive n ending with digits 08–15. Thus, for any n mod 100 ending in 11, 12, or 13, the array lookup returns undefined, leading to a suffix of th.

For any positive n ending in other digits than 08–15, the expression n % 100 >> 3 ^ 1 evaluates to a positive integer, invoking the expression n % 10 for array lookup, returning st,nd, or rd for n which ends with 1, 2, or 3. Otherwise, th.

Pyth, 33 bytes

+Q@.>+c"stndrd"2*]"th"7 1?qh`Q\1Z

Try it online!

Common Lisp, 63

(format t"~a~a"i(let((x(format()"~:r"i)))(subseq x(-(length x)2))))

Input is i. There's probably a more clever way to golf this.

Java 7, 91 81 bytes

String d(int n){return n+(n/10%10==1|(n%=10)<1|n>3?"th":n<2?"st":n<3?"nd":"rd");}

Port from @adrianmp's C# answer.

Old answer (91 bytes):

String c(int n){return n+((n%=100)>10&n<14?"th":(n%=10)==1?"st":n==2?"nd":n==3?"rd":"th");}

Ungolfed & test code:

Try it here.

class M{
  static String c(int n){
    return n + ((n%=100) > 10 & n < 14
                ?"th"
                : (n%=10) == 1
                   ? "st"
                   : n == 2
                      ? "nd"
                      : n == 3
                         ? "rd"
                         :"th");
  }

  static String d(int n){
    return n + (n/10%10 == 1 | (n%=10) < 1 | n > 3
                 ? "th"
                 : n < 2
                    ? "st"
                    : n < 3
                       ? "nd"
                       :"rd");
  }

  public static void main(String[] a){
    for(int i = 1; i < 201; i++){
      System.out.print(c(i) + ", ");
    }
    System.out.println();
    for(int i = 1; i < 201; i++){
      System.out.print(d(i) + ", ");
    }
  }
}

Output:

1st, 2nd, 3rd, 4th, 5th, 6th, 7th, 8th, 9th, 10th, 11th, 12th, 13th, 14th, 15th, 16th, 17th, 18th, 19th, 20th, 21st, 22nd, 23rd, 24th, 25th, 26th, 27th, 28th, 29th, 30th, 31st, 32nd, 33rd, 34th, 35th, 36th, 37th, 38th, 39th, 40th, 41st, 42nd, 43rd, 44th, 45th, 46th, 47th, 48th, 49th, 50th, 51st, 52nd, 53rd, 54th, 55th, 56th, 57th, 58th, 59th, 60th, 61st, 62nd, 63rd, 64th, 65th, 66th, 67th, 68th, 69th, 70th, 71st, 72nd, 73rd, 74th, 75th, 76th, 77th, 78th, 79th, 80th, 81st, 82nd, 83rd, 84th, 85th, 86th, 87th, 88th, 89th, 90th, 91st, 92nd, 93rd, 94th, 95th, 96th, 97th, 98th, 99th, 100th, 101st, 102nd, 103rd, 104th, 105th, 106th, 107th, 108th, 109th, 110th, 111th, 112th, 113th, 114th, 115th, 116th, 117th, 118th, 119th, 120th, 121st, 122nd, 123rd, 124th, 125th, 126th, 127th, 128th, 129th, 130th, 131st, 132nd, 133rd, 134th, 135th, 136th, 137th, 138th, 139th, 140th, 141st, 142nd, 143rd, 144th, 145th, 146th, 147th, 148th, 149th, 150th, 151st, 152nd, 153rd, 154th, 155th, 156th, 157th, 158th, 159th, 160th, 161st, 162nd, 163rd, 164th, 165th, 166th, 167th, 168th, 169th, 170th, 171st, 172nd, 173rd, 174th, 175th, 176th, 177th, 178th, 179th, 180th, 181st, 182nd, 183rd, 184th, 185th, 186th, 187th, 188th, 189th, 190th, 191st, 192nd, 193rd, 194th, 195th, 196th, 197th, 198th, 199th, 200th, 
1st, 2nd, 3rd, 4th, 5th, 6th, 7th, 8th, 9th, 10th, 11th, 12th, 13th, 14th, 15th, 16th, 17th, 18th, 19th, 20th, 21st, 22nd, 23rd, 24th, 25th, 26th, 27th, 28th, 29th, 30th, 31st, 32nd, 33rd, 34th, 35th, 36th, 37th, 38th, 39th, 40th, 41st, 42nd, 43rd, 44th, 45th, 46th, 47th, 48th, 49th, 50th, 51st, 52nd, 53rd, 54th, 55th, 56th, 57th, 58th, 59th, 60th, 61st, 62nd, 63rd, 64th, 65th, 66th, 67th, 68th, 69th, 70th, 71st, 72nd, 73rd, 74th, 75th, 76th, 77th, 78th, 79th, 80th, 81st, 82nd, 83rd, 84th, 85th, 86th, 87th, 88th, 89th, 90th, 91st, 92nd, 93rd, 94th, 95th, 96th, 97th, 98th, 99th, 100th, 101st, 102nd, 103rd, 104th, 105th, 106th, 107th, 108th, 109th, 110th, 111th, 112th, 113th, 114th, 115th, 116th, 117th, 118th, 119th, 120th, 121st, 122nd, 123rd, 124th, 125th, 126th, 127th, 128th, 129th, 130th, 131st, 132nd, 133rd, 134th, 135th, 136th, 137th, 138th, 139th, 140th, 141st, 142nd, 143rd, 144th, 145th, 146th, 147th, 148th, 149th, 150th, 151st, 152nd, 153rd, 154th, 155th, 156th, 157th, 158th, 159th, 160th, 161st, 162nd, 163rd, 164th, 165th, 166th, 167th, 168th, 169th, 170th, 171st, 172nd, 173rd, 174th, 175th, 176th, 177th, 178th, 179th, 180th, 181st, 182nd, 183rd, 184th, 185th, 186th, 187th, 188th, 189th, 190th, 191st, 192nd, 193rd, 194th, 195th, 196th, 197th, 198th, 199th, 200th, 

Mathematica 29 + 5 = 34 bytes

SpokenStringDump`SpeakOrdinal

+5 bytes because Speak function must be called before using this built-in.

Usage

SpokenStringDump`SpeakOrdinal[1]

"1st "

SpokenStringDump`SpeakOrdinal[4707]

"4,707th "

Scala 2.11.7, 151 Chars

def o(x:Int*)=x map{
case a if(a%10)==1&&a%100!=11=>a+"st"
case b if(b%10)==2&&b%100!=12=>b+"nd"
case c if(c%10)==3&&c%100!=13=>c+"rd"
case e=>e+"th"
}

Usage

   o(1,2,3,4) -> ArrayBuffer("1st", "2nd", "3rd", "4th")

C#, 62 bytes

n=>n+(n/10%10==1||(n%=10)<1||n>3?"th":n<2?"st":n<3?"nd":"rd");

Full program and verification:

using System;

namespace OutputOrdinalNumbers
{
    class Program
    {
        static void Main(string[] args)
        {
            Func<int,string>f= n=>n+(n/10%10==1||(n%=10)<1||n>3?"th":n<2?"st":n<3?"nd":"rd");
    
            for (int i=1; i<=124; i++)
                Console.WriteLine(f(i));
        }
    }
}

Python, 88 84 bytes

lambda x:x+((('th','st','nd','rd')+('th',)*6)[int(x[-1])]if('0'+x)[-2]!='1'else'th')

Ungolfed:

def y(x):
    return x + (('th', 'st', 'nd', 'rd') + ('th', ) * 6)[int(x[-1])] if ('0' + x)[-2] != '1' else 'th')

lambda x defines an anonymous function with parameter x. ((('th','st','nd','rd')+('th',)*6)[int(x[-1])] defines a tuple of the endings for numbers less than 10, the 0-th element is for 0, and so forth. the if ('0'+x)[-2] != '1' checks if there is 11, 12, or a 13 to fix, and adds then else 'th' adds th instead of st, rd, or nd.

Haxe, 83 chars

function o(n:Int){return ['th','st','nd','rd'][n%100>20||n%100<4?n%10>3?0:n%10:0];}

Based on the javascript version, fixed for n > 100 as well.

Javascript ES6, 52 chars

n=>n+(!/1.$/.test(--n)&&'snr'[n%=10]+'tdd'[n]||'th')

Oracle SQL 11.2, 101 bytes

SELECT:1||DECODE(ROUND(MOD(:1,100),-1),10,'th',DECODE(MOD(:1,10),1,'st',2,'nd',3,'rd','th'))FROM DUAL

PHP, 98 bytes

function c($n){$c=$n%100;$s=['','st','nd','rd'];echo$c>9&&$c<14||!$s[$c%10]?$n.'th':$n.$s[$c%10];}

The 11-13 bit is killing me here. Works for any integer $n >= 0.

For any integer $n:

PHP, 103 bytes

function c($n){$c=abs($n%100);$s=['','st','nd','rd'];echo$c>9&&$c<14||!$s[$c%10]?$n.'th':$n.$s[$c%10];}

Mathematica 39 45 bytes

Note: In recent versions of Mathematica, asking for the nth part of p, where p is undefined, generates an error message, but returns the correct answer anyway. I've added Quiet to prevent the error message from printing.

Quiet@StringSplit[SpokenString[p[[#]]]][[2]]&

Usage

Quiet@StringSplit[SpokenString[p[[#]]]][[2]] &[31]

31st

Quiet@StringSplit[SpokenString[p[[#]]]][[2]] &/@Range[21]

{"1st", "2nd", "3rd", "4th", "5th", "6th", "7th", "8th", "9th", "10th", "11th", "12th", "13th", "14th", "15th", "16th", "17th", "18th", "19th", "20th", "21st"}

How it works

SpokenString writes out an any valid Mathematica expression as it might be spoken. Below are two examples from the documentation for SpokenString,

SpokenString[Sqrt[x/(y + z)]]

"square root of the quantity x over the quantity y plus z" *)

SpokenString[Graphics3D[Sphere[{{0, 0, 0}, {1, 1, 1}}]], "DetailedGraphics" -> True]

"a three-dimensional graphic consisting of unit spheres centered at 0, 0, 0 and 1, 1, 1"

Now, for the example at hand,

Quiet@SpokenString[p[[#]]] &[31]

"the 31st element of p"

Let's represent the above string as a list of words:

StringSplit[%]

{"the", "31st", "element", "of", "p"}

and take the second element...

%[[2]]

31st

Javascript, 75

function s(i){return i+'thstndrd'.substr(~~(i/10%10)-1&&i%10<4?i%10*2:0,2)}

C - 95 83 characters

main(n,k){scanf("%d",&n);k=(n+9)%10;printf("%d%s\n",n,k<3?"st\0nd\0rd"+3*k:"th");}

Degolfed:

main(n,k)
{
    scanf("%d",&n);
    k=(n+9)%10; // xx1->0, xx2->1, xx3->2
    printf("%d%s\n",n,k<3?"st\0nd\0rd"+3*k:"th");
}

We could do k=(n-1)%10 instead of adding 9, but for n=0 we would get an incorrect behaviour, because in C (-1)%10 evaluates to -1, not 9.

K - 44 char

It so happens that this is exactly as long as the J, and works in almost the same way.

{x,$`th`st`nd`rd@{y*(y<4)*~1=x}.-2#0,.:'x$:}

Explained:

• x$: - First, we convert the operand x into a string, and then assign that back to x. We will need its string rep again later, so doing it now saves characters.
• .:' - Convert (.:) each (') digit back into a number.
• -2#0, - Append a 0 to the front of the list of digits (in case of single-digit numbers), and then take the last two.
• {y*(y<4)*~1=x}. - Use the two digits as arguments x and y to this inner function, which returns y if y is less than 4 and x is not equal to 1, otherwise 0.
• `th`st`nd`rd@ - Index the list of suffixes by this result.
• x,$ - Convert the suffix from symbol to string, and append it to the original number.

Usage:

  {x,$`th`st`nd`rd@{y*(y<4)*~1=x}.-2#0,.:'x$:} 3021
"3021st"
  {x,$`th`st`nd`rd@{y*(y<4)*~1=x}.-2#0,.:'x$:}' 1 2 3 4 11 12 13 14  /for each in list
("1st"
 "2nd"
 "3rd"
 "4th"
 "11th"
 "12th"
 "13th"
 "14th")

Javascript, 68 71

function o(n){return n+([,'st','nd','rd'][~~(n/10%10)-1?n%10:0]||'th')}

Joint effort with ItsCosmo.

EDIT: Wasn't working properly with numbers > 100

OCaml

I'm pretty new to OCaml, but this is the shortest i could get.

let n x =
   let v = x mod 100 and string_v = string_of_int x in
   let z = v mod 10 in
   if v=11 || v=12 || v=13 then string_v^"th" 
   else if v = 1 || z = 1 then string_v^"st" else if v = 2 || z = 2 then string_v^"nd" else if v = 3 || z = 3 then string_v^"rd" else string_v^"th";;

I created a function n that takes a number as a parameter and does the work. Its long but thought it'd be great to have a functional example.

C: 95 characters

A ridiculously long solution:

n;main(){scanf("%d",&n);printf("%d%s",n,n/10%10-1&&(n=n%10)<4&&n?n>2?"rd":n<2?"st":"nd":"th");}

It needs to be mangled more.

Scala 86

def x(n:Int)=n+{if(n%100/10==1)"th"else(("thstndrd"+"th"*6).sliding(2,2).toSeq(n%10))}

Scala 102:

def x(n:Int)=if((n%100)/10==1)n+"th"else if(n%10<4)n+("thstndrd".take(n+1)%5*2.drop(n%5*2))else n+"th"

102 as well:

def x(n:Int)=if((n%100)/10==1)n+"th"else if(n%10<4)n+("thstndrd".sliding(2,2).toSeq(n%10))else n+"th"

ungolfed:

def x (n: Int) =
  n + { if (((n % 100) / 10) == 1) "th" 
        else (("thstndrd"  + ("th"  * 6)).sliding (2, 2).toSeq (n % 10))
      }

PHP, 151

I know that this program is not comparable to the others. Just felt like giving a solution.

<?$s=explode(' ',trim(fgets(STDIN)));foreach($s as$n){echo$n;echo(int)(($n%100)/10)==1?'th':($n%10==1?'st':($n%10==2?'nd':($n%10==3?'rd':'th')))."\n";}

Golfscript, 34 characters

~.10/10%1=!1$10%*.4<*'thstndrd'2/=

PowerShell, 92

process{"$_$(switch -r($_){"(?<!1)1$"{'st'}"(?<!1)2$"{'nd'}"(?<!1)3$"{'rd'}default{'th'}})"}

Works with one number per line of input. Input is given through the pipeline. Making it work for only a single number doesn't reduce the size.

Python 2.7, 137 chars

def b(n):
 for e,o in[[i,'th']for i in['11','12','13']+list('4567890')]+[['2','nd'],['1','st'],['3','rd']]:
  if n.endswith(e):return n+o

n should be a string

I know I'm beaten by the competition here already, but I thought I'd provide my idea anyways

this just basically generates a list of key, value pairs with number (as a string) ending e and the ordinal o. It attempts to match 'th' first (hence why I didn't use a dictionary), so that it won't accidentally return 'st', for example, when it should be 'th'. This will work for any positive integer

Python, 68 characters

i=input()
k=i%10
print"%d%s"%(i,"tsnrhtdd"[(i/10%10!=1)*(k<4)*k::4])

Perl, 37 + 1 characters

s/1?\d\b/$&.((0,st,nd,rd)[$&]||th)/eg

This is a regexp substitution that appends the appropriate ordinal suffix to any numbers in $_ that are not already followed by a letter. To apply it to file input, use the p command line switch, like this:

perl -pe 's/1?\d\b/$&.((0,st,nd,rd)[$&]||th)/eg'

This is a complete Perl program that reads input from stdin and writes the processed output to stdout. The actual code is 37 chars long, but the p switch counts as one extra character.

Sample input:

This is the 1 line of the sample input...
...and this is the 2.
1 2 3 4 5 6 7 8 9 10
11 12 13 14 15 16 17 18 19 20
21 22 23 24 25 26 27 28 29 30
101 102 103 104 105 106 107 108 109 110

Output:

This is the 1st line of the sample input...
...and this is the 2nd.
1st 2nd 3rd 4th 5th 6th 7th 8th 9th 10th
11th 12th 13th 14th 15th 16th 17th 18th 19th 20th
21st 22nd 23rd 24th 25th 26th 27th 28th 29th 30th
101st 102nd 103rd 104th 105th 106th 107th 108th 109th 110th

Numbers already followed by letters will be ignored, so feeding the output again through the filter won't change it. Spaces, commas and periods between numbers are not treated specially, so they're assumed to separate numbers like any other punctuation. Thus, e.g. 3.14159 becomes 3rd.14159th.

How does it work?

• First, this is a global regexp replacement (s///g). The regexp being matched is 1?\d\b, where \d matches any digit and \b is a zero-width assertion matching the boundary between an alphanumeric and a non-alphanumeric character. Thus, 1?\d\b matches the last digit of any number, plus the previous digit if it happens to be 1.

• In the substitution, which is evaluated as Perl code due to the /e switch, we take the matched string segment ($&) and append (.) to it the suffix obtained by using $& itself as an integer index to the list (0,st,nd,rd); if this suffix is zero or undefined (i.e. when $& is zero or greater than three), the || operator replaces it with th.

Edit: If the input is restricted to a single integer, then this 35 character solution will suffice:

s/1?\d$/$&.((0,st,nd,rd)[$&]||th)/e

Ruby, 60

It's not as good as the Perl entry, but I figured I'd work on my Ruby skills.

def o(n)n.to_s+%w{th st nd rd}[n/10%10==1||n%10>3?0:n%10]end

Function takes one integer argument, n, and returns a string as the ordinal form.

Works according to the following logic:
If the tens digit is a 1 or the ones digit is greater than 3 use the suffix 'th'; otherwise find the suffix from the array ['th', 'st', 'nd', 'rd'] using the final digit as the index.