Python (52 bytes)

print sum(ord(c) for c in 'Happy new year to you!')

Ruby, 15

p Time.new.year

Temporary ;)

Jan Dvorak offers a great alternative in the comments:

Happy = Time
Happy.new.year

But it's so unenthusiastic. I prefer:

Happy = Time
class Time; alias year! year; end

Happy.new.year!

Or even:

class Have; def self.a; A.new; end; end
class A; def happy; Time; end; end
class Time; alias year! year; end

Have.a.happy.new.year!

And here's correct English punctuation:

def noop x = nil; end
alias a noop
alias happy noop
alias new noop
alias year! noop
def Have x
    p Time.new.year
end

Have a happy new year!

Okay okay, I couldn't help it:

def noop x = nil; end
eval %w[we wish you a merry christmas! christmas and a happy new].map{|x|"alias #{x} noop"}*"\n"
def year!; p Time.new.year; end

we wish you a merry christmas! we wish you a merry christmas!
we wish you a merry christmas and a happy new year!

Matlab (4 characters, 5 bytes)

The trick is to cast a string composed only of the character ߞ (of UTF-8 code point U+07DE, which corresponds to 2014 in decimal) to an integer, using the + operator:

+'ߞ'

Byte-count details:

• + is ASCII and counts for 1 byte
• ' is ASCII and counts for 1 byte (but appears twice in the expression)
• ߞ is a 2-byte UTF-8 character

Total: 5 bytes

TeX (32 26 characters, as many bytes)

\def~{\the\catcode`}~}~\\~\%\bye

An even shorter alternative (proposed by Joseph Wright) is

\number`^^T\number`^^N\bye

XeTeX/LuaTeX (13 characters, 14 bytes)

If XeTeX or LuaTeX are allowed, UTF-8 input can be used directly (as proposed by Joseph Wright):

\number`ߞ\bye

Python

newYear = '''
   _ _      _ _      /|     |     | 
  /   \    /   \    / |     |     | 
 |    /   |     |     |     |_ _ _| 
     /    |     |     |           | 
   _/_ _   \_ _/   _ _|_ _        | '''
print(newYear)

Python, 26

print int('bbc',ord("\r"))

Scala REPL

34 29 characters.

+"Happy new year to you!".sum

Well ok if you really want it golfed with any chars, you can use:

'@'*' '-'"'

or

"{yz}"map(_-'I'toChar)

which have 11 and 22 chars respectively.

72 45 characters in the code; Zero character codes

This is far from the shortest answer posted, but no one has yet posted an answer that

• doesn't use character codes as a substitute for numbers, and
• doesn't call the system date.

Using pure math (okay, and an automatic boolean conversion) in R, from the R console:

x<-(T+T);x+floor(exp(pi)^x)*x*x-(x*x)^(x*x)/x

Prints out the number 2014. T is a pre-defined synonym for true in R. The floor and exp functions are directly available in the base package, as is the pi constant. R doesn't have an increment operator, but repeating the (x*x) turned out to be fewer characters that doing increment and decrement twice each.

Original version in Javascript (72 characters)

For the simple reason that I could test out in the console, and it doesn't mind a complete lack of whitespace:

m=Math;p=m.pow;t=true;++t+m.floor(p(m.exp(m.PI),t))*t*t++-p(++t,t--)/--t

run in your console and it will print back the number 2014.

Props to xkcd (and also) for getting me to think about exp(pi):

P.S. If you can make the same algorithm shorter in a different language, post a comment with it.

C

main(){printf("%d",'A'*' '-'B');}

33 characters

PHP (9 bytes)

xxd needed because of binary data (so copying and pasting it would be easier). May return E_NOTICE, but it doesn't really matter, does it?

~ $ xxd -r > 2014.php
0000000: 3c3f 3d7e cdcf cecb 3b                   <?=~....;
~ $ php 2014.php
2014

Ruby, 20

p 'bbc'.to_i ?\r.ord

Explanation: bbc is 2014 in base 13. Shorter than Python. Not as short as Forth.

dc, 6 chars

$ dc <<< "DiBBCp"
2014

D pushes 13 on the stack, even tho the input radix is 10 initially. i changes input radix (to 13) and BBC is 2014 base 13. p prints.

Javascript / Base 64 Conversion

alert(atob("MjAxNA=="))

23 chars (16 with atob by itself)

Or

alert("MMXIV") // ;)

Morse Code 23

;)

..--- ----- .---- ....-

Mathematica, 14 characters (or 15 if you count the bitmap as a character)

TextRecognize@

Perl - 10 characters

This solution is courtesy of BrowserUK on PerlMonks, though I've shaved off some unnecessary punctuation and whitespace from the solution he posted. It's a bitwise "not" on a four character binary string.

say~"ÍÏÎË"

The characters displayed above represent the binary octets cd:cf:ce:cb, and are how they appear in ISO-8859-1 and ISO-8859-15.

Here's the entire script in hex, plus an example running it:

$ hexcat ~/tmp/ten-stroke.pl
73:61:79:7e:22:cd:cf:ce:cb:22
$ perl -M5.010 ~/tmp/ten-stroke.pl
2014

Update: altered to use say instead of print as per @PeterTaylor's comment. Not only does this shave off two further characters, it adds an attractive line break at the end of the output.

print'````'^RPQT

(I initially had the two strings the other way around, which required whitespace between print and RPQT to separate the tokens. @DomHastings pointed out that by switching them around I could save a character.)

GolfScript, 14

'-+,/'{)))))}%

How it works: ASCII goes like this:

...
+
,
-
.
/
0
1
2
3
4
...

So, this takes the ASCII codes of each character, subtracts five, and sticks it in a string.

{...}% yields an array of the characters of a string mapped, when given a string as an argument. So, it increments each character by 5 () means increment).

Powershell, 9

+"ߞ"[""]

ߞ (U+07DE NKO LETTER KA) is counted as two bytes according to the code-golf tag info.

[""] returns the first character from the string ("" is converted to 0). The unary plus opeartor (+) converts the character to an integer.

Yet another GolfScript entry, 9 printable characters

I believe this is shorter than any of the printable GolfScript entries so far:

"!="{*}*)

(Peter Taylor's 7-char entry beats it, but includes non-printable control characters.)

I call this the "that's so last year!" entry, because what it actually does is generate the number 2013 in 8 chars, as 33 × 61, and then increments it by one. ;-)

APL (6 bytes, 4 chars)

⊃⎕TS

Only works this year though.

Why it works:

      ⎕TS
2014 1 1 11 58 5 811
      ⊃⎕TS
2014

Without relying on the system date, it's 10 bytes (7 characters):

⎕UCS'ߞ'

Go - One unicode character (2 chars in utf16, 3 chars in utf8 format), output 2014 as part of an error

—

http://ideone.com/dRgKfk

can't load package: package : 
prog.go:1:1: illegal character U+2014 '—'

Python, 32 chars

print ord(',')*ord('-')+ord('"')

Probably possible to reduce it using the 2014 th Unicode char ߞ, but I didn't try.

Forth, 14

'> '" * '^ - .      

GolfScript (10 8 7 chars)

This solution contains non-printable characters. As xxd output:

0000000: 2714 0e27 7b7d 2f                        '..'{}/

As base 64:

JxQOJ3t9Lw==

Unpacks the ASCII codes for characters 20 and 14 and prints those numbers without any separation.

To actually generate the number 2014, I'm currently at 8 chars:

0000000: 2713 6a27 7b2a 7d2a                      '.j'{*}*

JxNqJ3sqfSo=

Takes a string containing characters with ASCII values 19 and 106 and multiplies them to get 2014.

Haskell, 23 characters, 24 bytes

main=print$fromEnum 'ߞ'

That's U+07DE N'Ko letter KA

Python 32 10 bytes, 91 85 (66 with math imported by default) bytes pure math

Had some fun writing this:

my_lst = []
for i in range(33, 126):
    for j in range(i, 126):
        if 2014 - 126 < i * j < 2014 - 33:
            if j not in range(48, 58):
                my_lst.append("ord('" + unichr(i) +
                              "')*ord('" + unichr(j) +
                              "')+ord('" + unichr(2014 - i * j) + "')")

for val in my_lst:
    print val, '->', eval(val)

Prints all the possible ways I can write 2014 using Bruno Le Floch's method (32 chars):

ord('!')*ord(':')+ord('d') -> 2014
ord('!')*ord(';')+ord('C') -> 2014
ord('!')*ord('<')+ord('"') -> 2014
ord('"')*ord(':')+ord('*') -> 2014
ord(')')*ord('/')+ord('W') -> 2014
ord('*')*ord('-')+ord('|') -> 2014
ord('*')*ord('.')+ord('R') -> 2014
ord('*')*ord('/')+ord('(') -> 2014
ord('+')*ord(',')+ord('z') -> 2014
ord('+')*ord('-')+ord('O') -> 2014
ord('+')*ord('.')+ord('$') -> 2014
ord(',')*ord(',')+ord('N') -> 2014
ord(',')*ord('-')+ord('"') -> 2014

But this is obviously redundant, so if your interpreter is set to utf-8 by default, then all it takes is:

>>> ord(u'ߞ')
2014

Also, thanks to AmeliaBR (for the idea), I tried my best to implement a pure math version:

from math import*
a,b,c=int(e),round(e),ceil(pi);print int(a**(b*c-(c-b))-a*a**c-a)

Mathematica, 23 bytes

Tr@ToCharacterCode@"ߞ"

(Using the 2014 unicode char. Credit for Tr goes to alephalpha)

Mathematica, 46 bytes (pure math)

a=⌊E⌋;b=⌈E⌉;c=⌈Pi⌉;a^(b*c-(c-b))-a*a^c-a

idea taken from here.

Javascript, 18 characters

alert(btoa('ÛMx'))

JSFiddle: http://jsfiddle.net/pBjL2/

D

ϯ has the numerical value 1007 in Unicode encoding.

as compiler message during compilation (20 chars)

pragma(msg,'ϯ'+'ϯ');

runtime version (45 chars)

import std.stdio;void main(){write('ϯ'+'ϯ');}

Golfscript - 11 10 chars

'&~I'{}/-*

computes 38 * 53 (ie an expansion of the prime factorization of 2014: 2 * 19 * 53)

Old Version

'f.'{}/.*\-

This computes 46²-102.

'f.'{}/ puts the ascii values of f and . on the stack.
.*      squares the value of the .
\-      swaps the two values, then subtracts