# triplegolf - not only the source code length counts!

The goal of this challenge is to write a program where the length of the following three are exactly the same:

• the length of the source code
• the length of the output it prints to the standard output
• the length of the compiler warnings it generates with the (unmodded) compiler of your choice

The winner is the shortest code to fulfill all three criteria, and has to be at least 1 character long.

To spice things up, if the output has less than three different characters, it gets a 5 point penalty added to the length for each one (so +10 if only a single type, +5 for two).

(Newlines count either as 1 or 2 characters, your choice, but it has to be consistent through all three criteria. Leading and trailing white-spaces are ignored in all lines in all three criteria)

In case of equal score, the one printing the most interesting result (instead of garbage) is the winner.

• What is a "type of character"? May 17 '12 at 18:36
• Also, what if the compiler produces non-warning output (e.g. "Compiling Main") ? Also, what if the name of the source file appears in warning messages (e.g. "foo.c:1:1: ...") ? May 17 '12 at 18:40
• My weapon of choice is VBA, which doesn't really have a compiler and does not output a binary file to run (as is the case for some other languages, I believe). Are there considerations for this, or am I just disqualified by default? May 17 '12 at 19:17
• I wonder if a triple-quine is possible in some language?
– Ry-
May 18 '12 at 0:00
• For interpreted languages, we get an output even on occurance of an error, in such a case, are those errors accepted or are only warnings accepted? May 18 '12 at 10:51

## Bash, 23 characters

Error:

bash: /: Is a directory


Source:

echo       $0-$01234;/;


Output:

/bin/bash-/bin/bash1234


## Brainf*ck, 32 characters

This code executes for about 3 seconds and stops and displays the following error and output.

Error:

bff: out of memory (-2058691272)


Source:

+++++[......-]..+[>>>>>>>>>>>>-]


Output: (Hexdump)

0505 0505 0505 0404 0404 0404 0303 0303
0303 0202 0202 0202 0101 0101 0101 0000


## C, 35 characters

Warning:

b.c:1:30: warning: division by zero


Source and Output:

main(){system("cat "__FILE__)/0;;;}


## PHP, 50 characters

Warning:

PHP Warning:  Division by zero in /tmp/3 on line 1


Source and Output:

<?php echo (0/0).''.file_get_contents(__FILE__);?>

• The bash example is an error, not a warning. May 18 '12 at 10:34
• What compiler are you using for the C solution? May 18 '12 at 16:10
• @breadbox gcc version 4.7.0 20120505 (prerelease) (GCC) May 18 '12 at 16:33

## C - 48 chars

main(i){while(++i<49)putchar(i);putchar('\z');}


Note: includes a final (Unix-style) newline.

Output from gcc a.c reads:

a.c:1:41: warning: unknown escape sequence '\z'


The output from a.out is mostly non-printing chars, so here's what it looks like after piping through hexdump:

00000000: 0203 0405 0607 0809 0A0B 0C0D 0E0F 1011  ................
00000010: 1213 1415 1617 1819 1A1B 1C1D 1E1F 2021  .............. !
00000020: 2223 2425 2627 2829 2A2B 2C2D 2E2F 307A  "#$%&'()*+,-./0z  • My understanding is that non-printable (whitepace?) are truncated, so this would not count. If that works, I can improve my own answer. May 17 '12 at 19:39 • Control characters aren't usually considered whitespace, except of course for \t ,\n, \r, \f, and sometimes \v. None of these are at the leading or trailing position, so I figured I was safe. May 17 '12 at 20:36 • I think this is perfectly acceptable. They are not whitespace, and in either case this is not ACM to be that strict with the requirements. The main reason for this "whitespace rule" was that some IDEs might heavily format their compiler output. – vsz May 18 '12 at 3:09 • @copy, Or main(i){i='\z';while(i-->74)putchar(i);} -- but reducing the program size is actually counter-productive. May 18 '12 at 16:09 • This is my favorite answer, I know I should have prohibited accessing the contents of the file (just as real quines do) – vsz May 22 '12 at 6:14 # JavaScript, 63 66 !function x(){console.log(x+'...');eval(Array(33).join('$'))}()


The output is:

function x(){console.log(x+'...');eval(Array(33).join('$'))}...  In Chrome, the error is: ReferenceError:  is not defined  # Visual Basic .NET, 185 Gee, vbc is pretty verbose with its compilation warnings. Anyway, the code is this: Public Module Main Public Sub Main() Console.WriteLine(New String("a"c,185)) End Sub Public Function A() 'This is actually just padding. 'Hi End Function End Module  (Note that they're supposed to be tabs, not spaces.) The output is this: aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa  And the compiler warning is this: warning BC42105: Function 'A' doesn't return a value on all code paths. A null reference exception could occur at run time when the result is used. End Function ~~~~~~~~~~~~  (This time, it's actually four spaces, not tabs.) # Zsh, 20 bytes <<<$-$_$_$_$_$_ [  The problem we face here is alluded to in a comment above: Zsh prints the program name in addition to the line number before every error. The solution here uses a one-character filename to abuse this. I've wrapped it up in a helper script to show exactly what it prints, and used wc -c on stderr, stdout, and the source file. Output: 569Xcatcatcatcatcat followed by a newline Errors: s:[:2: ']' expected followed by a newline EDIT: Alternate 20 byte solution with no file restriction: <<<$-$-$-$-$_>&1 >&2


Try it online!

<<<$_$_$_  -x flag enables xtrace. This again requires a single-character filename. Try it online! # Zshzsh -JNTwEDY, 12 bytes <<<$->&1 >&2


# Python, 57 bytes

Other interpreters may display the warning differently. This was only tested to work on TIO.

from warnings import*
print("prt"*19)
warn("Warning...!")


Try it online

Output

prtprtprtprtprtprtprtprtprtprtprtprtprtprtprtprtprtprtprt


Warning

.code.tio:3: UserWarning: Warning...!
warn("Warning...!")


Note that the leading spaces are not counted toward the byte count. If the leading spaces were not ignored, this could be done in 51 bytes.

# Javascript (ES6), 32 30 bytes

Screen(console.log(Function));


prints

ƒ Function() { [native code] }


in Chrome, and then throws

TypeError: Illegal constructor


## My original 32 byte solution:

(x=y=>console.log(x+!0+10)||z)()


first, prints

y=>console.log(x+!0+10)||ztrue10


And throws the error

ReferenceError: z is not defined


## VBA, 39 Bytes

Not sure if this qualifies, given the compiler constraint, but:

Input: (in the immediate window)

For i=1 To 39:a=a & Chr(i):Next:Print a


*The output includes non-printing characters that don't play well in this window.

• VBA does not give any compiler warnings so I do not think that this counts a being valid May 31 '17 at 16:52

# Perl 6, 10 bytes

dd say 1e9


Try it online!

Prints 1000000000 to STDOUT, and Bool::True to STDERR. Both produce a trailing newline, but trailing whitespace is ignored in this challenge. dd is a Rakudo specific debugging function