# A program that outputs itself

Gold Member
Is it possible to write a program (in any language) so that the output is the code of the program itself?

Borek
Mentor
10 LIST

On ZX Spectrum

Last edited:
OmCheeto
Gold Member
Is it possible to write a program (in any language) so that the output is the code of the program itself?
Yes.

I wrote a mini-compiler in interpreted basic about 30 years ago who's output was a machine level code that would perform the exact same function.

Of course the machine level program was about 10,000 times faster.

Though "10 LIST" is much simpler and doesn't take 10 hours to write. :tongue2:

CRGreathouse
Homework Helper
Such a program is called a "quine" (presumably after Willard Quine).

Hurkyl
Staff Emeritus
Gold Member
A formulaic way to do it is:

Choose a specific function f so that f(S) is a program that outputs the string S.

Let A be a program that inputs a string S, prints it, then prints f(S).

Now, run the program:
1. f(A)
2. A​

Tweak the details as appropriate for your programming language. An important variation on this method proves the important theoretical result that we can assume any program has access to its own source code.

matt grime
Homework Helper
#!/usr/bin/perl
while(<>){
print;
}

save in a file called "program", then do ./program program. Actually, it's entirely possible to do that without needing to pass the name as an argument to perl (this is stored in the variable $0, I think) so something like #!/usr/bin/perl open($_,"<$0"); while(<>){ print; } close($_);

should do it, but I've no intention of trying it to find out if I got the right interpolation in the string"<$0". in c++ (or any language which produces a separate compiled file) you can do it easily.. write a program which reads a text file with name same as that of itself.. int main(int argc, char* argv[]) { string filename = argv[0]; readFile(filename); } it ll work coz the executable and the source code are in different files.. dunno how u ll get it work for a programme spanning multiple files.. Borek Mentor The idea (although not listed in the OP) is that the program doesn't accept any input nor doesn't read external files. Besides, you are making several faulty assumptions. int main(int argc, char* argv[]) { string filename = argv[0]; readFile(filename); } It will not work - first of all, executable and the source files are different, second, files don't have to be in the same directory. This approach may work for interpreted languages. In fact LIST command does just that, although it doesn't need path nor filename. The idea (although not listed in the OP) is that the program doesn't accept any input nor doesn't read external files. Besides, you are making several faulty assumptions. It will not work - first of all, executable and the source files are different, second, files don't have to be in the same directory. it will work if u have only 1 source code file(*.cpp) and keep the exe file in the same directory. of course, u ll have to get the filename correctly. ya i know, i made several assumptions. however, in the OP when he says code of the program, does he mean the machine code? i d be very interested to know of a way to do that in c++.. Borek Mentor it will work if u have only 1 source code file(*.cpp) and keep the exe file in the same directory. of course, u ll have to get the filename correctly. Not only exe must be in the same directory, but both exe and cpp files have to share the same name. It won't work. I was going to recommend the Turing award acceptance speech of Ken Thompson, "Reflections on trusting trust", http://cm.bell-labs.com/who/ken/trust.html [Broken] but I see it's one of the links in the Wikipedia page cited above. So I did it anyway. :) Last edited by a moderator: I would do it like this in C++: Code: #include <iostream> // Standard header int main() { std::cout << "Hello, world!\n"; std::cout << "#include <iostream> // Standard header\n"; std::cout << "\n"; std::cout << "int main()\n"; std::cout << "{\n"; std::cout << "\n"; std::cout << "std::cout \<\< \"Hello, world!\\n\";\n"; std::cout << "}\n"; } but I just realised that now the parts that print out the code, would also be needed to be inlcuded, but in doing so would add more lines representing those lines, in effect it would be an infinate regress. I can't see how this would be solved? EDIT::: D'oh I just realised the solution is the method one poster above has done, just read a file and print it to screen, but just make sure that file is code itself! Last edited: Hurkyl Staff Emeritus Science Advisor Gold Member D'oh I just realised the solution is the method one poster above has done, just read a file and print it to screen, but just make sure that file is code itself! That's not a program producing it's source code: that's a program asking an external process for some text, and then displaying it. Reading in a file is a boring semi-solution to a clever challenge. The solutions which are discussed and linked to on the wikipedia page exploit much more clever mechanisms to avoid the infinite regress that Anhar found above. Ben Niehoff Science Advisor Gold Member I used to have a quine I wrote in C, but I lost it. This should do it in perl, without reading in a file: Code: #!/usr/bin/perl use strict; sub quote;$string = <<"HERE";
#!/usr/bin/perl
use strict;

sub quote;

\$string = <<"HERE"; placeholder \$quoted = quote(\$string); \$string ~= s/^placeholder\$/\$quoted/m;

print \$string; sub quote { my \$s = shift;
\$s ~= s/\\\$/\\\\\\\$/sg; \$s ~= s/\\\\/\\\\\\\\/sg;
\$s .= "\\nHERE"; return \$s;
}
HERE

$quoted = quote($string);
$string ~= s/^placeholder$/$quoted/m; print$string;

sub quote {
my $s = shift;$s ~= s/\$/\\\$/sg;
$s ~= s/\\/\\\\/sg;$s .= "\nHERE";
return \$s;
}