# Write a program that downloads itself

Write a program that connects to this site, downloads the very answer in which it is posted, extracts its own source code and prints it out. The output must be identical to the source code. Shortest code (in bytes) wins.

Rules:

• No URL shorteners allowed.
• The answer must have a regular format - a heading with the language name and size, optional description, code block, optional description and explanation. No unnatural delimiters allowed.
• The output must originate from the actual code block posted on the site.
• The functionality must not depend on the position in the answer list; it should work even if there are multiple pages and the answer it not on the first one.
• New: special note for answers that are supposed to be run in a browser: it is ok to require running them on the codegolf domain (to obey the same-origin policy) but the domain and path should be included in the solution in order to make it fair.
• Catch-22: How am I supposed to test my submission? – Martin Ender May 17 '14 at 18:11
• I forsee people posting answers and deleting them, so they can test their code. – Justin May 17 '14 at 18:18
• @m.buettner answers can be tested on other answers (to other questions) first, then posted, then edited to change the URL :) – aditsu May 17 '14 at 18:20
• @hexafraction if comments are able to interfere with an answer, then the answer is not very good... – aditsu May 17 '14 at 23:41
• A question stuck in my head: How to write a tweet that links to itself without using any URL shorteners, but by estimating the tweet id your tweet? – Ming-Tang May 18 '14 at 5:35

# Bash + coreutils + Lynx browser, 61 bytes

Thanks to @FDinoff for the tips:

lynx -dump codegolf.stackexchange.com/posts/28164/body|grep 2

• And what happens if I type that magic word grep is looking for? – Shade May 17 '14 at 22:36
• lynx lynx lynx lynx. This comment will be grepped out (and the heading as well) – hexafraction May 17 '14 at 22:38
• @hexafraction Awww. You had to go and ruin it! – Shade May 17 '14 at 22:40
• This url should work. codegolf.stackexchange.com/posts/28164/body And it ignores comments. I also think its within the rules that you can use it... – FDinoff May 18 '14 at 1:31
• @DigitalTrauma awww... damn. – haneefmubarak May 18 '14 at 14:50

# Ruby, 155186195148138110 97 characters

require'open-uri';puts open('http://codegolf.stackexchange.com/posts/28159/body').read[/req.+;/];


I had to make it one line, because otherwise it would output newlines as \n instead of actual newlines.

• +31 characters because I didn't notice some characters were being escaped.
• +9 characters to get rid of the annoying backslash.
• Thanks to Nathan Osman for saving 2 chars, and Ventero for saving 55 (!!!) by removing the need for most of the fixes listed above.

## The explanation

Let's beautify this a bit first. However, I'm going to have to use a somewhat... interesting notation in this code. I can't use semicolons at all in this post, for reasons explained later, so I will use {SEMI} in place of semicolons instead.

require 'open-uri'
resp = open('http://codegolf.stackexchange.com/posts/28159/body').read
puts resp.match(/req.+{SEMI}/){SEMI}


Alright, now let's walk through this. The first two lines are fairly self-explanatory -- they fetch the HTML text of this answer.

Now, the last line is the interesting one here. You see that seemingly useless semicolon at the end of the code? It's absolutely required, and here's why.

First, resp.match extracts the code to be printed. The regexp it uses for this is the trick: /req.+{SEMI}/. It grabs the start of the code, REQuire'net/http', by searching for req (re would grab my REputation). Then, it finds the end of the code by searching for a semicolon! Since + is greedy by default, it will keep going until it finds the semicolon that signifies the end of the code. See why I can't use semicolons anymore?

After that, I don't have to unescape anything thanks to Ventero's fix of not using \ at all anymore. All I have to do is fix {AMPERSAND} changing into {AMPERSAND}amp{SEMI}, which can be achieved simply by removing the amp{SEMI} part. No need for this anymore because of new URL. After that, the original code has been retrieved! (Note: I can't use the ampersand either, because that gets HTML-encoded which causes a semicolon to be created.)

• Some characters are getting escaped.. – aditsu May 17 '14 at 19:55
• @aditsu Gah; didn't notice that. Fixed. – Doorknob May 17 '14 at 19:57
• You're gonna hate this.. a backslash is getting duplicated. There's also a newline difference, but that's a minor thing. – aditsu May 17 '14 at 20:00
• @aditsu Argh! :P Fixed also. The newline thing is because of puts; it could be fixed with print but meh. Just pretend there's a trailing newline in the code, even though SE won't be able to show it. – Doorknob May 17 '14 at 20:05
• For the link, http://codegolf.stackexchange.com/a/28159 would give the same result as yours and would save some chars. – Mhmd May 19 '14 at 11:21

## PowerShell - 69 62

(irm codegolf.stackexchange.com/posts/28236/body).div.pre.code

• DOM in a shell. Nice! – bfred.it May 19 '14 at 20:49
• Doesn't irm require Azure Rights Management? Without that module, I think you could do it with Invoke-WebRequest. – Scott Leadley May 19 '14 at 23:29
• @ScottLeadley irm is the alias for Invoke-RestMethod and was introduced with PowerShell v3 core. computerperformance.co.uk/powershell/powershell3-alias.htm – Rynant May 19 '14 at 23:48
• Holy crap. A PowerShell code golf answer with a length in the same order of magnitude as the leading answers. +1 – Adam Maras May 20 '14 at 6:01
• @AdamMaras Ha, I know what you mean! It does happen occasionally though. codegolf.stackexchange.com/a/26811/4565 and codegolf.stackexchange.com/a/21982/4565 weren't too far from the lead. – Rynant May 20 '14 at 13:31

# JavaScript - 1231221019592918786 114

with(new XMLHttpRequest)send(open(0,/\codegolf.stackexchange.com\posts\28175\body/,0)),alert(/w.*/.exec(response))


Runs in the console of your web browser on this page. Tested on the latest Chrome and Firefox.

edit: +28 bytes to add the full domain.

Firefox doesn't like my Regex URL trick anymore with this update :(

Here's the rule-breaking 86 byte solution:

with(new XMLHttpRequest)send(open(0,/posts\28175\body/,0)),alert(/w.*/.exec(response))

• That made me awe. Multiple times. – bfred.it May 19 '14 at 2:46
• @bfred.it I just cut a byte using an interesting regex. I hope it makes you awe one more time. – nderscore May 19 '14 at 6:59
• If printing to the console is an acceptable method of output, you can shorten by 7 chars by removing the alert. – Tejas Kale May 19 '14 at 12:29
• Also , according to the new rule you gotta add codegolf.stackexchange.com/ to the url. – Tejas Kale May 19 '14 at 12:34
• @TejasKale From what I've seen people frown upon solutions that don't actually alert/document.write/console.log the answer. – nderscore May 19 '14 at 12:44

# Ruby + wget + gunzip, 1598682 71

Using tip of @FDinoff to use http://codegolf.stackexchange.com/posts/28173/body.

puts wget -qO- codegolf.stackexchange.com/posts/28173/body[/pu.*\]/]


Tested. Thanks to @ace and @Bob for command line optimization.

• You can combine the flags in wget, as in wget -qO- url. Also, in bash you do not need the double quotes for the url, so this may also work for you. – ace May 18 '14 at 10:46
• You can leave out the http://. – Bob May 18 '14 at 14:44

# CJam - 53

"codegolf.stackexchange.com/posts/28184/body"g54/1=);


I'm making this community wiki since I'm answering my own question and don't really want to compete :p
Credits to FDinoff for the URL choice.

• Woot, +1 for smiley face in code – Cruncher May 20 '14 at 14:18
• @Cruncher ); doesn't look too smiley to me... – MD XF May 19 '17 at 22:50

# Rebmu, 91 characters

Due to the Catch-22 I have to post to get this answer's URL. :-/ Okay, got it.

paTSrd http://codegolf.stackexchange.com/a/28154[th<a name="28154">th<code>cpCto</code>]prC


Rebmu is a dialect of Rebol, and you can read all 'bout it. The equivalent Rebol here would be:

parse to-string read http://codegolf.stackexchange.com/a/28154 [
thru <a name="28154">
thru <code>
copy c to </code>
]
print c


Rebol's PARSE is a sort of highly-literate answer to RegEx. It starts a parser position of the input (which can be any series, including structural blocks...binary data...or string types). The rules are a language for how the parse position moves.

Tags and URLs are really just strings under the hood in the language. But they are "flavored", and as Rebol is dynamically typed you can check that type. READ for instance knows that if you give it a URL-flavored string, then it should dispatch to a scheme handler to do the reading. (In this case, the one registered for HTTP). You get back UTF-8 bytes by default, so we use to-string to decode that and get a series of codepoints in a normal Unicode string.

In the case of the parse dialect, encountering a tag type is just matched as if it were a string that looked like the tag. THRU is an instruction meaning "skip until the ensuing rule is matched, and then place the match position at the end of what you just matched." (TO is the analogue that matches, but leaves the parse position before the element).

So we zip along past the <a name="28154">. Then we zip past the next occurrence of <code>, with our parse position now located right after the >. PARSE's COPY command then lets us copy data up to another rule, in this case that rule is [TO </code>]... so we get into the variable C everything up until right before that <.

Cool, huh? :-)

Technically I could shave more off it, for instance by seeking TO "</" and that saves three characters--there's no need to match the whole </code> end tag when just </ would do. Similar arguments could me made for the start tag. But Rebmu is about literate golfing...even if you might think it looks odd at first!

UPDATE: the /body trick is out of the bag, but I'm similarly going to leave it as-is...because I think it is more educational this way.

## Java now 634, 852, was 1004

Code has been updated; thanks for suggestions. Golfed: now replaces &gt with >

//bacchus
package golf;
import java.net.*;
import java.util.*;
public class G{
public static void main(String[] a) throws Exception {
Scanner z;
URL u;
int x=0;
String s;
u=new URL("http://codegolf.stackexchange.com/questions/28154/write-a-program-that-downloads-itself");
z=new Scanner(u.openConnection().getInputStream());
z.useDelimiter("\\s*//bacchus\\s*");
while(z.hasNext())
{
s=z.next();
s=s.replace("&gt;", ">");
if(x>0)System.out.println("//bacchus\n"+s);
x++;
if(x>2)break;
}
System.out.println("//bacchus\n");
}
}
//bacchus


Submitting for testing, I will edit and try golfing it shortly. Needed to change x>1 to x>2 because test string is also in my code. Note: Code golf replaces > symbol to &gt.

//bacchus
package golf;
import java.io.BufferedReader;
import java.io.IOException;
import java.io.InputStream;
import java.io.InputStreamReader;
import java.net.MalformedURLException;
import java.net.URL;
import java.net.URLConnection;

public class Golf {

public static void main(String[] args) throws IOException {
URL u;
URLConnection c;
InputStream i;
InputStreamReader r;
BufferedReader b;
String s;
int x=0;
try {
u=new URL("http://codegolf.stackexchange.com/questions/28154/write-a-program-that-downloads-itself");
c=u.openConnection();
i=c.getInputStream();
r=new InputStreamReader(i);
b=new BufferedReader(r);
while((s=b.readLine())!=null)
{
if(s.contains("//bacchus")) x++;
if(x>0)System.out.println(s);
if(x>2) break;
}
i.close();
b.close();
} catch (MalformedURLException ex) {

}
}

}
//bacchus

• How do you handle comments containing //bacchus? – hexafraction May 17 '14 at 23:51
• You could inline a lot of stuff, try-with-resources, and use * imports to save a lot of code. – Simon Kuang May 18 '14 at 7:23
• @SimonKuang - I'd also just leave the streams open rather than closing stuff. Also, throws Exception rather than trying to handle anything. Also, I think using a Scanner rather than a BufferedReader would be simpler, especially as you could set the delimiter to //bacchus, which would make things somewhat easier... – Jules May 18 '14 at 16:00

# Python, 175 167 bytes

This uses two external libraries; I didn't read that it was unauthorized.

import bs4,requests
print(bs4.BeautifulSoup(requests.get('http://codegolf.stackexchange.com/q/28154').text).select('#answer-28171')[0].select('pre > code')[0].string)


Longer, but nicer looking code:

import bs4, requests
request = requests.get('http://codegolf.stackexchange.com/q/28154')
soup = bs4.BeautifulSoup(request.text)
answer = soup.select('#answer-28171')[0]
code = answer.select('pre > code')[1].string
print(code)

• The questions in the url can be replaced with q: http://codegolf.stackexchange.com/q/28154 – Justin May 18 '14 at 0:37
• The space in bs4, requests (line 1) can be removed to reduce 1 byte. – ace May 18 '14 at 10:48

## JavaScript, 228

r=new XMLHttpRequest()
c='code'
r.open('GET','//'+c+'golf.stackexchange.com/posts/28157/body')
r.onreadystatechange=function(){this.readyState==4&&alert((a=r.responseText).substr(i=a.indexOf(c)+5,a.indexOf('/'+c)-i-1))}
r.send()


Runs on this page.

• How do you run it? – aditsu May 17 '14 at 19:05
• @aditsu It is supposed to be run on the JavaScript console of a browser. But I am still testing (and fixing) it, please wait – ace May 17 '14 at 19:09
• @aditsu It should work now. Open your browser console (press F12) and paste this code there. – ace May 17 '14 at 19:27
• you sir, need a if(this.readyState == this.DONE) inside the function. – Fabricio May 17 '14 at 20:07
• @ace I see :) I didn't saw the other js answer until now. Then take this upvote from me – C5H8NNaO4 May 20 '14 at 9:49

## Haskell, 563 613 bytes

import Control.Monad
import Data.List
import Network.HTTP
m%f=join(fmap f m)
q s=(simpleHTTP(getRequest"http://codegolf.stackexchange.com/questions/28154/write-a-program-that-downloads-itself?answertab=oldest#tab-top"))%getResponseBody%(putStrLn.head.filter((==)(s++show s)).map(take 613).tails)
main=q"import Control.Monad\nimport Data.List\nimport Network.HTTP\nm%f=join(fmap f m)\nq s=(simpleHTTP(getRequest\"http://codegolf.stackexchange.com/questions/28154/write-a-program-that-downloads-itself?answertab=oldest#tab-top\"))%getResponseBody%(putStrLn.head.filter((==)(s++show s)).map(take 613).tails)\nmain=q"


Tested. Has page support via "oldest posts" feature. Uses quine-line structure to find what to print. The import Control.Monad is only because >>= generates &gt; in HTML.

# Javascript +jQuery, 87, 67

I'm not sure wether I'm allowed to use jQuery, but:

$('body').load('//codegolf.stackexchange.com/posts/28268/body pre')  ### Javascript + jQuery, if excecuted in this page: 27, 25 For fun, if it would be excecuted here: $('[id$=268] pre').html()  $('[id$=28268] pre').html() • This outputs more than the source code. – nderscore May 20 '14 at 16:05 • 67: $('body').load('//codegolf.stackexchange.com/posts/28268/body pre') – nderscore May 20 '14 at 16:07
• You are correct, I wrongly asumed the whole reply instead of the code – Martijn May 20 '14 at 19:09

## w3m 45 characters

w3m codegolf.stackexchange.com/a/28336|grep ☻

• ☺. Does it still work? Edit: Looks like it does. – ace May 21 '14 at 12:04
• @ace I get your comment when I run it – FDinoff May 22 '14 at 2:26
• Fixed :) Anyone else? 😁 – gnibbler May 22 '14 at 2:42

# Dart, 164

I thought I'd try this in Dart, is pretty fun to use imo.

This can be run in the console in DartEditor, but does require the http package added in pubspec.yaml

import"package:http/http.dart"as h;h.read("http://codegolf.stackexchange.com/posts/28215/body").then((s){print(new RegExp(r"im.+(?:})").firstMatch(s).group(0));});}


Ungolfed version:

import "package:http/http.dart" as h;

void main()
{
h.read("http://codegolf.stackexchange.com/posts/28215/body").then((s)
{
print(new RegExp(r"im.+(?:})").firstMatch(s).group(0));
});
}


## R 114 characters

library(XML);cat(xpathSApply(xmlParse("http://codegolf.stackexchange.com/posts/28216/body"),'//code',xmlValue)[1])


No real magic here: it takes the value of the field between the html tags <code></code>. Uses library XML (as one can see in the code quite obviously). Outputs the result as stdout.

# Java, 300 294

import java.net.*;import java.util.*;public class G{public static void main (String [] a) throws Exception{Scanner s=new Scanner(new URL("http://codegolf.stackexchange.com/posts/28189/body").openConnection().getInputStream()).useDelimiter("./?[c]ode\\W");s.next();System.out.print(s.next());}}


An improved version of bacchusbeale's answer which:

• doesn't close resources unnecessarily
• doesn't declare unnecessary variables
• uses a Scanner to avoid having to loop over the input
• uses a regexp that doesn't match itself to avoid having to skip over a middle occurrence of the start/end marker.

Updated:

• Use a direct URL to the post, so we don't need a unique comment to identify the start/end of the code; now uses <code>[...]</code> as the delimiters to search for (actually using the regular expression "./?[c]ode\W", so as to avoid having to decode &lt; and &gt; -- the "\W" is necessary rather than the shorter "." to avoid it matching part of the URL to the post, unfortunately, which costs 2 characters, and the square brackets around c prevent the regex matching itself).
• You have a bunch of unnecessary spaces. Also your class doesn't need to be public. – aditsu May 19 '14 at 8:54
• openConnection().getInputStream() can also be shortened to openStream() – aditsu May 19 '14 at 9:00

# w3m 55 bytes

w3m codegolf.stackexchange.com/posts/28242/body|grep x


Based on @DigitalTrauma

# Ruby, 237215146 132

require'mechanize'
a=Mechanize.new
puts a.get('http://codegolf.stackexchange.com/a/28159').search('.lang-rb code:nth-child(1)').text

• Pretty sure you can remove a few spaces here and there to save a few bytes. – MisterBla May 19 '14 at 10:48
• @richard who cares I'm not going to win anyway. – Mhmd May 19 '14 at 10:52
• Do it for the lols, not for winning. – MisterBla May 19 '14 at 10:55
• @RichardA done, and I also removed few chars from the regexp. – Mhmd May 19 '14 at 11:29

# Processing, 90

print(loadStrings("http://codegolf.stackexchange.com/posts/28657/body")[2].substring(11));


Edit: Finally got it!

# bash + awk, 71 bytes

curl -sL codegolf.stackexchange.com/q/28154 |awk -F\> '/\#/ {print \$3}'

• Doesn't seem to work - it outputs a lot of other stuff along with this answer. – Riking May 18 '14 at 9:53
• @Riking true, it seems to be position-dependent (breaking the last rule) – aditsu May 18 '14 at 10:01
• You can leave out the http://. – Bob May 18 '14 at 14:45
• @user155406: note that each answer has a URL - this one is codegolf.stackexchange.com/a/28179/14710 – Phil H May 19 '14 at 15:26

# Javascript, 138

a=window.open("http://codegolf.stackexchange.com/posts/28160/body");setTimeout('alert(a.document.body.innerHTML.match(/a=.*9\\)/)[0])',99)


This works assuming that the page loads in under 99 ms. It also has to be run via a console opened on a codegolf.SE page, because of the same origin policy.

• Just a note: you don't need the slug in the URL, and questions can be replaced by q. – Schism May 17 '14 at 21:59
• Note that you could do http://codegolf.stackexchange.com/a/28160 instead of http://codegolf.stackexchange.com/a/28160/12551 – Justin May 18 '14 at 0:36
• Chrome doesn't like this: "Uncaught TypeError: Cannot read property 'document' of undefined" – Spedwards May 19 '14 at 7:00
• @Spedwards you should disable the popup blocker. – nderscore May 19 '14 at 15:55

## Perl 5.10, 155127122 117 bytes

use XML::LibXML;say XML::LibXML->new->parse_file('http://codegolf.stackexchange.com/posts/28330/body')->find('//pre')


Using XML::LibXML.

## Shell and xmllint, 82 bytes

xmllint --xpath 'string(//pre)' http://codegolf.stackexchange.com/posts/28333/body


# Python, 164

Works by extracting the text between the code tags. It is quite long but it will always function correctly unless the html page is edited directly or a new code block is added before the one below (having a block of code after should have no effect on the output of the program).

import urllib2
print urllib2.urlopen("http://codegolf.stackexchange.com/posts/28617/body").read().split(chr(60)+"code"+chr(62))[1].split(chr(60)+"/code"+chr(62))[0]


## protected by TimtechMay 20 '14 at 21:31

Thank you for your interest in this question. Because it has attracted low-quality or spam answers that had to be removed, posting an answer now requires 10 reputation on this site (the association bonus does not count).

Would you like to answer one of these unanswered questions instead?