Second Order FizzBuzz

The FizzBuzz problem requires the construction of a well-known list of interleaved numbers and nonsense words. The list has been studied extensively, and recent work at CERN has constrained its utility to less than 10-16 of a mosquito’s left nut. In modern times, the task of its production is typically relegated to computer programs. It is an open question why some institutions continue to use FizzBuzz as a test of programming ability, given universal knowledge of the test and the wide availability of solutions on Google and StackOverflow.

Programs that solve FizzBuzz are typically constructed by hand in an ad-hoc manner, often by inexperienced computer scientists, and entail a hazard of boredom-induced mortality. Below is presented an algorithmic solution to the second-order problem of generating such programs. In general, we define a class of nth-order FizzBuzz problems which require the production of a program which solves the (n-1)th-order FizzBuzz problem.

We use a variation on the stacksort algorithm due to Munroe (2013) and first implemented by Koberger (2013), which mimics a technique commonly employed in the wild. Here, the availability of solutions on StackOverflow is exploited to solve the second-order FizzBuzz problem in quadratic time[citation needed].

With reasonable probability, StackFizzBuzz finds, downloads, and executes code from questions about FizzBuzz on StackOverflow until a solution is found. Also, with small but nonzero probability, FizzBuzz may root your computer. It is recommended that StackFizzBuzz not ever be run.

The solution on GitHub



# (c) 2015 Tim Babb

import os
import sys
import json
import re
import urllib2      as ul
import zlib         as zl
import gzip         as gz
import StringIO     as sio
import HTMLParser   as html
from htmlentitydefs import name2codepoint

# TODO: This will never find a correct answer if an intermediate test 
#       program loops forever. It should be a trivial matter to write 
#       some routine which detects ahead of time whether the test 
#       program will halt. This is left as an exercise to the reader.

SO_API         = ''
NUMBERSTEXT_RE = re.compile('\d+|\w+', re.MULTILINE)

class FindCode(html.HTMLParser):
    # parse through html finding nodes that look like code blocks.
    # code blocks are simultaneously inside <code> and <pre> tags.
    def __init__(self):
        self._codedepth = 0
        self._predepth  = 0
        self._code      = []
        self._cur_code  = ""
    def in_codeblock(self):
        return self._codedepth > 0 and self._predepth > 0
    def handle_starttag(self, tag, attrs):
        if tag.lower() == 'code':
            self._codedepth += 1
        elif tag.lower() == 'pre':
            self._predepth  += 1
    def handle_endtag(self, tag):
        popped = True
        if tag.lower() == 'code':
            self._codedepth -= 1
        elif tag.lower() == 'pre':
            self._predepth  -= 1
            popped = False
        if popped and len(self._cur_code) > 0:
            self._cur_code = ""
    def handle_data(self, data):
        if self.in_codeblock():
            self._cur_code += data
    def handle_entityref(self, name):
        if self.in_codeblock():
            self._cur_code += unichr(name2codepoint[name])
    def handle_charref(self, name):
        if self.in_codeblock():
            if name.startswith('x'):
                c = unichr(int(name[1:], 16))
                c = unichr(int(name))
            self._cur_code += c
    def get_code(self):
        return self._code

def decode(response):
    """Decode an HTTP response packet."""
    enc ="content-encoding")
    content =
    if enc in ('gzip', 'x-gzip'):
        content = gz.GzipFile('', 'rb', 9, sio.StringIO(content)).read()
    elif enc == 'deflate':
        content = zlib.decompress(content)
    return content

def get_request(req):
    """Make an HTTP request."""
    opener  = ul.build_opener()
    opener.addheaders = [('Accept-Encoding', 'gzip')]
    response =
    data = decode(response)
    return data

def stackoverflow_req(call, params):
    """Make a StackOverflow query."""
    ext = "&".join(['%s=%s' % (ul.quote(k), ul.quote(str(v))) for k,v in params.iteritems()])
    url = SO_API + ( "?".join( (call, ext) ) )
    res = get_request(url)
    return json.loads(res)

def runcode(code, interactive=True):
    """Execute some python code `code` and capture its output, returning it as a string.
    If `code` is not a valid python program, return False. If `interactive` is true,
    present the code and prompt the user before executing."""
    # in case the sample program expects user input/interaction,
    # generate a list of 100 numbers for the program to read.
    s = "\n".join([str(x+1) for x in range(100)])
    numinput   = sio.StringIO(s)
    fizzoutput = sio.StringIO()
    if interactive:
        print "Running the following code:"
        print "---------------------------"
        print code
        print "---------------------------"
        if not raw_input("OK? (y/n): ").startswith('y'): return False
    sys.stdin  = numinput
    sys.stdout = fizzoutput
    success    = True
    ### DANGER DANGER ###
        success = False
    sys.stdout = sys.__stdout__
    sys.stdin  = sys.__stdin__
    if success:
        return False

def verify_fizzbuzz(code, interactive=True):
    """Verify that the python program `code` produces a valid FizzBuzz list.""" 
    data = runcode(code, interactive)
    if not data: return False
    # we are forgiving with separating junk
    # (whitespace, commas, quotes, etc.)
    # pull out the meaty bits and make sure they are correct
    matches = NUMBERSTEXT_RE.findall(data)
    # if we don't get a single "fizzbuzz", then we 
    # don't really have any evidence the program works, do we?
    if len(matches) < 15: return False
    for i,val in enumerate(matches):
        j = i + 1
        if   j % 3 == 0 and j % 5 == 0: expected  = 'fizzbuzz'
        elif j % 3 == 0:                expected  = 'fizz'
        elif j % 5 == 0:                expected  = 'buzz'
        else:                           expected  = str(j)
        if expected != val.lower(): return False
    return True

def second_order_fizzbuzz(interactive=True):
    """Find some code on StackOverflow that solves the FizzBuzz problem."""
    page = 1
    while True:
        # find some questions that match the tag
        results = stackoverflow_req('questions', 
                                    {'sort'     : 'activity', 
                                     'tagged'   : 'fizzbuzz;python',
                                     'page'     : str(page),
                                     'pagesize' : '25',
                                     'order'    : 'desc'})
        if len(results['items']) < 1: return False
        # iterate over the questions
        for q in results['items']:
            # list answers belonging to this question
            answers = stackoverflow_req('questions/%s/answers' % q['question_id'],
                                        {'sort'   : 'votes',
                                         'filter' : 'withbody'})['items']
            for a in answers:
                fc = FindCode()
                # parse HTML looking for nested <pre> and <code>
                code = fc.get_code()
                # check each codeblock
                for c in code:
                    if len(code) > 0 and verify_fizzbuzz(c, interactive):
                        # got a valid fizzbuzz!
                        return c
        page += 1

############# Main program #############

if __name__ == "__main__":
    import argparse
    parser = argparse.ArgumentParser(
                        description="Solve the FizzBuzz problem by looking it up on StackOverflow. "
                       "This program downloads random code from the internet and executes it. "
                       "This is a terrible idea and should not be done ever.")
    parser.add_argument('--interactive', default=False, action="store_true",
                        help='Present code and prompt before executing each candidate')
    parser.add_argument('--yes', default=False, action="store_true",
                        help='Pass this flag to affirm that running this program is a bad idea.')
    parser.add_argument('--order', type=int, default=2,
                        help='Which meta-level of solution is desired?\n'
                        '1 = just print the damn list\n'
                        '2 = produce a program that prints the list\n'
                        '3 = produce a program that prints a program which produces the list')
    args = parser.parse_args()
    if not args.yes:
        print "\nRunning this program is a terrible idea because it downloads " \
              "random code from the internet and executes it. "
        print "Pass --yes to affirm that you understand how dumb it is to do this.\n"
    if args.order >= 3:
        with open(__file__, 'r') as f:
        prog = second_order_fizzbuzz(args.interactive)
        if args.order == 1:
        elif args.order == 2:
            print prog
Share this article:Share on FacebookShare on Google+Tweet about this on TwitterShare on RedditShare on StumbleUponEmail this to someone

3 thoughts on “Second Order FizzBuzz

  1. JTurner

    Absolutely hilarious! “mimics a technique commonly employed in the wild” aka “Google search for a while to see if I can find some code only to give up and copy/paste something from StackExchange”.

    [fizzbuzz, for those unfamiliar, is a simple programming exercise meant to filter out people who claim to be able to write computer code, from those who can. It prints the word fizz or buzz based on divisibility of a number]

    Until now, my favorite fizzbuzz solution was “enterprisey fizzbuzz” which I saw a few years ago. (For those who haven’t seen, it’s an absurdly complex implementation of a fizzbuzz program which should be about ten lines long, which imcorporates, GoF patterns such as factories, facade, etc.

    However this “second order” solution rises above the fray to look at the meta-problem being asked. And the idea of building a watchdog to observe the execution, being left as an “exercise for the reader” is also LOL brilliant!!

    Similar to Nigel Tufnel’s guitar in Spinal Tap (the guiltar which has never been played), it is agreed this code should never be run. It’s just too special.

  2. Andrei

    “write some routine which detects ahead of time whether the test program will halt”

    I wrote the patch: it searches on StackOverflow for a solution. If no solution can be found, it posts a question, sleeps for 5 minutes and then tries again. Hopefully it won’t block until it finds one.

    Where do I create the PR?


Leave a Reply to JTurner Cancel reply

Your email address will not be published. Required fields are marked *