The domain server requires that all employees have a strong, random password conforming to the following rules:

• Exactly 15 characters long.
• Keyboard-typeable characters only (as shown in code-type below). Teaching the sales to use ALT+NUMPAD codes is not permitted.
• At least 1 lower case letter: abcdefghijklmnopqrstuvwxyz
• At least 1 upper case letter: ABCDEFGHIJKLMNOPQRSTUVWXYZ
• At least 1 numeric digit: 0123456789
puts$_  I also used puts instead of p in this one because p prints out the string enclosed in "quotation marks" and some characters escaped with a backslash. # Ruby, 70 bytes As Ventero points out, ~ can be skipped in front of the regexes, and print can replace puts$_. But with the ugly output this causes you might as well print all the rejected passwords too, squeezing it into a one-liner:

• Could you please explain your code a bit? What does .all?{|r|~r} do? What does $_= do? – Martin Thoma Jan 5 '14 at 23:41 • The sample line is clever and all, but I think it violates "It must be able to generate all permutations of all allowable characters." Nowhere does it says that the password may only contain a-d as far as letters are concerned. If z is an allowed character, there should be a chance >0 that z is in the password. – nitro2k01 Jan 6 '14 at 5:07 • Does \W in Ruby include the underscore _? In most regex dialects I know it does not. And if your code could not generate passwords where a single _ was the only non-alphanumeric symbol, then it would violate one requirement. The second approach very obviously violates that requirement, but I guess it wasn't phrased correctly at that time. – MvG Jan 6 '14 at 5:09 • @MvG: You're correct. \W does not contain an underscore in Perl-compatible RegEx (source). – Martin Thoma Jan 6 '14 at 8:42 • Furthermore, your solution is affected by the same problem @moose and I had with Python: sample does not repeat elements, so passwords with repeated elements cannot be generated by your code.Can you fix these two issues to make your answer conform to the question? Seeing how yours is the leading solution, excepting Wolfram Alpha, it would be nice to see whether you can conform and still keep the lead. I guess it shouldn't be too hard. – MvG Jan 6 '14 at 10:26 # Java 8 - 354329319275 267 characters Just for fun, using lambdas with Java 8 - each possible output has the same probability of being found. It uses the fact that the allowed characters have consecutive ascii codes, from 33 to 126. class A { //flags for, respectively, small caps, large caps, digits, punctuation static int a, A, d, p; public static void main(String[] x) { String s; do { //Using special String constructor that takes an int[] s = new String(new java.util.Random().ints(15, 33, 127) .toArray(), 0, 15); a = A = d = p = 0; s.chars() .map(c -> c > 96 & c < 123 ? a = 1 : c > 64 & c < 90 ? A = 1 : c > 47 & c < 58 ? d = 1 : (p = 1)) .min(); } while (a + A + d + p < 4); System.out.println(s); } }  Sample output: .*;Tm?svthiEK3 o.dzMgtW5|Q?ATo FUmVsu<4JF4eB]1  Compressed program: class A{static int a,A,d,p;public static void main(String[]x){String s;do{s=new String(new java.util.Random().ints(15,33,127).toArray(),0,15);a=A=d=p=0;s.chars().map(c->c>96&c<123?a=1:c>64&c<90?A=1:c>47&c<58?d=1:(p=1)).min();}while(a+A+d+p<4);System.out.println(s);}} • How about while(a+A+d+p<4) together with a|=1 instead of a++? Or use bitmasks, i.e. stuff like a|=1 through a|=8, with a<15 as loop condition. This saves another 13 chars if I counted correctly. – MvG Jan 6 '14 at 5:04 • @MvG good point - did something similar, saving a couple extra chars I believe. – assylias Jan 6 '14 at 9:33 • @MvG And using new String(int[],int,int) saves another 40-odd chars! – assylias Jan 6 '14 at 9:41 # Python 2.X + 3.X (229 characters): Generate and replace ## Idea 1. First make a list with 15 allowed symbols 2. Replace a random position r by a random digit 3. Replace a random position s, with s != r, by an upper case letter 4. The same for lower case letter and symbol as in 2 and 3. ## Code from random import randint as r, shuffle as s a=list(range(15)) p=a[:] for i in range(15): a[i]=chr(r(32,126)) s(p) a[p.pop()]=chr(r(48,57)) a[p.pop()]=chr(r(65,90)) a[p.pop()]=chr(r(97,122)) a[p.pop()]=chr(r(33,47)) print(a)  # Python 2.X + 3.X (194 characters): Generate and check import random from re import search as s p='' while not all([s("\d",p),s("[a-z]",p),s("[A-Z]",p),s("[\W_]",p)]): p=str(map(chr,[random.choice(list(range(33,127))) for i in range(15)])) print(p)  • Thanks to MvG who told me that \u and \l does not exist in Python regex. • Thanks to grc who told me that random.sample is without replacement, be to get every possible allowed password we need sampling with replacement. # Using flaw in the problem description Currently, the problem description does not demand that every symbol / digit appears with the same probability. With the following solution, you cannot make any assumption about a single symbol and/or position. But you can do it with multiple ones. ## Python 2.X+ 3.X (62 characters) from random import sample print(sample("0123abcdABCD-+/<",15))  Thanks to daniero for the idea to use sample. • Very smooth finding the flaw! I've plugged that one, but bonus points for identifying it. :-) – Hand-E-Food Jan 5 '14 at 22:59 • Your gen&check is similar to my approach. Out of curiosity:where is this \l and so on for python regexes documented? Don't see it in the reference. My Python 3.3.3 won't even accept "\u". The str(…) does not join the letters in either 3.3.3 or 2.7.6. One suggestion for optmization: all(s("\\"+i,p)for i in "dluW"). – MvG Jan 6 '14 at 4:09 • random.sample chooses elements without replacement, so not all passwords are possible. – grc Jan 6 '14 at 7:15 • @MvG: Thank you. I've just seen that \u and \l is vim-only. – Martin Thoma Jan 6 '14 at 8:36 # Bash on *nix (109) while ! grep -Pq [A-Z].*[a-z].*[0-9].*[\\W_]<<<$a$a$a$a do a=tr -dc !-~</dev/urandom|head -c15 done echo$a


To work correctly, $a must not be set to a valid but non-random password up front. If you want to include a= and a line break up front, that's three more characters but it allows you to run the thing repeatedly. You can obviously also replace all newlines with ; so you have a one-liner which you can execute as often as you whish. Furthermore, you should have set LC_ALL=C or not set any locale-specific environment variables (LANG and LC_CTYPE in particular), since the character ranges depend on collation order being equal to ascii order. /dev/urandom is the a source of random bytes. !-~ is the range of all permissible characters, as specified in the question. tr -dc removes all characters not listed in its next argument. head takes 15 of the remaining characters. grep checks whether each of the required kinds does occur at least once. Its input consists of four copies of the candidate, so order of the symbols does not matter, hence all possible passwords stand a chance of getting selected. The -q to grep suppresses output. For reasons unknown, /dev/random instead of /dev/urandom takes ages. It seems like entropy got exhausted pretty quickly. If you cd into /dev, you can avoid some more bytes, but that feels a bit like cheating. # Python 2 (138) import re,random a='' while not re.search('[A-Z].*[a-z].*[0-9].*[\W_]',a*4): a=''.join(random.sample(map(chr,range(33,127))*15,15)) print a  To make the code readable I added a newline and indentation after the loop which is not neccessary and which I did not count. This is essentially the same idea as in the bash version. The random source here is random.sample, which will not repeat elements. To counter this fact, we use 15 copies of the list of permissible letters. That way, every combination can still occur, although those with repeated letters will occur less often. But I decide to consider this a feature, not a bug, since the question did not require equal probability for all permutations, only the possibility. # Python 3 (145) import re,random a='' while not re.search('[A-Z].*[a-z].*[0-9].*[\W_]',a*4): a=''.join(random.sample(list(map(chr,range(33,127)))*15,15)) print(a)  One newline and one indent again not counted. Apart from some Python-3-specific syntax overhead this is the same solution as for Python 2. # JavaScript (161) a=[];for(i=33;i<127;)a.push(s=String.fromCharCode(i++)); while(!/[A-Z].*[a-z].*[0-9].*[\W_]/.test(s+s+s+s)) for(i=0,s="";i<15;++i)s+=a[Math.random()*94|0];alert(s)  I added the newlines for readability, but did not count them. # R (114) s<-"" while(!grepl("[A-Z].*[a-z].*[0-9].*(\\W|_)",paste(rep(s,4),collapse=""))) s<-intToUtf8(sample(33:126,15,T)) s  Linebreak and indentation inside loop added but not counted. If you feel like it, you can again move this to a single ;-separated line. • Ha! I was just about to point out that you could have used grepl in your R code. If only I'd thought of repeating the test password four times so you could do all the checks in one. And you know, if only I'd thought about sample and intToUtf8. However, you need to add replace=TRUE (or more concisely, you just need to add ,T) to your sample method to make sure you're getting all possible passwords. – AmeliaBR Jan 6 '14 at 6:26 • @AmeliaBR: You are right, fixed that replace=T mistake, thanks for pointing this out. Finding intToUtf8 by guessing likely names with tab completion took me quite a while; I knew such a function has to exist, but the more common names like chr and so on were not used. – MvG Jan 6 '14 at 6:33 • @MvG: I don't understand why your Python code terminates at all. Why do you need that *4? I thought your regex would match any string, that ad first one uppercase letter, then anything, then one lowercase letter, than anything ... what did I get wrong? – Martin Thoma Jan 6 '14 at 8:49 • @moose: As you already noticed, my regex checks for the required categories in a specific order. But by taking the concatenation of four copies of the current candidate, I can ensure that the order no longer matters: Even if my password is symbols followed by digits followed by lowercase followed by uppercase, then there would still be a match. The only way a match can fail is if a category is missing altogether. Also note that I do re.search not re.match, so the regex might match anywhere in the candidate password. Does this explain why it will terminate eventually? – MvG Jan 6 '14 at 8:55 • Ah, I did not notice that you use re.search instead of re.match. That explains it. But I still think you don't need the *4. Thank you for the explanation (+1) – Martin Thoma Jan 6 '14 at 9:09 # C# (123 - 139 103 - 127 characters compacted): Using a perfectly adequate framework method in System.Web.dll: class P { static void Main() { Console.WriteLine(System.Web.Security.Membership.GeneratePassword(15, 1)); } }  Compacted: class P{static void Main() {Console.WriteLine(System.Web.Security.Membership.GeneratePassword(15,1));}}  Example: b+m2ae0K:{dz7:A  Alternatively, take the value of the second parameter (int numberOfNonAlphanumericCharacters) from the command line: class P { static void Main(string[] a) { Console.WriteLine(System.Web.Security.Membership.GeneratePassword(15, int.Parse(a[0]))); } }  • GeneratePassword does not support the full set of symbols specified in the question. Neither did I find any guarantees about the minimum number of occurrences of each character category. – MvG Jan 6 '14 at 8:39 • You could compact further by using class P and string[] a. – d3dave Jan 6 '14 at 14:45 • @MvG, that's interesting. It looks like it excludes any symbol that's commonly used to write accented characters in languages such as French. Probably a smart move. Changing the keyboard language would be enough to stuff up your password. – Hand-E-Food Jan 6 '14 at 21:19 # R (301 322 characters) Correction forgot to check for digits. a='abcdefghijklmnopqrstuvwxyz'; f=as.factor(strsplit(paste(a,toupper(a), sep="0123456789~!@#$%^&*()_+-={}|[]\\:\";'<>?,./"),"")[[1]]);
g=gsub("(.):","\\1",levels(q:q:q:q:q:q:q:q:q:q:q:q:q:q:q));
repeat{p=g[runif(1)*length(g)];
if(grepl("[A-Z]",p)&&grepl("[a-z]",p)&&grepl("[0-9]",p)&&grepl("[^A-Za-z0-9]",p))break;};
print(p);


Generates all possible 15-character permutations of the 94 characters. Then randomly selects one until it matches the criteria.

The magic is in the q:q operation, which generates a new factor data type that is the interaction of all the factors in the first q list with all the factors in the second list, with every possible combination of those two lists being included in the list of "levels" of that factor. Interact 15 copies of the list of allowed characters, and you get (94^15) possible levels.

Please do not try this at home. The code takes a couple seconds to figure out all the three-character permutations, I really can't imagine how long it would take to figure out all the 15-character permutations, if your computer didn't just run out of memory in the meantime. When I ran the finished (three-character password) script to check it, the first password it spit out was "oO=", which I think about sums up the reaction you should have to this code.

• @MvG has an R script that is both much more practical and much shorter, if much less awesome: codegolf.stackexchange.com/a/17700/12413 – AmeliaBR Jan 6 '14 at 6:28
• Nevertheless, I like your idea. Many code-golf snippets I saw let specific features of the language do the hard work. And your code certainly does that for R, with those interactions. – MvG Jan 7 '14 at 13:13

# Mathematica 170

r=RandomSample;f[i_]:=(FromCharacterCode/@Range@@i);
{t,A,a,n}=f/@{{33,126},{65,90},{97,122},{48,57}};
s=Complement[t,A,a,n];
""<>r[Join[RandomChoice/@{A,a,n,s},r[t,11]],15]


Examples

"<]}Pg3/e?3+Z~Oz"
"X/8jWe@f(_x5P:="
"2wz2VQhtJC?*R7^"

## Python 2.7 (182)

import random as r,string as s
z=r.sample
j=list(z(s.ascii_lowercase,12)+z(s.ascii_uppercase,1)+z(s.digits,1)+z('~!@#$%^&*()_+-={}|[]\\:";\'<>?,./',1)) r.shuffle(j) print ''.join(j)  • You can get 9 digits less by removing the join as it is not required by the problem descript. Another 2 less by removing spaces. – Martin Thoma Jan 5 '14 at 22:47 • @moose I took out the spaces right before you commented :-) I feel like the join kinda has to be there: Would users be expected to understand python list syntax from the output: ['q', 'u', ...]? – Jonathon Reinhart Jan 5 '14 at 22:50 • I thought about removing print at all. When size in bytes is important, they might live in the punch card time. That way, they might be able to read the memory ... just by looking at it. Or they are "real programmers": xkcd.com/378 – Martin Thoma Jan 5 '14 at 22:53 • If I'm reading the code correctly, this does not fulfill the all permutations requirement, it will always have 12 lowercase characters, making passwords with more than one of the other groups (like aA$bc1111111111) impossible. – IQAndreas Jan 5 '14 at 23:31
• In Johnathon's defence, I think the permutations rule was added 5 minutes after his post. – Hand-E-Food Jan 5 '14 at 23:36

Golfscript (60)

Since the obl. golfscript is missing and as a noob I need the practice anyway :)

[48 10{rand}:r~+65 26r+97 26r+33 15r+11,{;32 96r+}%~]{r}$''+  It just builds an array with the 4 required + 11 random characters and sorts in random order. • +1 for {r}$. That's a pretty dirty way to shuffle a list — I like it! ;-) – Ilmari Karonen Jan 7 '14 at 2:47
• ...however, I don't think this can ever output e.g. 0Aa~~~~~~~~~~~~. :-( – Ilmari Karonen Jan 7 '14 at 13:43

# JavaScript 258240233 225

R=Math.random;a=b=>b[b.length*R()|0];for(x=[a(l="abcdefghijklmnopqrstuvwxyz"),a(u=l.toUpperCase()),a(n="0123456789"),a(s="~!@#$%^&*()_+-={}|[]\\:\";'<>?,./")];15>x.length;x.push(a(l+u+n+s)));alert(x.sort(y=>.5-R()).join(""))  Using a rule where: function(x){return x*x} can be re-written as function(x)x*x. Only seems to work for functions returning a value. Next iteration, reduced x.sort(function().5-R()) to x.sort(y=>.5-R()) Next iteration, reduced further with fat arrow notation, which sadly only works for Firefox 22 and above. • Nice compacting! :D – IQAndreas Jan 6 '14 at 9:01 ## JavaScript (269 characters compacted) For clarity, this is the code before I compacted it down JS-Fiddle of it: var lowerLetters = "abcdefghijklmnopqrstuvwxyz"; var upperLetters = "ABCDEFGHIJKLMNOPQRSTUVWXYZ"; var numbers = "0123456789"; var symbols = "~!@#$%^&*()_+-={}|[]\\:\";'<>?,./";
var allCharacters = lowerLetters + upperLetters + numbers + symbols;

String.prototype.randomChar = function() {
return this[Math.floor(this.length * Math.random())];
}

var minLength = 15;
var result = [];

// Start off by picking one random character from each group
result.push(lowerLetters.randomChar());
result.push(upperLetters.randomChar());
result.push(numbers.randomChar());
result.push(symbols.randomChar());
// Next, pick a random character from all groups until the desired length is met
while(result.length < minLength) {
result.push(allCharacters.randomChar());
}
result.shuffle(); // Finally, shuffle the items (custom function; doesn't actually exist in JavaScript, but is very easy to add) -> http://stackoverflow.com/questions/2450954/how-to-randomize-shuffle-a-javascript-array
result.join("");


Here it is compacted down to 269 characters (JS-Fiddle of it):

l="abcdefghijklmnopqrstuvwxyz";
u=l.toUpperCase();
n="0123456789";
s="~!@#$%^&*()_+-={}|[]\\:\";'<>?,./"; R=Math.random; function r(t){ return t[~~(t.length*R())] } for(x=[r(l),r(u),r(n),r(s)];x.length<15;x.push(r(l+u+n+s))); x.sort(function(){return .5-R()}); alert(x.join(""));  • Since I'm ending lines with semi-colons, all removable whitespace was ignored for character counting, but left in for clarity. – IQAndreas Jan 5 '14 at 23:57 • What do you mean by shuffle() being a "custom function". Is it part of JavaScript or code you would have to write it yourself? – Hand-E-Food Jan 6 '14 at 1:51 • @Hand-E-Food I meant it is not built into JavaScript, and since any developers here should know how to shuffle an array, I felt including the function in the code was unnecessary. It is available in the JS-Fiddle though (line 16). – IQAndreas Jan 6 '14 at 1:58 • My point is, it counts towards your byte-count. But I see now you have implemented it in the compacted version, so please ignore me. :-) – Hand-E-Food Jan 6 '14 at 4:39 # Clojure (63): (->> (map char (range 33 127)) (shuffle) (take 15) (apply str))  But need to be improved to ensure that containing at least 1 character of each category (Upper, Lower, Digit, Symbol). # In sql-server declare @a nvarchar(28) set @a='abcdefghijklmnopqrstuvwxyz' declare @b nvarchar(max) set @b='ABCDEFGHIJKLMNOPQRSTUVWXYZ' declare @c nvarchar(max) set @c='0123456789' declare @d nvarchar(max) set @d='~!@#$%^&*()_+-={}|[]\:";<>?,./'

select left(substring(@a,cast(rand()*10 as int),3)+substring(@b,cast(rand()*10 as int),6)+substring(@c,cast(rand()*10 as int),3)+substring(@d,cast(rand()*10 as int),5),15)


# see it in Action--2

• I'm having trouble following the last line, but the code does not seem to fulfill the all permutations requirement. – IQAndreas Jan 6 '14 at 9:15
• Your code will never generate any password starting with ~0Aa, nor any password where b is followed by a. – Heinzi Jan 6 '14 at 9:39
• @Heinzi:yes i agree all permutations required r not taken into account it just displays a 15 length..characters randomly choosen from a...z,A..Z,0..9,!...+ :(... – vhadalgi Jan 6 '14 at 9:47

## SAS (191)

%macro c(p);compress(p,,"&p")ne''%mend;data x;length p$15;do x=1by 1;do t=1to 15;substr(p,t,1)=byte(ranuni(7)*94+33);end;if %c(kd)and %c(kl)and %c(ku)and %c(ad)then do;put p;stop;end;end;run;  *TQP,(f=h10*)S= Commented/indented: %macro c(p); /*compress removes or keeps certain classes of characters*/ compress(p,,"&p")ne'' %mend; data x; length p$15;
do x=1by 1;
do t=1to 15;
substr(p,t,1)=byte(ranuni(7)*94+33); /*give it a 33-126, byte discards the noninteger portion rounding down*/
end;
if %c(kd)and %c(kl)and %c(ku)and %c(ad)then do; /*k=keep d=digit l/u=lower/upper ad=remove digits and alphas*/
put p;
stop;  /*met our requirement, head home*/
end;
end;
run;


# PowerShell: 119

Gofled Code

for(;!($x-cmatch'.*(?=.*[a-z])(?=.*[A-Z])(?=.*\d)(?=.*[!-/:-@[-{-~]).*')){$x='';1..15|%{$x+=[char](33..126|random)}}$x


Un-golfed and Commented

# Start for loop definition.
for(
# Skip variable initialization, start definition of run condition.
;
# Loop runs if $x does not meet complexity requirements. # Length requirement is not tested here because it is enforced by the generator later. # Much thanks to @VasiliSyrakis for the RegEx help. !($x-cmatch'.*(?=.*[a-z])(?=.*[A-Z])(?=.*\d)(?=.*[!-/:-@[-{-~]).*')
)
{
# Reset $x in case the script block has already run.$x='';
# Use ForEach-Object, via the % alias, to run a loop 15 times.
1..15|%{
# Append a random ASCII character from 33-126 to $x. # Note: Use get-random instead of random for faster performance.$x+=[char](33..126|random)
}
}
# Display $x.$x
# Variable cleanup - not included in golfed code.
rv x

• I think this regex may make it ever so slightly shorter: ^.*(?=.{15,})(?=.*[a-z])(?=.*[A-Z])(?=.*\d)(?=.*[!#$%&? "]).*$, you can do one match with this, which will only match when there is one Upper,Lower,Digit,Symbol. – Vasili Syrakis Jan 9 '14 at 6:01
• @VasiliSyrakis Ok, You might have to walk me through that a bit. Feel free to start up a chat room if you think it will take awhile. A couple things I'm confused on: 1.) I see the number 15 included there. Is that to make sure the string is exactly 15 characters? If so, this can be omitted, since the script naturally only generates 15-character strings. 2.) What do you mean "will only match when there is one Upper, Lower, Digit, Symbol"? Does that mean it will only match when there is exactly one of each, or at least one of each? The former will break things. – Iszi Jan 9 '14 at 6:07
• Also, does your RegEx ignore ordering of characters? For example, if tuned down to match 4-character strings, would both 1aZ% and (p3R match? I had some difficulty finding ways to do that online. – Iszi Jan 9 '14 at 6:08
• Tested the new RegEx against output from my current script. It doesn't seem quite reliable. Code: $x-cmatch'^.*(?=.{15,})(?=.*[a-z])(?=.*[A-Z])(?=.*\d)(?=.*[!#$%&? "]).*$' Good matches: C>suD1?hTwbDx(z j%4O]HyeG|u[U$5 O/rGeD0$hJk=GO/ Failed matches: 3evthX3_X^nBrR .nA~uYzrR4YV-r. u-IjZE48ntQ;HxV – Iszi Jan 9 '14 at 6:17 • How do I open a chat room? – Vasili Syrakis Jan 9 '14 at 6:17 Python 2.7 (149) from random import* ''.join(map(lambda x:chr(randint(*(x[1]or(32,126)))),sorted(map(None,sample(range(15),15),((48,57),(65,90),(97,122),(33,47))))))  Written out in a more readable (and not executable) way; from random import * ''.join( # Concatenate characters to string map( # Map all characters using below lambda lambda x:chr(randint(*(x[1] or (32, 126)))), # Map a single range to a random character # within a specific range if supplied, # otherwise the default "all" range. sorted( # After distributing ranges, sort map(None, # zip_longest alternative, distributes the # required ranges over 4 random positions sample(range(15), 15), # 0-14 in random order ((48, 57), (65, 90), (97, 122), (33, 47)) # The 4 required ranges ) ) ) )  Fairly straight forward and surprisingly not much longer than a "generate, retry on match fail" version. • Are you sure this can really generate all eligible passwords, including e.g. 0Aa~~~~~~~~~~~~? (Note that '~' == chr(126).) – Ilmari Karonen Jan 7 '14 at 13:39 PSQL (189) Feels like PSQL is a bit verbose... :) SELECT ARRAY_TO_STRING(ARRAY_AGG(CHR((TRUNC((b-a)*RANDOM()+a))::int)ORDER BY RANDOM()),'')FROM(SELECT 32 a,127 b FROM generate_series(1,11)UNION ALL VALUES(48,58),(65,91),(97,123),(33,48))a  ## PHP, 235 225 This script shuffles characters around and then is checked via RegEx to make sure the password is strong (or it is regenerated). <?php while(!preg_match('/^(?=.*[A-Z])(?=.*[^A-Za-z])(?=.*[0-9])(?=.*[a-z]).{15}$/',$p)){$p = substr(str_shuffle('ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789~!@#$%^&*()_+-={}|[]\:";\'<>?,./'),0,15); } echo$p;

• Clever, but doesn't allow duplicate characters. – Hand-E-Food Jan 6 '14 at 1:44
• Instead of while(true) ... if (condition) break you can use while (!condition) – exussum Jan 6 '14 at 8:12

Javascript (209)

r=Math.random;function t(x,y){return String.fromCharCode(Math.floor(y*r()+x))};x=[t(33,14),t(48,10),t(65,26),t(97,26)];for(i=0;i<11;i++)x.push(t(32,95));console.log(x.sort(function(){return r()-0.5}).join(''))


Semi-ungolfed;

// Return a character in the range [x,x+y)
function t(x,y) { return String.fromCharCode(Math.floor(y*Math.random()+x)) }
// Prefill required ranges
x=[ t(33,14), t(48,10), t(65,26), t(97,26)]
// Push 11 totally random (valid) characters
for(i=0; i<11; i++)
x.push(t(32,95))
// Shuffle and output as string
console.log(x.sort(function(){return Math.random()-0.5})
.join(''))


## Perl, 92

Not as concise as the Ruby answer, but I'm sure a Perl wizard could make this even shorter... I'm not too happy with all the m//s at the end, but seems to work and should satisfy the conditions to eventually generate all permutations.

do{$_=join"",map{(map{chr}33..127)[rand 94]}0..14}while!(/[A-Z]/&/[a-z]/&/\d/&/[\W_]/);print  Sample usage: perl -e 'do{$_=join"",map{(map{chr}33..127)[rand 94]}0..14}while!(/[A-Z]/&/[a-z]/&/\d/&/[\W_]/);print'


Edited to fix validation and change [[:punct:]] to [\W_] after MvGs comments.

• Your generation part is nice, but your selection criterion in the loop condition is plain wrong: e.g. a password of aaaaaaaaaaaaaa would cause the loop to terminate. You should test criteria with non-random passwords to make sure they do what you want them to. – MvG Jan 7 '14 at 13:10
• Indeed, you are correct, I've fixed this and saved some bytes! Thanks! – Dom Hastings Jan 7 '14 at 13:50
• Are you sure about this [[:punct:]]? I guess I'd prefer '[\W_], which is shorter and of which I'm even more sure that it is correct, at least combined with your 33..127 range. – MvG Jan 7 '14 at 15:04
• A good point, I think I was concerned that \W didn't include _, however you're absolutely right, it's not needed: gist.github.com/anonymous/8301237. Thank you! – Dom Hastings Jan 7 '14 at 15:41

## Java 7 (270 234 characters)

The premise is the same used by @assylias with java 8 (generate random passwords until valid password). However, instead of using lambdas, the password is generated by iterating a char array and validated by matching a regex.

class A {
public static void main(String [] a) {
byte[] b = new byte[15];
String s;
do {
new java.util.Random().nextBytes(b);
s = new String(b);
} while(!s.matches("(?=.*?[a-z])(?=.*?[A-Z])(?=.*?\\d)(?=.*?[!-/:-@\$-]).*")); System.out.println(s); } }  Minified Code: class A {public static void main(String[] a){byte[] b=new byte[15];String s;do{new java.util.Random().nextBytes(b);s=new String(b);}while(!s.matches("(?=.*?[a-z])(?=.*?[A-Z])(?=.*?\\d)(?=.*?[!-/:-@\$$-]).*"));System.out.println(s);}}  # Powershell ## One Liner version (143 bytes) sal g random;1..11|%{P+=[char](33..126|g)};(65..90|g),(97..122|g),(48..57|g),(33..47+58..64+123..126|g)|%{P=P.insert((1..11|g),[char]_)};P  ## Mini version (146 bytes) sal g random 1..11|%{P+=[char](33..126|g)} (65..90|g),(97..122|g),(48..57|g),(33..47+58..64+123..126|g)|%{P=P.insert((1..11|g),[char]_)} P  ## Readable version (860 bytes) function pwgen { # Fulfill Upper,Lower,Digit,Symbol requirement by predefining ASCII ranges for each # These will be added into the string beginning at line 24 [array[]]symbolrange = (33..47),(58..64),(123..126) [char]upper = (get-random (65..90)) [char]lower = (get-random (97..122)) [char]digit = (get-random (48..57)) [char]symbol = symbolrange | get-random [char[]]requirement = upper + lower + digit + symbol # Create the first 11 characters using any ASCII character between 32 - 126 foreach (number in (1..11)) { [string]pass += [char](get-random (33..126)) } # Insert each requirement character at a random position in the string foreach (char in requirement) { [string]pass = pass.insert((Get-Random (1..11)),char) } return pass }  Credit to Iszi for various tips to shorten the code. • This doesn't cover all permutations. As an example, abcd1234ABCD{|}~ will never come up because symbol forces at least one of the symbols to be between ASCII 33 and 47. – Hand-E-Food Jan 6 '14 at 4:43 • Dangit! Must you point out my laziness!? Just kidding... I've edited it now. I also made each "requirement" character go to a separate index within the string, instead of clumping the four in together at the same index. Now if only I could shorten this... – Vasili Syrakis Jan 6 '14 at 6:23 • Is there any reason you can't shave a couple more characters by shortening SR to, perhaps Q? – Iszi Jan 8 '14 at 23:54 • You should also be able to cut stuff like (g(65..90)) down to 65..90|g'. And change the foreach statements to foreach-object loops using the % alias. Example: foreach(N in (1..11)){...} should be doable as 1..11|%{...}. I'm pretty sure there's other optimizations which are possible, but I've actually got a completely different implementation in mind that I'm planning to try later. – Iszi Jan 9 '14 at 0:03 • Nice tips :) I shortened it to 213 bytes if I take out the carriage returns and replace with semicolons. – Vasili Syrakis Jan 9 '14 at 0:20 # Factor, 196 characters Same algorithm as MvG and moose's. It is not the shortest but should satisfy all the (current) criteria in the question: USING: io kernel math pcre random sequences sequences.repeating ; [ 15 94 random-integers [ 33 + ] "" map-as dup 60 cycle "[A-Z].*[a-z].*[0-9].*[\\W_]" findall { } = not ] [ drop ] until print  • I might be misinterpreting the regex, but I think something like ~{}|1234abcdABCD will fail the regex. – Hand-E-Food Jan 8 '14 at 23:41 • No it will work: "~{}|1234abcdABCD" 60 cycle "[A-Z].*[a-z].*[0-9].*[\\W_]" findall empty? not => t – Björn Lindqvist Jan 9 '14 at 8:41 • I'll take your word for it. :-) – Hand-E-Food Jan 9 '14 at 22:51 # C - 154 characters char p[16],c,f,w;main(){srand(time());while(f^15){c=p[15]=f=0;while(c^15){w=33+rand()%94;f|=w >96&&w<123?1:w>47&&w<59?2:w>64&&w<91?4:8;p[c++]=w;}}puts(p);}  How do I hate srand()? Let me count the ways. # Haskell, 192 import System.Random main=getStdGen>>= \g->(print.(take 15))until((\x->all(any(elemx))[['a'..'z'],['A'..'Z'],['0'..'9'],['!'..'/']++":;<=>?@[\$^_{|}~"]).(take 15))tailrandomRs('"','~')g  The printed string has quotes around it and escapes the backslash and quote characters; if that's unacceptable, print can be replaced with putStrLn for 3 more bytes. Here's a more readable version: import System.Random main = do g <- getStdGen let chars = randomRs ('"', '~') g let password = take 15  until (hasChars.(take 15)) tail chars print password hasChars :: String -> Bool hasChars x = all (any (elem x))  [ ['a'..'z'] , ['A'..'Z'] , ['0'..'9'] , "!\"#%&'()*+,-./:;<=>?@[\$$^_{|}~"
]


It's quite straightforward, it just creates an infinite/lazy list of random ASCII characters in the range '!' to '~', then tosses out the first element until the first 15 characters have at least one character from each string of required characters.

### Excel VBA, 209 bytes

For i = 1 To 15
x = x + Chr(Int(90 * Rnd + 33))
Next
p = "^(?=.*[a-z])(?=.*[A-Z])(?=.*\d)(?=.*(_|[^\w])).+$" With New RegExp .Pattern = p Set m = .Execute(x) If m.Count = 0 Then MsgBox "redo" Else MsgBox x End If End With  Randomly generates 15 ASCII characters so all possible combinations are possible. Then uses a regular expression pattern to check if it contains at least one of each criteria. If it does then the password is displayed, if not "redo" is displayed. Credit to Bart Kiers for the Regular Expression pattern: https://stackoverflow.com/questions/1559751/regex-to-make-sure-that-the-string-contains-at-least-one-lower-case-char-upper ## AutoHotkey 352 global o:={} loop 4 o[c()]:=o(A_index-1) loop 11 o[c()]:=o(m(r(),4)) loop 15 s.=o[A_index-1] msgbox % s r(){ Random,z return z } m(z,r){ return mod(z,r) } c(){ while o[t:=m(r(),15)]!="" j++ return t } o(f){ r0:=48,l0:=10,r1:=97,l1:=l2:=26,r2:=65 r := chr(r%f%+m(r(),l%f%)) if f=3 r:=Substr("~!@#$%^&*()_+-={}|[]\:"";'<>?,./",m(r(),32)+1,1)
return r
}


Using - Just run the script

## Python (121 characters)

Makes use of the fact that you can multiply lists in Python [1,2,3] * 2 gives [1,2,3,1,2,3]. Imports random. Numbers in a list multiplied by three are borders between ranges in ascii table for needed characters, e.g. [65, 90] maps to uppercase letters.

print "".join([random.choice([chr(i) for i in range(z[0],z[1])]) for z in [[33,48],[48,58],[58,65],[65,90],[90,123]]* 3])

• "It must be able to generate all permutations of all allowable characters.". I don't think it does that since the ranges are always applied in the same order...? – Joachim Isaksson Jan 6 '14 at 13:01
• You are right, thank you. Indeed I didn't notice that ranges should be applied in random order, they should be shuffled, I'll edit that in a moment. – Pawelmhm Jan 6 '14 at 13:57
• This needs to actually include import random in the code. – Mego May 21 '16 at 11:06

PHP 5.5 (230 bytes)

echo preg_replace_callback('/./', function ($a) { return chr([rand(33, 126), rand(48, 57), rand(65, 90), rand(97, 122), ord(str_shuffle('~!@#$%^&*()_+-={}|[]\:";\'<>?,./')[0])][$a[0]]); } , str_shuffle(str_pad('1234', 15, 0)));  Or on one line (211 bytes) echo preg_replace_callback('/./',function($a){return chr([rand(33,126),rand(48,57),rand(65,90),rand(97,122),ord(str_shuffle('~!@#$%^&*()_+-={}|[]\:";\'<>?,./')[0])][$a[0]]);},str_shuffle(str_pad('1234',15,0)));