Create regex from samples algorithm - regex

AFAIK no one have implemented an algorithm that takes a set of strings and substrings and gives back one or more regular expressions that would match the given substrings inside the strings. So, for instance, if I'd give my algorithm this two samples:
string1 = "fwef 1234 asdfd"
substring1 = "1234"
string2 = "asdf456fsdf"
substring2 = "456"
The algorithm would give me the regular expression "[0-9]*" back. I know it could give more than one regex or even no possible regex back and you might find 1000 reasons why such algorithm would be close to impossible to implement to perfection. But what's the closest thing?
I don't really care about regex itself also. Basically what I want is an algorithm that takes samples as the ones above and then finds a pattern in them that can be used to easily find the "kind" of text I want to find in a string without having to write any regex or code manually.

I don't have proof but I suspect no such discrete algorithm with a finite output could exist since you are asking for the creation of a regular language which could be "large" in respect to the input size.
With that, I suggest you peek at txt2re which can break down sample texts one-by-one and help you build regexes.

FlashFill a new feature of MS Excel 2013 would do exactly the task you want, but it does not give you the regular expression. It's a NP-complete problem and an open question for practical purposes. If you're interested in how to synthesise string manipulation from multiple examples, Go Flash Fill official website and read a few papers. They have pseudo-code and demo. movies as well.

There are many such algorithm in fact. This is a research area called "Grammatical inference".
I know RPNI, for example. (you could also look on the probabilistic branch, alergia, MDI, DEES). These algorithms generate DSA (Deterministic State Automata). In fact you absolutely don't need to enter the strings in your example. Only substrings.
There are also some algorithms to generate directly Non deterministic automata.
Of course, get the regular expression from an Non Deterministic Automata is easy.
The main ideas are simple:
Generate a PTSA (Prefix Tree State Automata) from your sample.
Then, you have to try to "merge" some states. From these merge, will emerge loops (i.e. * in the regular expression). All the difficulty being to choose the right rule to merge.

Here you go:
re = '|'.join(substrings)
If you want anything more general, your algorithm is going to have to make educated guesses about what type of strings are acceptable as matches, and it's trivial to demonstrate that no procedure can account for all possible sets of possible inputs without simply enumerating them all. For instance, consider some of these scenarios:
Match all prime numbers
Match hexadecimal strings, but no strings containing 'f' are in the sample set
Match the same string repeated twice
Match any even-length string
The root problem is that your question is incompletely specified. If you have a more specific requirement, that might be solvable, depending on what it is.

Related

How to detect deviations in a sequence from a pattern?

i want to detect the passages where a sequence of action deviates from a given pattern and can't figure out a clever solution to do this, although the problem sound really simple.
The objective of the pattern is to describe somehow a normal sequence. More specific: "What actions should or shouldn't be contained in an action sequence and in which order?" Then i want to match the action sequence against the pattern and detect deviations and their locations.
My first approach was to do this with regular expressions. Here is an example:
Example 1:
Pattern: A.*BC
Sequence: AGDBC (matches)
Sequence: AGEDC (does not match)
Example 2:
Pattern: ABCD
Sequence: ABD (does not match)
Sequence: ABED (does not match)
Sequence: ABCED (does not match)
Example 3:
Pattern: ABCDEF
Sequence: ABXDXF (does not match)
With regular expressions it is simple to detect a error but not where it occurs. My approach was to successively remove the last regex block until I can find the pattern in the sequence. Then i will know the last correct actions and have at least found the first deviation. But this doesn't seem the best solution to me. Further i can't all deviations.
Other soultions in my mind are working with state machines, order tools like ANTLR. But I don't know if they can solve my problem.
I want to detect errors of omission and commission and give an user the possibility to create his own pattern. Do you know a good way to do this?
While matching your input, a regular expression engine has information on the location of a mismatch -- it may however not provide it in a way where you can easily access it.
Consider e.g. a DFA implementing the expression. It fetches characters sequentially, matching them with expectation, and you are interested at the point in the sequence where there is no valid match.
Other implementations may go back and forth, and you would be interested in the maximum address of any character that was fetched.
In Java, this could possibly be done by supplying a CharSequence implementation to
java.util.regex.Pattern.matches(String regex, CharSequence input)
where the accessor methods keep track of the maximum index.
I have not tried that, though. And it does not help you categorizing an error, either.
have you looked at markov chains? http://en.wikipedia.org/wiki/Markov_chain - it sounds like you want unexpected transitions. maybe also hidden markov models http://en.wikipedia.org/wiki/Hidden_Markov_Models
Find an open-source implementation of regexs, and add in a hook to return/set/print/save the index at which it fails, if a given comparison does not match. Alternately, write your own RE engine (not for the faint of heart) so it'll do exactly what you want!

Distance between regular expression

Can we compute a sort of distance between regular expressions ?
The idea is to mesure in which way two regular expression are similar.
You can build deterministic finite-state machines for both regular expressions and compare the transitions. The difference of both transitions can then be used to measure the distance of these regular expressions.
There are a few of metrics you could use:
The length of a valid match. Some regexs have a fixed size, some an upper limit and some a lower limit. Compare how similar their lengths or possible lengths are.
The characters that match. Any regex will have a set of characters a match can contain (maybe all characters). Compare the set of included characters.
Use a large document and see how many matches each regex makes and how many of those are identical.
Are you looking for strict equivalence?
I suppose you could compute a Levenshtein Distance between the actual Regular Experssion strings. That's certainly one way of measuring a "distance" between two different Regular Expression strings.
Of course, I think it's possible that regular expressions are not required here at all, and computing the Levenshtein Distance of the actual "value" strings that the Regular Expressions would otherwise be applied to, may yield a better result.
If you have two regular expressions and have a set of example inputs you could try matching every input against each regex. For each input:
If they both match or both don't match, score 0.
If one matches and the other doesn't, score 1.
Sum this score over all inputs, and this will give you a 'distance' between the regular expressions. This will give you an idea of how often two regular expressions will differ for typical input. It will be very slow to calculate if your sample input set is large. It won't work at all if both regexes fail to match for almost all random strings and your expected input is entirely random. For example the regex 'sgjlkwren' and the regex 'ueuenwbkaalf' would probably both never match anything if tested on random input, so this metric would say the distance between them is zero. That might or might not be what you want (probably not).
You might be able to analyze the structure of the regex and use biased random sampling to deliberately hit strings that match more frequently than in completely random input. For example, if both regex require that the string starts with 'foo', you could make sure that your test inputs also always start with foo, to avoid wasting time testing strings that you know will fail for both.
So in conclusion: unless you have a very specific situation with a restricted input set and/or restricted regular expression language, I'd say its not possible. If you do have some restrictions on your input and on the regular expression, it might be possible. Please specify what these restrictions are and maybe I can come up with something better.
There's an answer hidden in an earlier question here on SO: Generating strings from regexes. You can calculate an (asymmetric) distance measure by generating strings using one regex and checking how many of those match the other regex.
This can be optimized by stripping out shared prefixes/suffixes. E.g. a[0-9]* and a[0-7]* share the a prefix, so you can calculate the distance between [0-9]* and [0-7]* instead.
I think first you need to understand for yourself how you see a "difference" between two expressions. Basically, define a distance metric.
In general case, it would be quite different to make. Depending on what you need to do, you may see allowing one different character in some place as a big difference. In the other case, allowing any number of consequent but same characters may not yield much difference.
I'd like to emphasize as well that normally when they talk about distance functions, they apply them to..., well, let's call them, tokens. In our case, character sequences. What you are willing to do, is to apply this method not to those tokens, but to the rules a multitude of tokens will match. I'm not quite sure it even makes sense.
Still, I believe we could think of something, but not in general, but for one particular and quite restricted case. Do you have some sort of example to show us?

In "aa67bc54c9", is there any way to print "aa" 67 times, "bc" 54 times and so on, using regular expressions?

I was asked this question in an interview for an internship, and the first solution I suggested was to try and use a regular expression (I usually am a little stumped in interviews). Something like this
(?P<str>[a-zA-Z]+)(?P<n>[0-9]+)
I thought it would match the strings and store them in the variable "str" and the numbers in the variable "n". How, I was not sure of.
So it matches strings of type "a1b2c3", but a problem here is that it also matches strings of type "a1b". Could anyone suggest a solution to deal with this problem?
Also, is there any other regular expression that could solve this problem?
Do you know why "regular expressions" are called "regular"? :-)
That would be too long to explain, I'll just outline the way. To match a pattern (i.e. decide whether a given string is "valid" or "invalid"), a theoretical informatician would use a finite state automaton. That's an abstract machine that has a finite number of states; each tick it reads a char from the input and jumps to another state. The pattern of where to jump from particular state when a particular character is read is fixed. Some states are marked as "OK", some--as "FAIL", so that by examining state of a machine you can check whether your text is "valid" (i.e. a valid e-mail).
For example, this machine only accepts "nice" as its "valid" word (a pic from Wikipedia):
A set of "valid" words such a machine theoretically can distinguish from invalid is called "regular language". Not every set is a regular language: for example, finite state automata are incapable of checking whether parentheses in string are balanced.
But constructing state machines was a complex task, compared to the complexity of defining what "valid" is. So the mathematicians (mainly S. Kleene) noted that every regular language could be described with a "regular expression". They had *s and |s and were the prototypes of what we know as regexps now.
What does it have to do with the problem? The problem in subject is essentially non-regular. It can't be expressed with anything that works like a finite automaton.
The essence is that it should contain a memory cell that is capable to hold an arbitrary number (repetition count in your case). Finite automata and classical regular expressions can not do this.
However, modern regexps are more expressive and are said to be able to check balanced parentheses! But this may serve as a good example that you shouldn't use regexps for tasks they don't suit. Let alone that it contains code snippets; this makes the expression far from being "regular".
Answering the initial question, you can't solve your problem with using anything "regular" only. However, regexps could be aid you in solving this problem, as in tster's answer
Perhaps, I should look closer to tster's answer (do a "+1" there, please!) and show why it's not the "regular expression" solution. One may think that it is, it just contains print statement (not essential) and a loop--and loop concept is compatible with finite state automaton expressive power. But there is one more elusive thing:
while ($line =~ s/^([a-z]+)(\d+)//i)
{
print $1
x # <--- this one
$2;
}
The task of reading a string and a number and printing repeatedly that string given number of times, where the number is an arbitrary integer, is undoable on a finite state machine without additional memory. You use a memory cell to keep that number and decrease it, and check for it to be greater than zero. But this number may be arbitrarily big, and it contradicts with a finite memory available to the finite state machine.
However, there's nothing wrong with classical pattern /([abc]*){5}/ that matches something "regular" repeated fixed number of times. We essentially have states that correspond to "matched pattern once", "matched pattern twice" ... "matched pattern 5 times". There's finite number of them, and that's the gist of the difference.
how about:
while ($line =~ s/^([a-z]+)(\d+)//i)
{
print $1 x $2;
}
Answering your question directly:
No, regular expressions match text and don't print anything, so there is no way to do it solely using regular expressions.
The regular expression you gave will match one string/number pair; you can then print that repeatedly using an appropriate mechanism. The Perl solution from #tster is about as compact as it gets. (It doesn't use the names that you applied in your regex; I'm pretty sure that doesn't matter.)
The remaining details depend on your implementation language.
Nope, this is your basic 'trick question' - no matter how you answer it that answer is wrong unless you have exactly the answer the interviewer was trained to parrot. See the workup of the issue given by Pavel Shved - note that all invocations have 'not' as a common condition, the tool just keeps sliding: Even when it changes state there is no counter in that state
I have a rather advanced book by Kenneth C Louden who is a college prof on the matter, in which it is stated that the issue at hand is codified as "Regex's can't count." The obvious answer to the question seems to me at the moment to be using the lookahead feature of Regex's ...
Probably depends on what build of what brand of regex the interviewer is using, which probably depends of flight-dynamics of Golf Balls.
Nice answers so far. Regular expressions alone are generally thought of as a way to match patterns, not generate output in the manner you mentioned.
Having said that, there is a way to use regex as part of the solution. #Jonathan Leffler made a good point in his comment to tster's reply: "... maybe you need a better regex library in your language."
Depending on your language of choice and the library available, it is possible to pull this off. Using C# and .NET, for example, this could be achieved via the Regex.Replace method. However, the solution is not 100% regex since it still relies on other classes and methods (StringBuilder, String.Join, and Enumerable.Repeat) as shown below:
string input = "aa67bc54c9";
string pattern = #"([a-z]+)(\d+)";
string result = Regex.Replace(input, pattern, m =>
// can be achieved using StringBuilder or String.Join/Enumerable.Repeat
// don't use both
//new StringBuilder().Insert(0, m.Groups[1].Value, Int32.Parse(m.Groups[2].Value)).ToString()
String.Join("", Enumerable.Repeat(m.Groups[1].Value, Int32.Parse(m.Groups[2].Value)).ToArray())
+ Environment.NewLine // comment out to prevent line breaks
);
Console.WriteLine(result);
A clearer solution would be to identify the matches, loop over them and insert them using the StringBuilder rather than rely on Regex.Replace. Other languages may have compact idioms to handle the string multiplication that doesn't rely on other library classes.
To answer the interview question, I would reply with, "it's possible, however the solution would not be a stand-alone 100% regex approach and would rely on other language features and/or libraries to handle the generation aspect of the question since the regex alone is helpful in matching patterns, not generating them."
And based on the other responses here you could beef up that answer further if needed.

Efficiently querying one string against multiple regexes

Lets say that I have 10,000 regexes and one string and I want to find out if the string matches any of them and get all the matches.
The trivial way to do it would be to just query the string one by one against all regexes. Is there a faster,more efficient way to do it?
EDIT:
I have tried substituting it with DFA's (lex)
The problem here is that it would only give you one single pattern. If I have a string "hello" and patterns "[H|h]ello" and ".{0,20}ello", DFA will only match one of them, but I want both of them to hit.
This is the way lexers work.
The regular expressions are converted into a single non deterministic automata (NFA) and possibily transformed in a deterministic automata (DFA).
The resulting automaton will try to match all the regular expressions at once and will succeed on one of them.
There are many tools that can help you here, they are called "lexer generator" and there are solutions that work with most of the languages.
You don't say which language are you using. For C programmers I would suggest to have a look at the re2c tool. Of course the traditional (f)lex is always an option.
I've come across a similar problem in the past. I used a solution similar to the one suggested by akdom.
I was lucky in that my regular expressions usually had some substring that must appear in every string it matches. I was able to extract these substrings using a simple parser and index them in an FSA using the Aho-Corasick algorithms. The index was then used to quickly eliminate all the regular expressions that trivially don't match a given string, leaving only a few regular expressions to check.
I released the code under the LGPL as a Python/C module. See esmre on Google code hosting.
We had to do this on a product I worked on once. The answer was to compile all your regexes together into a Deterministic Finite State Machine (also known as a deterministic finite automaton or DFA). The DFA could then be walked character by character over your string and would fire a "match" event whenever one of the expressions matched.
Advantages are it runs fast (each character is compared only once) and does not get any slower if you add more expressions.
Disadvantages are that it requires a huge data table for the automaton, and there are many types of regular expressions that are not supported (for instance, back-references).
The one we used was hand-coded by a C++ template nut in our company at the time, so unfortunately I don't have any FOSS solutions to point you toward. But if you google regex or regular expression with "DFA" you'll find stuff that will point you in the right direction.
Martin Sulzmann Has done quite a bit of work in this field.
He has a HackageDB project explained breifly here which use partial derivatives seems to be tailor made for this.
The language used is Haskell and thus will be very hard to translate to a non functional language if that is the desire (I would think translation to many other FP languages would still be quite hard).
The code is not based on converting to a series of automata and then combining them, instead it is based on symbolic manipulation of the regexes themselves.
Also the code is very much experimental and Martin is no longer a professor but is in 'gainful employment'(1) so may be uninterested/unable to supply any help or input.
this is a joke - I like professors, the less the smart ones try to work the more chance I have of getting paid!
10,000 regexen eh? Eric Wendelin's suggestion of a hierarchy seems to be a good idea. Have you thought of reducing the enormity of these regexen to something like a tree structure?
As a simple example: All regexen requiring a number could branch off of one regex checking for such, all regexen not requiring one down another branch. In this fashion you could reduce the number of actual comparisons down to a path along the tree instead of doing every single comparison in 10,000.
This would require decomposing the regexen provided into genres, each genre having a shared test which would rule them out if it fails. In this way you could theoretically reduce the number of actual comparisons dramatically.
If you had to do this at run time you could parse through your given regular expressions and "file" them into either predefined genres (easiest to do) or comparative genres generated at that moment (not as easy to do).
Your example of comparing "hello" to "[H|h]ello" and ".{0,20}ello" won't really be helped by this solution. A simple case where this could be useful would be: if you had 1000 tests that would only return true if "ello" exists somewhere in the string and your test string is "goodbye;" you would only have to do the one test on "ello" and know that the 1000 tests requiring it won't work, and because of this, you won't have to do them.
If you're thinking in terms of "10,000 regexes" you need to shift your though processes. If nothing else, think in terms of "10,000 target strings to match". Then look for non-regex methods built to deal with "boatloads of target strings" situations, like Aho-Corasick machines. Frankly, though, it seems like somethings gone off the rails much earlier in the process than which machine to use, since 10,000 target strings sounds a lot more like a database lookup than a string match.
Aho-Corasick was the answer for me.
I had 2000 categories of things that each had lists of patterns to match against. String length averaged about 100,000 characters.
Main Caveat: The patters to match were all language patters not regex patterns e.g. 'cat' vs r'\w+'.
I was using python and so used https://pypi.python.org/pypi/pyahocorasick/.
import ahocorasick
A = ahocorasick.Automaton()
patterns = [
[['cat','dog'],'mammals'],
[['bass','tuna','trout'],'fish'],
[['toad','crocodile'],'amphibians'],
]
for row in patterns:
vals = row[0]
for val in vals:
A.add_word(val, (row[1], val))
A.make_automaton()
_string = 'tom loves lions tigers cats and bass'
def test():
vals = []
for item in A.iter(_string):
vals.append(item)
return vals
Running %timeit test() on my 2000 categories with about 2-3 traces per category and a _string length of about 100,000 got me 2.09 ms vs 631 ms doing sequential re.search() 315x faster!.
You'd need to have some way of determining if a given regex was "additive" compared to another one. Creating a regex "hierarchy" of sorts allowing you to determine that all regexs of a certain branch did not match
You could combine them in groups of maybe 20.
(?=(regex1)?)(?=(regex2)?)(?=(regex3)?)...(?=(regex20)?)
As long as each regex has zero (or at least the same number of) capture groups, you can look at what what captured to see which pattern(s) matched.
If regex1 matched, capture group 1 would have it's matched text. If not, it would be undefined/None/null/...
If you're using real regular expressions (the ones that correspond to regular languages from formal language theory, and not some Perl-like non-regular thing), then you're in luck, because regular languages are closed under union. In most regex languages, pipe (|) is union. So you should be able to construct a string (representing the regular expression you want) as follows:
(r1)|(r2)|(r3)|...|(r10000)
where parentheses are for grouping, not matching. Anything that matches this regular expression matches at least one of your original regular expressions.
I would recommend using Intel's Hyperscan if all you need is to know which regular expressions match. It is built for this purpose. If the actions you need to take are more sophisticated, you can also use ragel. Although it produces a single DFA and can result in many states, and consequently a very large executable program. Hyperscan takes a hybrid NFA/DFA/custom approach to matching that handles large numbers of expressions well.
I'd say that it's a job for a real parser. A midpoint might be a Parsing Expression Grammar (PEG). It's a higher-level abstraction of pattern matching, one feature is that you can define a whole grammar instead of a single pattern. There are some high-performance implementations that work by compiling your grammar into a bytecode and running it in a specialized VM.
disclaimer: the only one i know is LPEG, a library for Lua, and it wasn't easy (for me) to grasp the base concepts.
I'd almost suggest writing an "inside-out" regex engine - one where the 'target' was the regex, and the 'term' was the string.
However, it seems that your solution of trying each one iteratively is going to be far easier.
You could compile the regex into a hybrid DFA/Bucchi automata where each time the BA enters an accept state you flag which regex rule "hit".
Bucchi is a bit of overkill for this, but modifying the way your DFA works could do the trick.
I use Ragel with a leaving action:
action hello {...}
action ello {...}
action ello2 {...}
main := /[Hh]ello/ % hello |
/.+ello/ % ello |
any{0,20} "ello" % ello2 ;
The string "hello" would call the code in the action hello block, then in the action ello block and lastly in the action ello2 block.
Their regular expressions are quite limited and the machine language is preferred instead, the braces from your example only work with the more general language.
Try combining them into one big regex?
I think that the short answer is that yes, there is a way to do this, and that it is well known to computer science, and that I can't remember what it is.
The short answer is that you might find that your regex interpreter already deals with all of these efficiently when |'d together, or you might find one that does. If not, it's time for you to google string-matching and searching algorithms.
The fastest way to do it seems to be something like this (code is C#):
public static List<Regex> FindAllMatches(string s, List<Regex> regexes)
{
List<Regex> matches = new List<Regex>();
foreach (Regex r in regexes)
{
if (r.IsMatch(string))
{
matches.Add(r);
}
}
return matches;
}
Oh, you meant the fastest code? i don't know then....

Random string that matches a regexp [duplicate]

This question already has answers here:
Using Regex to generate Strings rather than match them
(12 answers)
Closed 1 year ago.
How would you go about creating a random alpha-numeric string that matches a certain regular expression?
This is specifically for creating initial passwords that fulfill regular password requirements.
Welp, just musing, but the general question of generating random inputs that match a regex sounds doable to me for a sufficiently relaxed definition of random and a sufficiently tight definition of regex. I'm thinking of the classical formal definition, which allows only ()|* and alphabet characters.
Regular expressions can be mapped to formal machines called finite automata. Such a machine is a directed graph with a particular node called the final state, a node called the initial state, and a letter from the alphabet on each edge. A word is accepted by the regex if it's possible to start at the initial state and traverse one edge labeled with each character through the graph and end at the final state.
One could build the graph, then start at the final state and traverse random edges backwards, keeping track of the path. In a standard construction, every node in the graph is reachable from the initial state, so you do not need to worry about making irrecoverable mistakes and needing to backtrack. If you reach the initial state, stop, and read off the path going forward. That's your match for the regex.
There's no particular guarantee about when or if you'll reach the initial state, though. One would have to figure out in what sense the generated strings are 'random', and in what sense you are hoping for a random element from the language in the first place.
Maybe that's a starting point for thinking about the problem, though!
Now that I've written that out, it seems to me that it might be simpler to repeatedly resolve choices to simplify the regex pattern until you're left with a simple string. Find the first non-alphabet character in the pattern. If it's a *, replicate the preceding item some number of times and remove the *. If it's a |, choose which of the OR'd items to preserve and remove the rest. For a left paren, do the same, but looking at the character following the matching right paren. This is probably easier if you parse the regex into a tree representation first that makes the paren grouping structure easier to work with.
To the person who worried that deciding if a regex actually matches anything is equivalent to the halting problem: Nope, regular languages are quite well behaved. You can tell if any two regexes describe the same set of accepted strings. You basically make the machine above, then follow an algorithm to produce a canonical minimal equivalent machine. Do that for two regexes, then check if the resulting minimal machines are equivalent, which is straightforward.
String::Random in Perl will generate a random string from a subset of regular expressions:
#!/usr/bin/perl
use strict;
use warnings;
use String::Random qw/random_regex/;
print random_regex('[A-Za-z]{3}[0-9][A-Z]{2}[!##$%^&*]'), "\n";
If you have a specific problem, you probably have a specific regular expression in mind. I would take that regular expression, work out what it means in simple human terms, and work from there.
I suspect it's possible to create a general regex random match generator, but it's likely to be much more work than just handling a specific case - even if that case changes a few times a year.
(Actually, it may not be possible to generate random matches in the most general sense - I have a vague memory that the problem of "does any string match this regex" is the halting problem in disguise. With a very cut-down regex language you may have more luck though.)
I have written Parsley, which consist of a Lexer and a Generator.
Lexer is for converting a regular expression-like string into a sequence of tokens.
Generator is using these tokens to produce a defined number of codes.
$generator = new \Gajus\Parsley\Generator();
/**
* Generate a set of random codes based on Parsley pattern.
* Codes are guaranteed to be unique within the set.
*
* #param string $pattern Parsley pattern.
* #param int $amount Number of codes to generate.
* #param int $safeguard Number of additional codes generated in case there are duplicates that need to be replaced.
* #return array
*/
$codes = $generator->generateFromPattern('FOO[A-Z]{10}[0-9]{2}', 100);
The above example will generate an array containing 100 codes, each prefixed with "FOO", followed by 10 characters from "ABCDEFGHKMNOPRSTUVWXYZ23456789" haystack and 2 numbers from "0123456789" haystack.
This PHP library looks promising: ReverseRegex
Like all of these, it only handles a subset of regular expressions but it can do fairly complex stuff like UK Postcodes:
([A-PR-UWYZ]([0-9]([0-9]|[A-HJKSTUW])?|[A-HK-Y][0-9]([0-9]|[ABEHMNPRVWXY])?) ?[0-9][ABD-HJLNP-UW-Z]{2}|GIR0AA)
Outputs
D43WF
B6 6SB
MP445FR
P9 7EX
N9 2DH
GQ28 4UL
NH1 2SL
KY2 9LS
TE4Y 0AP
You'd need to write a string generator that can parse regular expressions and generate random members of character ranges for random lengths, etc.
Much easier would be to write a random password generator with certain rules (starts with a lower case letter, has at least one punctuation, capital letter and number, at least 6 characters, etc) and then write your regex so that any passwords created with said rules are valid.
Presuming you have both a minimum length and 3-of-4* (or similar) requirement, I'd just be inclined to use a decent password generator.
I've built a couple in the past (both web-based and command-line), and have never had to skip more than one generated string to pass the 3-of-4 rule.
3-of-4: must have at least three of the following characteristics: lowercase, uppercase, number, symbol
It is possible (for example, Haskell regexp module has a test suite which automatically generates strings that ought to match certain regexes).
However, for a simple task at hand you might be better off taking a simple password generator and filtering its output with your regexp.