3

I have a long list of regular expressions that should be tested. Instead of walking through the list one by one and writing an applicable test for each expression - if possible - I want to create a generator function that will consider the type of regular expression and generate relevant input string.

For example: 

rgx = re.compile(r'^item_(?P<item_number>\d+)$')

Is it possible to determine the type of input that a regex would match? Considering the above example, is it possible to determine the input (such as item_23567) that the regex needs from the compiled regular expression?

Sufian Latif
  • 13,086
  • 3
  • 33
  • 70
altunyurt
  • 2,821
  • 3
  • 38
  • 53
  • Related: http://stackoverflow.com/questions/2905229/reverse-regular-expressions-to-generate-data and http://stackoverflow.com/questions/205411/random-string-that-matches-a-regexp – Alex L Jan 19 '12 at 14:59
  • It's not clear what you mean. What would the generator function look like? – Fred Foo Jan 19 '12 at 15:01
  • 3
    If you want to test the regular expressions, wouldn't generating the expected in and outputs from the regular expression entirely defeat the point of testing them in the first place? – klennepette Jan 19 '12 at 15:36
  • i have a bloated django project to fix, and don't know which parts of it works or not. For it is a huge project, the only way seems to me is to test the views by generating the urls on the fly according to regexes in the url patterns. – altunyurt Jan 19 '12 at 16:18
  • 1
    @hinoglu: It's better if you describe your **real** problem, rather than some approach might seem workable. You want to generate sample URL's that match the paths in your Django URL's? Is that the **real** question? You might want to close this and ask the **real** question. Django's `reverse()` function is the first thing to read, if that's your question. – S.Lott Jan 19 '12 at 17:04
  • Generating a variety of valid matches from a pattern could be very informative - it could be quickly obvious that there are problems, in some cases. eg if the + was a * – Highly Irregular Jan 23 '12 at 01:44

2 Answers2

0

This answer does a good job of defining the problem abstractly: https://stackoverflow.com/a/208112/1092724

Python solutions: Reversing a regular expression in Python

A Java library solution: http://code.google.com/p/xeger/

These solutions are not without limitations: see http://code.google.com/p/xeger/wiki/XegerLimitations

Pulling from a comment made on the Xeger limitations page

"it is worth noting that the current implementation of xeger randomly traverses an fsm to generate its strings. if you have regular expressions with characters to match after a ., transitions are added back to the state for . for every single state afterwards. it becomes increasingly improbable with each state after the .* that xeger will ever reach an accepting state."

(@Highly Irregular) voiced a similar concern in their comment.

This is a common theme among all implementations you will find. The solutions will only work well with carefully constructed expressions. Otherwise your code may go on an infinite Journey, never finding what you seek. If what you need is indeed regular and well defined, one of these approaches may work for you.

Community
  • 1
  • 1
coagulateblue
  • 23
  • 4
0

impossible in general sense for extended regular expressions, see e.g.

Do all regular expressions halt?

most importantly if you had an oracle that generated some obscure, convoluted text that matched any regular expression, who would you ever use it as a test? how can you ever know if given megabyte of text rubbish is in fact what the given regexp was designed to match or should match?

Community
  • 1
  • 1
Dima Tisnek
  • 11,241
  • 4
  • 68
  • 120