4

I'm using the "fuzzy match" functionality of the Regex module.

How can I get the "fuzziness value" of a "match" which indicates how different the pattern is to the string, just like the "edit distance" in Levenshtein?

I thought I could get the value in the Match object, but it's not there. The official docs said nothing about it, neither.

e.g.:

regex.match('(?:foo){e}','for')

a.captures() tells me that the word "for" is matched, but I'd like to know the fuzziness value, which should be 1 in this case.

Is there any way to achieve that?

Casimir et Hippolyte
  • 88,009
  • 5
  • 94
  • 125
tslmy
  • 628
  • 1
  • 6
  • 21
  • It's certainly not ideal, but if all else fails you could try repeated attempts with `(?:foo){e<=i}` where you loop over some integer `i`. The first time you get a match, your `i` is the Levenshtein distance. – Martin Ender Jun 10 '13 at 12:53
  • Or if you are working with a limited number of errors you could use something like `(foo)|((?:foo){e=1})|((?:foo){e=2})` and check which group matched, if first e = 0, second e = 1, etc. – Qtax Jun 10 '13 at 13:10

2 Answers2

2
>>> import difflib
>>> matcher = difflib.SequenceMatcher(None, 'foo', 'for')
>>> sum(size for start, end, size in matcher.get_matching_blocks())
2
>>> max(map(len, ('foo', 'for'))) - _
1
>>>
>>>
>>> matcher = difflib.SequenceMatcher(None, 'foo', 'food')
>>> sum(size for start, end, size in matcher.get_matching_blocks())
3
>>> max(map(len, ('foo', 'food'))) - _
1

http://docs.python.org/2/library/difflib.html#difflib.SequenceMatcher.get_matching_blocks http://docs.python.org/2/library/difflib.html#difflib.SequenceMatcher.get_opcodes

falsetru
  • 357,413
  • 63
  • 732
  • 636
  • Errrr... This workaround is great, and difflib is good stuff, too. But I have to use the "groupdict" thing in re/regex, so, I don't think difflib is suitable for that. Thanks all the same. – tslmy Jun 10 '13 at 13:42
  • I replaced the regex module with the difflib like you said, and it works like a charm! So, I wonder is this "SequenceMatcher" thing comparable to the Levenshtein algrithm? – tslmy Jun 13 '13 at 15:27
  • In [source code](http://hg.python.org/cpython/file/464e8fd7300d/Lib/difflib.py), there is no mention about Levenshtein. (I didn't read throughly.) – falsetru Jun 13 '13 at 15:34
0
a = regex.match('(?:foo){e}','for')
a.fuzzy_counts

this returns a tuple (x,y,z) where:

x = number of substitutions

y = number of insertions and

z = number of deletions

But this is not always a reliable count, ie: the regex match fuzziness night not equate to the true Levinstein distance in some cases

Python regex module fuzzy match: substitution count not as expected

Community
  • 1
  • 1
Colin Anthony
  • 1,141
  • 12
  • 21