views:

152

answers:

3

This is similar to my previous efforts (wordEnds and repeatEnd): as a mental exercise, I want to solve this toy problem using regex only.

Description from codingbat.com:

Given a string and a non-empty word string, return a version of the original string where all chars have been replaced by pluses ("+"), except for appearances of the word string which are preserved unchanged.

plusOut("12xy34", "xy") → "++xy++"
plusOut("12xy34", "1") → "1+++++"
plusOut("12xy34xyabcxy", "xy") → "++xy++xy+++xy"

There is no mention whether or not to allow overlap (e.g. what is plusOut("+xAxAx+", "xAx")?), but my non-regex solution doesn't handle overlap and it passes, so I guess we can assume non-overlapping occurrences of word if it makes it simpler (bonus points if you provide solutions for both variants!).

In any case, I'd like to solve this using regex (of the same style that I did before with the other two problems), but I'm absolutely stumped. I don't even have anything to show, because I have nothing that works.

So let's see what the stackoverflow community comes up with.

A: 

I think you could leverage a negated range to do this. As this is just a hint, it's not tested though!

Turn your "xy" into a regexp like this: "[^xy]"

...and then wrap that into a regexp which replaces strings matched by that expression with "+".

Carl Smotricz
That won't work; `"xy"` _string_ needs to occur exactly to be kept; not just `x`, not just `y`. Please test your solution on codingbat if possible.
polygenelubricants
Oops, you're right. I hastily read those strings as being sets of characters you want matched, not complete strings. Sorry, I give up!
Carl Smotricz
Thanks for trying! I really do appreciate it! Most people would just ignore these silly little exercises!
polygenelubricants
+3  A: 

This passes all their tests:

public String plusOut(String str, String word) {
  return str.replaceAll(
    String.format("(?<!(?=\\Q%s\\E).{0,%d}).", word, word.length()-1),
    "+"
  );  
}

Also, I get:

plusOut("1xAxAx2", "xAx") → "+xAxAx+"

If that's the result you were looking for then I pass your overlap test as well, but I have to admit, that one's by accident. :D

Alan Moore
I was stuck playing around with looking behind for `\G` (if previous character wasn't a match, then you also can't match if you're within the span of `word`, etc), which _perhaps_ can solve the non-overlap variant. Also previously I also tried something like yours but with the mistake of putting the positive lookahead AFTER the finite rep (which is also the problem I had before with `repeatBegin`). Anyway, good job again! I wish I can double-upvote this!
polygenelubricants
One question: how safe is using `\Q` and `\E` to quote arbitrary input string? Your solution would break if the `word` is `"\\E"`, for example.
polygenelubricants
@poly: Yes, to be perfectly safe I should have used the `quote()` method, as you did in your `wordEnds` solution. It actually works by adding the `\Q` and `\E`, but it also escapes any `\E` that might already be present.
Alan Moore
A: 

This is provided here just for reference. This is essentially Alan's solution, but using replace instead of String.format.

public String plusOut(String str, String word) {
  return str.replaceAll(
    "(?<!(?=word).{0,M})."
      .replace("word", java.util.regex.Pattern.quote(word))
      .replace("M", String.valueOf(word.length()-1)),
    "+"
  );  
}
polygenelubricants