CODEWITHSUNDEEP
javascriptregex
webwires
webwires

Reputation: 1907

Regular Expression "AND"

I'm doing some basic text matching from an input. I need the ability to perform a basic "AND". For "ANY" I split the input by spaces and join each word by the pipe ("|") character but I haven't found a way to tell the regular expression to match any of the words.

switch (searchOption) {
  case "any":
    inputArray = input.split(" ");
    if (inputArray.length > 1) { input = inputArray.join("|"); }
    text = input;
    break;
  case "all":
    inputArray = input.split(" ");
    ***[WHAT TO DO HERE?]***
    text = input;
    break;
  case "exact":
    inputArray = new Array(input);
    text = input;
    break;
}

Seems like it should be easy.

Upvotes: 4

Views: 401

Answers (4)

b w
b w

Reputation: 4663

Use lookahead. Try this:

if( inputArray.length>1 ) rgx = "(?=.*" + inputArray.join( ")(?=.*" ) + ").*";

You'll end up with something like

(?=.*dog)(?=.*cat)(?=.*mouse).*

Which should only match if all the words appear, but they can be in any order.

  • The dog ate the cat who ate the mouse.
  • The mouse was eaten by the dog and the cat.
  • Most cats love mouses and dogs.

But not

  • The dog at the mouse.
  • Cats and dogs like mice.

The way it works is that the regex engine scans from the current match point (0) looking for .*dog, the first sub-regex (any number of any character, followed by dog). When it determines true-ness of that regex, it resets the match point (back to 0) and continues with the next sub-regex. So, the net is that it doesn't matter where each word is; only that every word is found.

EDIT: @Justin pointed out that i should have a trailing .*, which i've added above. Without it, text.match(regex) works, but regex.exec(text) returns an empty match string. With the trailing .*, you get the matching string.

Upvotes: 7

Gumbo
Gumbo

Reputation: 655219

Do a simple for loop and search for every term, something like this:

var n = inputArray.length;
if (n) {
    for (var i=0; i<n; i++) {
        if (/* inputArray[i] not in text */) {
            break;
        }
    }
    if (i != n) {
        // not all terms were found
    }
}

Upvotes: 3

wds
wds

Reputation: 32283

My regular expressions cookbook does feature a regular expression that can possibly do this using conditionals. However, it's quite complicated, so I'd go for the currently top rated answer which is iterating over the options. Anyway, trying to adapt their example I think it would be something like:

\b(?:(?:(word1)|(word2))(\b.*?)){2,}(?(1)|(?!))(?(2)|(?!))

No guarantees that this'll work as is, but it's the basic idea I think. See what I mean about complicated?

Upvotes: 1

VoteyDisciple
VoteyDisciple

Reputation: 37803

The problem with "and" is: in what combination do you want the words? Can they appear in any order, or must they be in the order given? Can they appear consecutively or can there be other words in between?

These decisions impact heavily what search (or searches) you do.

If you're looking for "A B C" (in order, consecutively), the expression is simply /A B C/. Done!

If you're looking for "A foo B bar C" it might be /A.*?B.*?C/

If you're looking for "B foo A foo C" you'd be better off doing three separate tests for /A/, /B/, and /C/

Upvotes: 3

Related Questions