JavaScript Alternative To Possessive Quantifiers I

问题:

I'm using JavaScript to extract a subset of "siblings" from a comma-delimited string of members I call a "generation" string.

Metaphorically speaking, the members are all from the same generation, but they are not all siblings (from the same parents). Here's an example:

// This is the generation string to search
var generation  = 'ABAA,ABAB,ABAC,ABAD,ABBA,ACAA,ACAB,ACAD,AEAB,AEAD,AFAA';

// This is the member for whom to extract siblings (member included)
var member      = 'ACAA';

The generation string and its members have the following characteristics:

• Each member has the same number of characters as the others
• All members of the string are alpha sorted
• Each set of siblings will always be adjacent to one another
• Siblings are those members who share the exact same combination of letters except the last letter

Continuing the example...

// This is how I go about extracting the desired result: ACAA,ACAB,ACAD
var mParent     = member.substr(0, member.length - 1) ;
var mPattern    = mParent + '[A-Z]';
var mPattern    = '(.*)((' + mPattern + ')(,$1)*)(.*)'; // Trouble is here
var mRegex      = new RegExp(mPattern);
var mSiblings   = generation.replace(mRegex, '$2');

The trouble spot identified above concerns regex quantifiers in the constructed pattern. As it is above, everything is set to greedy, so the value of mSiblings is:

ACAD

That's only the last member. Changing mPattern to be less greedy in hopes of extracting the other members yields the following

// Reluctant first expression yields ACAA
var mPattern = '(.*?)((' + mPattern + ')(,$1)*)(.*)'; 

// Reluctant last expression yields ACAD,AEAB,AEAD,AFAA
var mPattern = '(.*)((' + mPattern + ')(,$1)*)(.*?)'; 

// Reluctant first and last yields ACAA,ACAB,ACAD,AEAB,AEAD,AFAA
var mPattern = '(.*?)((' + mPattern + ')(,$1)*)(.*?)';

If I could make the middle expression possessive, this would be problem solved. Something like this:

// Make as many "middle" matches as possible by changing (,$1)* to (,$1)*+
var mPattern = '(.*?)((' + mPattern + ')(,$1)*+)(.*?)';

But as I have read (and have the syntax errors to prove it), JavaScript doesn't support possessive regular expression quantifiers. Can someone suggest a solution? Thank you.

回答1:

The most obvious problem is the $1. Within a regex, you would refer to capturing group #1 using \1, not $1. The (,$1)* in your regex is never going to match anything. But a group reference isn't going to do any good anyway.

When you use a group reference in a regex, you aren't applying that part of the regex again, you're simply matching the same thing that it matched the first time. That is, (ACA[A-Z])(,\1)* will match ACAA,ACAA, but not ACAA,ACAB or ACAA,ACAC. If you want to do that, you need to repeat the actual regex: (ACA[A-Z])(,ACA[A-Z])*. Since you're generating the regex dynamically, that shouldn't be a problem.

Note that that's the whole regex: ACA[A-Z](,ACA[A-Z])*. There's no need to match the stuff preceding or following the part that interests you; that's just making the job more complicated (and the results more confusing). You can access the match result directly, instead of using that "replace" gimmick:

var match = mRegex.exec(generation);
if (match != null) {
    mSiblings = match[0];
}