match a word that is not contained in parentheses

Question

hello, i am trying to split a string by some keywords that aren't contained in parentheses...

so, let's say i have the string "i will meet you where we talked (but not where he said)". i want to split the string into 2 pieces. one containing: "i will meet you", the other containing "we talked (but not where he said)". in other words, i want a regex to match the "where" that isn't in parentheses and ignore the one that is.

thank you very much!
all the best

p.s. i asked a related question yesterday, but i realised that i thinked the whole thing wrong... so now i ask the right question...

Answer 1

depends on how variable the input can be. If you are certain that it will always be in that format...(non-parenthetical-sentence followed by parenthetical sentence), you can be very simplistic and just search for the first "where".

Can you be more specific about what can be expected from the input? Can it have more than one parenthetical section? Can the sections be in any order?

Edit: It sounds like you can safely assume your input will be valid SQL. I would just search for all instances of "where" and then iterate through each one and count the number of open and closing parentheses before it. If the number of open parentheses before a given "where" string is equal to the number of close parentheses before it, it is the one you are looking for.

Answer 2

If you know that your "where" of interest will always happen first and that there will only be one "where" of interest, then this will work:

$array=explode('where',$string,2);

Answer 3

Try the following regex:

/(?<!\([^\)]*)where/

The (?<!...) is a negative lookbehind, which asserts that the string matched (where) must not be preceded by an open paren \( following by any number of characters that aren't a close paren [^\)]*.

Answer 4

It was proved way back in the 60s that traditioal Regexes can't solve your specific problem. Fortunately many new features have been added over the years. The feature you need is called "balancing groups", and allows a kind of a counter to be used inside a Regex.

Here is a simple solution to your problem:

^
(?'BeforeWhere'
  ((?'Open'\()|(?'Close-Open'\))|[^()])*  # count balanced parens
  (?(Open)^|)  # make sure parens actually balanced
)
(?'Where'\ where\ )
(?'AfterWhere'
  ((?'Open'\()|(?'Close-Open'\))|[^()])*  # count balanced parens
  (?(Open)^|)  # make sure parens actually balanced
)
$

This will result in BeforeWhere, Where and AfterWhere named groups. If you want to find multiple 'Where' instances outside of parenthesis you can add a ( before (?'Where' and a ) just before the final $

The only tricky part of this regex is making sure parens are actually balanced. In that line, if Open is still defined the parens are unbalanced so I match ^ which of course fails. Otherwise I match nothing, and the match proceeds.

Answer 5

i finnally got it solved with Brian Schroth solution... i made a php function that return every main part of a sql select statement (SELECT, FROM, JOINS, WHERE, GROUP BY, HAVING, ORDER BY, LIMIT) including any number of subqueries contained in that part... if anyone wants this function, i'll post it...

thank you all very much!