Question

I need to get content of executable comments in mysqldump results, but for regexp

/\/\*\!\d+\s+(.*?)\*\//s

and input data like this:

/*!50003 text
some text else
/*
comment
also comment
*/
text...
and also text...
*/

I get wrong result because it get data only from “text” to “also comment” lines. How can I to skip comment into comment?
Thanks.

UPD: I cannot use “^” and “$” to mark start and end of input, because I have a lot ot executable statements in input.

UPD2: Output I want:

text
some text else
/*
comment
also comment
*/
text...
and also text...

NOT all input how in comment below. It’s very strange, I think, get the same output as input.

UPD3:
Start of executable comment must be /*!ANYNUMBER. It must be skipped and not included in output. End of executable comment is simply */ Right output example is presented in “UPD2”.

Answer 1

Pure regular expressions can’t handle nesting, but PHP’s flavor can by using recursion. Using the PCRE_EXTENDED modifier so we can have whitespace and comments:

%(               # opening RE delimiter, group start
  /\*            # comment open marker
    (  [^/*]     # non comment-marker characters
     | /(?!\*)   # '/' not followed by '*', so not a comment open
     | \*(?!/)   # '*' not followed by '/', so not a comment close
     | (?R)      # recursive case
    )*           # repeat any number of times
  \*/            # comment close marker
)%x              # group end, closing RE delimiter, PCRE_EXTENDED

In short:

%(/\*([^/*]|/(?!\*)|\*(?!/)|(?R))*\*/)%x

In use:

<?php

$commentRE = '%(/\*([^/*]|/(?!\*)|\*(?!/)|(?1))*\*/)%';
$doc = <<<EOS

USE database;

/* comment
and a
/* nested comment /* me too */
   now exiting
 */
the comment */


/*!50003 text
some text else
/*
comment
also comment
*/
text...
and also text...
*/

CREATE TABLE IF NOT EXISTS ...

EOS;

preg_match_all($commentRE, $doc, $parts);
var_export($parts[0]);

Result:

array (
  0 => '/* comment
  and a
  /* nested comment /* me too */
     now exiting
   */
  the comment */',
  1 => '/*!50003 text
  some text else
  /*
    comment
    also comment
  */
  text...
  and also text...
*/',
)