Section 10.4.


10.4. "Missing" Preg Functions

PHP's built-in preg functions provide a good range of functionality, but there have been times that I've found certain aspects a bit lacking. One example we've already seen is my special version of preg_match (˜ 454).

Another area where I've felt the need to build my own support functions involves situations where regular expressions are not provided directly in the program via literal pattern-argument strings, but brought in from outside the program (e.g., read from a file, or provided by a user via a web form). As we'll see in the next section, converting from a raw regular-expression string to a preg-appropriate patter n-argument can be tricky.

Also, before using such a regular expression, it's generally a good idea to validate that it's syntactically correct. We'll look into that as well.

As with all the code samples in this book, the functions on the coming pages are all available for download at my web site: http://regex. info /.





If you have a raw regular expression in a string (perhaps read from a configuration file, or submitted via a web form) that you'd like to use with a preg function, you must first wrap it in delimiters to make a preg-appropriate pattern argument.

10.4.1.

10.4.1.1. The problem

In many cases, converting a regular expression into a pattern argument is as simple as wrapping the regex with forward slashes . This would convert, for example, a regular-expression string ' [a-z]+ ' to ' /[a-z]+/ ', a string appropriate for use as a preg pattern argument.

However, the conversion becomes more complex if the regular expression actually contains the delimiter in which you choose to wrap it. For example, if the regex string is ' ^http://([^/:]+) ', simply wrapping it in forward slashes yields ' /^http://([^/:]+)/ ', which results in an "Unknown modifier / " error when used as a pattern modifier.

As described in the sidebar on page 448, the odd error message is generated because the first and second forward slashes in the string are taken as the delimiters, and whatever follows (in this case, the third forward slash in the string) is taken as the start of the pattern-modifier list.

10.4.1.2. The solution

There are two ways to avoid the embedded-delimiter conflict. One is to choose a delimiter character that doesn't appear within the regular expression, and this is certainly the recommend way when you're composing a pattern-modifier string by hand. That's why I used {‹} as the delimiters in the examples on pages 444, 449, and 450 (to name only a few).

It may not be easy (or even possible) to choose a delimiter that doesn't appear in the regex, because the text could contain every delimiter, or you may not know in advance what text you have to work with. This becomes a particular concern when working programatically with a regex in a string, so it's easier to simply use a second approach: select a delimiter character, then escape any occurrence of that character within the regex string.

It's actually quite a bit trickier than it might seem at first, because you must pay attention to some important details. For example, an escape at the end of the target text requires special handling so it won't escape the appended delimiter.

Here's a function that accepts a regular-expression string and, optionally , a pattern-modifiers string, and returns a pattern string ready for use with preg functions. The code's cacophony of backslashes (both regex and PHP string escapes ) is one of the most complex you're likely to see; this code is not light reading by any measure. (If you need a refresher in PHP single-quoted string semantics, refer to page 444.)

 /*     *  Given a raw regex in a string (and, optionally, a pattern-modifiers string), return a string suitable  *  for use as a preg pattern. The regex is wrapped in delimiters, with the modifiers (if any) appended  . */ function preg_regex_to_pattern($raw_regex, $modifiers = "") { /* *  To convert a regex to a pattern, we must wrap the pattern in delimiters (we'll use a pair of  *  forward slashes) and append the modifiers. We must also be sure to escape any unescaped  *  occurrences of the delimiter within the regex, and to escape a regex-ending escape  *  (which, if left alone, would end up escaping the delimiter we append)  . * *  We can't just blindly escape embedded delimiters, because it would break a regex containing  *  an already-escaped delimiter. For example, if the regex is '\/', a blind escape results  *  in '\/' which would not work when eventually wrapped with delimiters: '/\//'  . * *  Rather, we'll break down the regex into sections: escaped characters, unescaped forward  *  slashes (which we'll need to escape), and everything else. As a special case, we also look out  *  for, and escape, a regex-ending escape  . */ if (! preg_match('{\\(?:/;$)}', $raw_regex)) /* '/' followed by '\' or EOS */ { /*  There are no already-escaped forward slashes, and no escape at the end, so it's  *  safe to blindly escape forward slashes  . */ $cooked = preg_replace('!/!', '\/', $raw_regex); } else { /*  This is the pattern we'll use to parse $raw_regex  . *  The two parts whose matches we'll need to escape are within capturing parens  . */ $pattern = '{ [^\\/]+ \\. ( / \\$ ) }sx'; /*  Our callback function is called upon each successful match of $pattern in $raw-regex  . *  If $matches[1] is not empty, we return an escaped version of it  . *  Otherwise, we simply return what was matched unmodified  . */ $f = create_function('$matches', '     //  This long  if (empty($matches[1]))           //  singlequoted  return $matches[0];           //  string becomes  else                              //  our function  return "\\" . $matches[1];  //  code  . '); /*  Actually apply $pattern to $raw_regex, yielding $cooked  */ $cooked = preg_replace_callback($pattern, $f, $raw_regex); } /*  $cooked is now safe to wrap -- do so, append the modifiers, and return  */ return "/$cooked/$modifiers"; } 

This is a bit more involved than I'd like to recode each time I need it, which is why I've encapsulated it into a function (one I'd like to see become part of the built-in preg suite).

It's instructive to look at the regular expression used in the lower half of the function, with preg__replace_callback , and how it and the callback work to walk through the pattern string, escaping any unescaped forward slashes, yet leaving escaped ones alone.

10.4.2. Syntax-Checking an Unknown Pattern Argument

After wrapping the regex in delimiters, you've ensured that it's in the proper form for a preg pattern argument, but not that the original raw regex is syntactically valid in the first place.

For example, if the original regex string is ' *.txt 'perhaps because someone accidentally entered a file glob (˜ 4) instead of a regular expressionthe result from our preg_regex_to_pattern is /*.txt/ . That doesn't contain a valid regular expression, so it fails with the warning (if warnings are turned on):

 Compilation failed:   nothing to repeat at offset 0 

PHP doesn't have a built-in function to test whether a pattern argument and its regular expression are syntactically valid, but I have one for you below.

preg_pattern_error tests the pattern argument simply enough, by trying to use itthat's the one-line preg_match call in the middle of the function. The rest of the function concerns itself with PHP administrative issues of corralling the error message that preg_match might try to display.

 /*     *  Return an error message if the given pattern argument or     its underlying regular expression  *  are not syntactically valid. Otherwise (if they are valid),     false is returned  .     */     function preg_pattern_error($pattern)     {     /*     *  To tell if the pattern has errors, we simply try to use it  .     *  To detect and capture the error is not so simple, especially     if we want to be sociable and not  *  tramp on global state (e.g., the value of $php_errormsg).     So, if 'track_errors' is on, we preserve  *  the $php_errormsg value and restore it later.     If 'track_errors' is not on, we turn it on (because  *  we need it) but turn it off when we're done  .     */     if ($old_track = ini_get("track_errors"))     $old_message = isset($php_errormsg) ? $php_errormsg : false;     else     ini_set('track_errors', 1);     /*  We're now sure that track_errors is on  . */     unset($php_errormsg);     @  preg_match($pattern, "")  ; /*  actually give the pattern a try!  */     $return_value = isset($php_errormsg) ? $php_errormsg : false;     /*  We've now captured what we need; restore global state to what it was  . */     if ($old_track)     $php_errormsg = isset($old_message) ? $old_message : false;     else     ini_set('track_errors', 0);     return $return_value;     } 

10.4.3. Syntax-Checking an Unknown Regex

Finally, here's a function that utilizes what we've already developed to test a raw regular expression (one without delimiters and pattern modifiers). It returns an appropriate error string if the regular expression is not syntactically valid, and returns false if it is syntactically valid.

 /*     *  Return a descriptive error message if the given regular     expression is invalid  .     *  If it's valid, false is returned  .     */     function preg_regex_error($regex)      {         return preg_pattern_error(preg_regex_to_pattern($regex));     } 



Mastering Regular Expressions
Mastering Regular Expressions
ISBN: 0596528124
EAN: 2147483647
Year: 2004
Pages: 113

Similar book on Amazon

flylib.com © 2008-2017.
If you may any questions please contact us: flylib@qtcs.net