PHP 5.6正则表达式意外行为

Question

I have come across a strange behaviour in PHP 5.6 (not tested with other versions)

var_dump(preg_match('#\b(39||90)\b#', '42')); // int(1)
var_dump(preg_match('#\b(39||90)\b#', '')); // int(0)

https://regex101.com says the pattern \\b(39||90)\\b is invalid but PHP preg_match does not return FALSE as it should if the pattern is invalid.

As you can see 42 produces a match and the empty string produces a non-match. I'd expect the other way round as || should stand for empty string.

What's happening here?

Answer 1

This regex:

\b(39||90)\b

Will return a successful match if any of the alternation is matched. These are:

1. Complete word 39
2. Complete word 90
3. A word boundary anywhere in the input (because of empty || )

However in empty string there is no word boundary. A word boundary \\b is asserted true between a word \\w and a non-word \\W .

Eg see these results:

// no word character hence false
var_dump(preg_match('#\b(39||90)\b#', '#@'));
int(0)

# a word char hence true
php > var_dump(preg_match('#\b(39||90)\b#', 'a'));
int(1)

// no word character hence false
php > var_dump(preg_match('#\b(39||90)\b#', "\t\n"));
int(0)