Regex - Operator "." - Meta character capture

Question

Situation

I’m conducting a survey with regex in a specific word inválido, but I decided to use inv.lido. Which I knew I had in the test string but didn’t return.

Testing

vr = var_dump
pr = print_r

$string = 'até, atenção, Hipótese, você, português, café, órgão';

vr(preg_match('~at.~', $string, $match));
pr($match);
vr(preg_match('~aten..o~', $string, $match));
pr($match);
vr(preg_match('~Hip.tese~', $string, $match));
pr($match);
vr(preg_match('~voc.~', $string, $match));
pr($match);
vr(preg_match('~portugu.s~', $string, $match));
pr($match);
vr(preg_match('~caf.~', $string, $match));
pr($match);
vr(preg_match('~.rg.o~', $string, $match));
pr($match);

Out

int(1)
Array([0] => at�)

int(0)
Array()

int(0)
Array()

int(1)
Array([0] => voc�)

int(0)
Array()

int(1)
Array([0] => caf�)

int(0)
Array()

Question

As you can see he didn’t capture the words, except for a few, but even the ones he captured, I don’t know what the �, because even using utf8_decode or even utf8_encode it does not return me the correct character.

From what little I know C and binary, I suppose it has to do with the fact that these characters are formed by two 8bit squares, however they are present in the table ASCII and as far as I know regex follows the table ASCII.

Why did this happen?

Answer 1

PHP regular expressions do not support Unicode by default, unless you use the flag u:

preg_match('~aten..o~', $string, $match);
print_r($match);

Array ( )

preg_match('/aten..o/u', $string, $match);
print_r($match);

Array ( [0] => attention )

Example in the ideone.

What results you are getting (e.g..: at�), the reason is that accented characters are usually represented by more than one byte, for example in UTF-8 encoding. A pattern:

at.

without the flag u will marry 3 bytes, the first one a, the second one t and the third the first byte of é. Since this first byte is not a valid ASCII (nor Unicode) character, the function print_r does not know how to represent it, so prints a �. Already the standard:

aten..o

When applied to the word atenção house the first . with the first byte of the ç, the second point with the second byte of the ç, and when he tries to marry the o with the first byte of the ã cannot, and marriage fails.

By activating the flag u, the engine takes full characters (and not just bytes) into account in the match, so that the first point matches ç, the second with ã, and the result is right as expected.

Answer 2

As already quoted by @mgibsonbr, by default, PHP does not support unicode in regular expressions of preg.

In addition to the solution already presented, what can be done is to use regular expression functions of Multibyte String.

Example:

$str = 'inválido';

var_dump(mb_ereg_match('inv.lido', $str)); // bool(TRUE);

Observing:

According to that answer in SOEN, the functions mb_ereg_* are not marked as obsolete. Therefore, it is OK to use them.