Confused about lookbehind behaviour-CodePudding

In the following code:

$a = ['aaa_bbb', 'aaa_bbb_ccc_AAA', 'aaa_bbb_ccc_BBB', 'aaa_bbb_ddd'];

foreach ($a as $s) {
    if (preg_match('/aaa_bbb(?:_.*)?_.*$/', $s, $g)) {
        print_r($g[0].PHP_EOL);
    }
}
echo '====='.PHP_EOL;
foreach ($a as $s) {
    if (preg_match('/aaa_bbb(?:_.*)?_.*$(?<!_ccc_BBB)/', $s, $g)) {
        print_r($g[0].PHP_EOL);
    }
}

which outputs

aaa_bbb_ccc_AAA
aaa_bbb_ccc_BBB
aaa_bbb_ddd
=====
aaa_bbb_ccc_AAA
aaa_bbb_ddd

WHY does the second loop successfully ignores the 'aaa_bbb_ccc_BBB' string?

How does the lookbehind (?<!_ccc_BBB) knows that it should start AFTER aaa_bbb? It was my understanding that lookbehinds "specify a group that can not match before the main expression" (from RegExr Reference section). I thought the main expression was the entire RegEx.

CodePudding user response：

The lookbehind, as its name suggests, looks behind the current position in the string. In your case, it does not look "after aaa_bbb", instead it looks before the end of the line (the $ marker).

The regular expression can be seen as follows:

aaa_bbb                            # literally, this
       (?:_.*)?                    # an optional and noncaptured underscore stuff
               _.*                 # a compulsory underscore stuff
                  $                # end of line
                   (?<!_ccc_BBB)   # last 8 chars must not be these.