Tricky big data-processing problems (regular expressions)

Some time ago, I collected more than 40,000 wallpaper data, format similar to the following:

-------------------a
1
2
3
-------------------b
AK
al
am
an
ao
AP
aq
ar
-------------------C
1
/
71

Because to guide the MySQL database, I want the final format is this:

A-1
a-2
a-3

b-ak
b-al
b-am

b-an b-ao b-ap b-aq b-ar c-1
c-11
c-21
c-31
c-41
c-51
c-61

With Sublime Text3 the regular, and even the recursion has learned in sublime text3 or can not be achieved.
(doing, killing people, even with Excel is trouble.) ）

Finally I wrote a PHP to deal with this thing:

<?php
$headPattern  = '-* (\w) ';          Match head
$childPattern = ' \ r \ n (\d+|\w+) ';    Match body
$pattern 	  = "/{$headPattern} ((?: $childPattern) (? 2) *)/";
$pattern   = "/-* (\w) (?: \ r\n (\d+|\w+)) (? 2) *)/";

$subject = "
-------------------a
1
2
3
4
5
6 7 8------------ -------b
AK
al
am
an
ao
ap
aq
ar
-------------------c
1 One by one
;

Echo preg_replace_callback ($pattern, function ($e) use ($childPattern) {
    //two-time regular match
    preg_match_all ('/'. $ Childpattern. '/', $e [2], $match);
    $result = ';
    foreach ($match [1] as $key => $value) {
        $result. = $e [1]. '-' . $value. "\ n";
    }
    return $result;
}, $subject);

Do not parse, look at the code.