Have you tried something until now? the answer hevily depends on rules of the pattern and the string composition.

But something almost simple like this can suffice (pay attention to the [^AX] part: it may need to be more elaborate).

# warning win32 doublequotes oneliner                 
perl -e " print $1 if $ARGV[0]=~/(AAA)[^AX].+XXX/"    ABBBAAACCCXXX
AAA
[download]

that explained by YAPE::Regex::Explain becomes:

perl -MYAPE::Regex::Explain -e " print YAPE::Regex::Explain->new(qr/(A
+AA)[^AX].+XXX/)->explain(); "
The regular expression:

(?-imsx:(AAA)[^AX].+XXX)

matches as follows:

NODE                     EXPLANATION
----------------------------------------------------------------------
(?-imsx:                 group, but do not capture (case-sensitive)
                         (with ^ and $ matching normally) (with . not
                         matching \n) (matching whitespace and #
                         normally):
----------------------------------------------------------------------
  (                        group and capture to \1:
----------------------------------------------------------------------
    AAA                      'AAA'
----------------------------------------------------------------------
  )                        end of \1
----------------------------------------------------------------------
  [^AX]                    any character except: 'A', 'X'
----------------------------------------------------------------------
  .+                       any character except \n (1 or more times
                           (matching the most amount possible))
----------------------------------------------------------------------
  XXX                      'XXX'
----------------------------------------------------------------------
)                        end of grouping
----------------------------------------------------------------------
[download]

L*

Your example suggests that the entire search is to be performed on data appearing once in a single line. If so, you have a fine suggestion above.

But if your problem case involves data that is multi-line with more than one instance of each target and reference datum (or similarly, if the data is all on a single line but may contain repetitions of the target and reference data) -- perhaps like this:

YYY
AAA
BBB
CCC
AAA
freida
XXX
GGG
FFF
AAA.
XXX
ozymandis
BBB...
[download]

... then you'll need to inspect each line for your initial string, and save it (and perhaps, its index if that's going to be useful as you go forward, which seems likely as you've stated your question) or a reference (etc) before continuing to look for your "reference data" discarding any intervening data.

Then process the saved hit (AAA) before continuing the search for additional instances of AAA before XXX.

Your example suggests that the entire search is to be performed on data appearing once in a single line.

This is due to missing <code> tags. As evidenced by the HTML source code, the OP is as follows:

---

Hi,

I would like to know how to grep the first occurrence of a string above(upwards) a reference string.

for (e.x.)in the below set, I would like to get the first occurrence of "AAA" above "XXX".

AAA
BBB
AAA
CCC
XXX

Thanks,
Venkat

---

To the OP - if you rephrase your problem, you have the solution almost there:

I want to keep track of a pattern I seek for, and output the found string after finding a reference string

which is what ww did for you above already. Hence a possible solution is

#!/usr/bin/perl
# file grep.pl

my $sought    = shift; # from @ARGV
my $reference = shift;
$sought && $reference or die "usage: $0 soughtstring refstring files\n
+";

my $found;
while(<>) {
    chop; # strip newline

    /$sought/ and $found = $_;

    # or, if you want the very first occurence, don't
    # overwrite the variable (see perlop for '||='):
    # /$sought/ and $found ||= $_;

    if (/$reference/) {
        print "$ARGV: '$found'\n" if $found;
        $found = '';
    }
}
[download]

to be used as

$ perl grep.pl AAA XXX example.txt
[download]

which invoked upon this example.text

  1 YYY
  2 first AAA
  3 BBB
  4 CCC
  5 second AAA
  6 freida
  7 XXX
  8 GGG
  9 FFF
 10 third AAA
 11 XXX
 12 ozymandis
 13 BBB
 14 blorflydick
 15 XXX
 16 fourth AAA
[download]

produces this:

example.txt: '  5 second AAA'
example.txt: ' 10 third AAA'
[download]

Lines 2 and 16 are not printed. If you use the commented alternative, line 2 is printed instead of line 5.

perl -le'print map{pack c,($-++?1:13)+ord}split//,ESEL'