Shell Programming and Scripting

Hello All,

I have a following task that I need to accomplish through a script or program and I am looking for some help as I have exhausted my ideas.

1. given: a text file with thousands of lines
2. find: pattern A in file and get line number ( grep -n works)
3. find: the first occurence of pattern B up from line number obtained in step2
4. find: pattern A again but after the line number previously found
5. find: the first occurence of pattern B up from the new line number in step4
6. repeat the process until the file is completely processed.

Possible solutions?
a) will vi work programmatically with forward search '/', go to line 'nG' and reverse searching '?' ?
b) I think C program will work with file pointer, seek and other facilities. But I have no C compiler.
c) will java work - RandomAccessFile - seek?
d) how about awk? not sure if it has reverse searching capability.
e) how about a combination of grep -n file, head -n, tac |grep

this may be a possibility but could be tedious with a few intermediary files

Please let me know if you have any suggestions on how I could go about this.
Thanks.

I need to clarify step 3 and 5.

suppose I find pattern A on line 25 of the input text file, now I need to find the first occurence of pattern B from lines 25 through 1 of the file - in reverse. Say pattern B occurs on line 20 and line 15 and line 10. I need line 20, but not lines 15 & 10.

Thanks.

If I understand the logic correctly, wouldn't it suffice to remember the B hits and rattle off one each time you see an A? This is basically a two-liner in Perl, I guess it could be done in awk as well but the lack of ordered arrays is a bit of a drag. (Or does "first up" in steps 3 and 5 mean the latest B before the current A? Then it's even easier, and doesn't require an array.)

"Or does "first up" in steps 3 and 5 mean the latest B before the current A?"

Yes, I only need to get the latest pattern B preceeding the current pattern A.

Thanks.

Code:

awk '/B/ { lastb=$0 } /A/ { print; print lastb; lastb="" }' file

Code:

awk '/B/{a=$0}/A/ && a {print a;a=""}' file

a will be the last occurrence of B.

Intuitive (-;

Actually the && a is a useful addition, I meant to remember to not forget it, but failed. But you're not printing the A lines at all, did I misunderstand that part of the requirements? (Also actually not clear whether the matches should be printed or somehow collected, say, just collect the line numbers for later processing? Easy enough to do, NR is the current line number in awk.)