ksh string manipulation $##!? - shell

Question

So I have this spool file having this kind of content.
SQL> select file_name from dev_files;
FILE_NAME
------------------------------------------------------------------
file1.txt
file2.doc
file3.pdf
total.xls
4 rows selected.
SQL> spool off
I am writing a ksh script to load these files into a ftp server and update a log file. and I am stuck up bigtime. Here is the portion of my poor code after many tries.
dump="spoolfile.txt"
while read line;
do
if [[`expr match "$line" 'SQL'` !=3]] && [[`expr match "$line" 'FILE_NAME'` !=9]] && [[`expr match "$line" '---------'` !=9]]
then
ftp -inv $tgt_server <<EOT
quote user $uname
quote password $pword
mput $src_path/$line
quit
EOT
echo "sent $line" >> sent_files.log
fi
done < $dump
how do i ensure that "no rows selected" and say "4 rows selected." are not read? there can be any number instead of 4 corresponding to number of files. In the case of no files the spool file looks like this. the '.' is also missing.
SQL> select file_name from dev_files;
no rows selected
SQL> spool off

Answer 1

Here is a much more simple soltion: Instead of trying to find lines to ignore, mark the lines to process:
select concat("JHFDGFSH ",file_name) from dev_files
Now all you need to do is to ignore any line that doesn't start with JHFDGFSH.
[EDIT] If you can't do that: Start reading when you hit a line that's only ---- and stop at the first empty line. That should work unless Oracle starts to page.
And open an issue for upstream to tell them that their interface is brittle and that it will break eventually.
If that's OK, it's not your problem and not your fault when it breaks and production is down for a couple of days; just show the issue and say "see? You wanted it this way."
Also KSH is probably not the right tool; have a look at awk. This should work:
/^---+/,/^[ \t]*$/ { print; }
If you can't use awk, how about treating every line as a file name and ignore those which don't exist?
As long as no one creates a file named FILE_NAME, 4 rows selected. or SQL> spool off, you should be find as long as you make sure you properly quote all strings.

Answer 2

I agree with most of the comments about 'turn off headings' etc in the SQLPlus,
but this is Easy Peasy ;-)
while read line;
do
case ${line} in
-------* ) continue ;;
FILE_NAME ) continue ;;
SQL[>] ) continue ;;
*rows\ selected ) continue ;;
esac
# with enough case match targets, you can probably eliminate
# the if/fi completely. Good luck.
if [[`expr match "$line" 'SQL'` !=3]] .....
ftp ...
fi
done < $dump
You'll have to experiment with what values to put in the case statement to clean it up completely.
Also note that for any unusual characters, like '>', it is probably better to include them as a character class [>] (again requiring some experimentation on your part)
And finally, note that if you want to include white-space chars in the case match targets, either surround the phrase in dbl-quotes (") or escape each WS char like '\'.