Once upon a time, way back in the mists of time (i.e. sometime in 1990 :-) I started my first version of "sybperl" like this: write the SQL to a temp file, run isql and parse the output. Then I got an email from Larry about using pseudo-TTYs and running isql as a sub-processes - I still have that code somewhere (1990 vintage perl...) if anyone is interested.

Assuming that this is Sybase (maybe MS-SQL has it as well) you should use the "-s" command-line argument to specify a column separator that is unlikely to occur in the data. You should also specify the "-w" argument with a large integer to make sure that isql doesn't wrap any of the output lines. So something like this could work:

open(TMP, ">/tmp/t$$.sql") || die "can't open temp file: $!";
# Set nocount to avoid the "xxx rows affected" message in isql
print TMP "set nocount on\ngo\nselect * from the_table\ngo\n";
close(TMP);
# Note: -b flag suppresses column headers.
open(ISQL, "isql -Ujoe -Pjoespwd -Smyserver -s'%%^^%%' -w 1024 -b -i /
+tmp/t$$.sql") || die "Can't run isql: $!";
while(<ISQL>) {
    chomp;
    my @row = split(/%%^^%%/, $_);
    .... do something with the @row of data....
}
close(ISQL);
[download]

That being said - using DBI and the appropriate DBD driver is a lot better. It's faster, and error handling is much easier.

Michael