The issue is fundamental.

You clearly understand SQL but have a problem processing the results of the query in unix Shell.

The problem is:
All the output from the SQL session is going into one environment variable called $sql_rows. There is no record structure and no record delimiters - just a very long list of fields. When you do "for sql_records in $sql_rows" it is reading each field one-by-one (left-to-right on a very wide row).

I would recommend using a SQL "spool filename" statement before the query and a "spool off" statement after the query in order to output the results of the query to a disc file (not a Shell environment variable). With judicious use of SQL commands early in the script like "set feedback off" we can reduce the amount of rubbish appearing in the file. I notice that you have "-s" (silent) on the command line - it all helps.

The file output by "spool filename" will have a proper record structure (one line per hit on the database).
If you post the contents of the file (in code tags) the scripting wizards on this board should be able to fish out the filesystem name, run the appropriate "df" and present the data in an organised manner.

The script is finding out where the database segments reside from Oracle system tables and checking the related disc space. I like the concept of this script and I think that a good clean solution should go into the FAQ on this board.

I have written a huge commercial quality script to monitor multiple Oracle databases and alert if new segments will be required in the forseeable future. It has certain commonality with your script but my SQL side goes a lot further into finding the "high water" mark of the database within multiple fixed size segments. In the commercial world we would never let an Oracle database extend dynamically.


Ps. Whenever you post Oracle questions, please mention the word "Oracle" and state:
Operating System and version.
Oracle version.
What Shell you use.