How do I spool to a CSV formatted file using SQLPLUS?

Question

I want to extract some queries to a CSV output format. Unfortunately, I can\'t use any fancy SQL client or any language to do it. I must use SQLPLUS.

How do I do it?

Answer 1

You could also use the following, although it does introduce spaces between fields.

set colsep ,     -- separate columns with a comma
set pagesize 0   -- No header rows
set trimspool on -- remove trailing blanks
set headsep off  -- this may or may not be useful...depends on your headings.
set linesize X   -- X should be the sum of the column widths
set numw X       -- X should be the length you want for numbers (avoid scientific notation on IDs)

spool myfile.csv

select table_name, tablespace_name 
  from all_tables
 where owner = 'SYS'
   and tablespace_name is not null;

Output will be like:

    TABLE_PRIVILEGE_MAP           ,SYSTEM                        
    SYSTEM_PRIVILEGE_MAP          ,SYSTEM                        
    STMT_AUDIT_OPTION_MAP         ,SYSTEM                        
    DUAL                          ,SYSTEM 
...

This would be a lot less tedious than typing out all of the fields and concatenating them with the commas. You could follow up with a simple sed script to remove whitespace that appears before a comma, if you wanted.

Something like this might work...(my sed skills are very rusty, so this will likely need work)

sed 's/\s+,/,/' myfile.csv 

Answer 2

There is a problem using sqlplus to create csv files. If you want the column headers only once in the output and there are thousands or millions of rows, you cannot set pagesize large enough not to get a repeat. The solution is to start with pagesize = 50 and parse out the headers, then issue the select again with pagesize = 0 to get the data. See bash script below:

#!/bin/bash
FOLDER="csvdata_mydb"
CONN="192.168.100.11:1521/mydb0023.world"
CNT=0376
ORD="0376"
TABLE="MY_ATTACHMENTS"

sqlplus -L logn/pswd@//${CONN}<<EOF >/dev/null
set pagesize 50;
set verify off;
set feedback off;
set long 99999;
set linesize 32767;
set trimspool on;
col object_ddl format A32000;
set colsep ,;
set underline off;
set headsep off;
spool ${ORD}${TABLE}.tmp;
select * from tblspc.${TABLE} where rownum < 2;
EOF
LINES=`wc -l ${ORD}${TABLE}.tmp | cut -f1 -d" "`
[ ${LINES} -le 3 ] && {
  echo "No Data Found in ${TABLE}."
}
[ ${LINES} -gt 3 ] && {
  cat ${ORD}${TABLE}.tmp | sed -e 's/ * / /g' -e 's/^ //' -e 's/ ,/,/g' -e 's/, /,/g' | tail -n +3 | head -n 1 > ./${ORD}${TABLE}.headers
}

sqlplus -L logn/pswd@//${CONN}<<EOF >/dev/null
set pagesize 0;
set verify off;
set feedback off;
set long 99999;
set linesize 32767;
set trimspool on;
col object_ddl format A32000;
set colsep ,;
set underline off;
set headsep off;
spool ${ORD}${TABLE}.tmp;
select * from tblspc.${TABLE};
EOF
LINES=`wc -l ${ORD}${TABLE}.tmp | cut -f1 -d" "`
[ ${LINES} -le 3 ] && {
  echo "No Data Found in ${TABLE}."
}
[ ${LINES} -gt 3 ] && {
  cat ${ORD}${TABLE}.headers > ${FOLDER}/${ORD}${TABLE}.csv
  cat ${ORD}${TABLE}.tmp | sed -e 's/ * / /g' -e 's/^ //' -e 's/ ,/,/g' -e 's/, /,/g' | tail -n +2 | head -n -1 >> ${FOLDER}/${ORD}${TABLE}.csv
}

Answer 3

If you are using 12.2, you can simply say

set markup csv on
spool myfile.csv

Answer 4

Use vi or vim to write the sql, use colsep with a control-A (in vi and vim precede the ctrl-A with a ctrl-v). Be sure to set the linesize and pagesize to something rational and turn on trimspool and trimout.

spool it off to a file. Then...

sed -e 's/,/;/g' -e 's/ *{ctrl-a} */,/g'  {spooled file}  > output.csv

That sed thing can be turned into a script. The " *" before and after the ctrl-A squeezes out all the useless spaces. Isn't it great that they bothered to enable html output from sqlplus but NOT native csv?????

I do it this way because it handles commas in the data. I turns them to semi-colons.