Array/ In Ref cursor

Hi all,
I need to create a stored procedure which takes 30 arguments. However specifying them will become very messy. Is there a shorter way to pass in so many arguments? Perhaps an In Ref Cursor or an array?
Any help would be much appreciated.
Thanks
Rich

> Is there a shorter way to pass in so many arguments?
Yes. You can define a PL/SQL record structure (like a C struct). And if I'm not mistaken (last used Pro*C in '99), you can actually define a C struct wrapper for a PL/SQL struct in Pro*C.
Another option is defining an ADT (Advance Data Type). Superior to a PL/SQL struct as this is usable in both SQL and PL/SQL engines, is object orientated and therefore support subclassing, methods and so on.
Slightly more complex to use from Pro*C perhaps? I know the OCI supported an "object" interface for ADT's since Oracle 8i, but have never used it myself.
> Perhaps an In Ref Cursor or an array?
Not sure how a ref cursor can be used to pass parameter values.. A ref cursor is pointer to a SQL cursor in the Shared Pool - which can be passed to an external client (like Pro*C program) to use.
An array is also an option. Oracle uses it themselves for the two parameter calling method for mod_plsql (Oracle Web Cartridge). Your PL/SQL web enabled proc is called with a name list and a value list array.

Similar Messages

REF CURSOR to Associative Array

I have a procedure that builds up a collection (associative array / index-by table). Is there any way to hand the collection over to ODP.NET without putting the contents into a temporary table? REF CURSOR seems to fit, but I cannot find the syntax to OPEN a REF CURSOR for the collection.
I am trying to turn the GetResourceGroup() a_Resources parameter into a ref cursor in the example below (which, as I understand it, would also hold the collection in memory until the cursor is closed).
Alternatively, is there a direct method? I.e. is it possible to get the collection directly into a DataTable or similar?
Any advice gratefully received,
Peter
TYPE RsrcRow IS RECORD
ResourceId Resource.ResourceId%TYPE,
ResourceName Resource.ResourceName%TYPE
TYPE Resources IS TABLE OF RsrcRow INDEX BY VARCHAR2(80);
PROCEDURE GetResourceGroup(
a_ResourceType IN VARCHAR2,
a_Resources OUT Resources
IS
l_LocalCollection Resources;
BEGIN
GetResourcesByType(a_ResourceType, l_LocalCollection);
-- Return resource set (need to port this to ODP.NET).
a_Resources := l_LocalCollection;
RETURN;
END GetResourceGroup;

odp.net does not have collection support right now.
PL/SQL Associative Array should be available in next release (see other forum threads)
1. Support for Varchar2 indexes is unlikely (since OCI does not provide such functionality). Which means you can only return an C style "array" like structure at best.
2. Your 'GetResourceGroup' need to return REF CURSOR (or other primitive type like VARCHAR2), until Associative Array become available.
FB

ORA-01008 with ref cursor and dynamic sql

When I run the follwing procedure:
variable x refcursor
set autoprint on
begin
Crosstab.pivot(p_max_cols => 4,
   p_query => 'select job, count(*) cnt, deptno, row_number() over (partition by job order by deptno) rn from scott.emp group by job, deptno',
   p_anchor => Crosstab.array('JOB'),
   p_pivot => Crosstab.array('DEPTNO', 'CNT'),
   p_cursor => :x );
end;I get the following error:
^----------------
Statement Ignored
set autoprint on
begin
adsmgr.Crosstab.pivot(p_max_cols => 4,
p_query => 'select job, count(*) cnt, deptno, row_number() over (partition by
p_anchor => adsmgr.Crosstab.array('JOB'),
p_pivot => adsmgr.Crosstab.array('DEPTNO', 'CNT'),
p_cursor => :x );
end;
ORA-01008: not all variables bound
I am running this on a stored procedure as follows:
create or replace package Crosstab
as
    type refcursor is ref cursor;
    type array is table of varchar2(30);
    procedure pivot( p_max_cols       in number   default null,
                     p_max_cols_query in varchar2 default null,
                     p_query          in varchar2,
                     p_anchor         in array,
                     p_pivot          in array,
                     p_cursor in out refcursor );
end;
create or replace package body Crosstab
as
procedure pivot( p_max_cols          in number   default null,
                 p_max_cols_query in varchar2 default null,
                 p_query          in varchar2,
                 p_anchor         in array,
                 p_pivot          in array,
                 p_cursor in out refcursor )
as
    l_max_cols number;
    l_query    long;
    l_cnames   array;
begin
    -- figure out the number of columns we must support
    -- we either KNOW this or we have a query that can tell us
    if ( p_max_cols is not null )
    then
        l_max_cols := p_max_cols;
    elsif ( p_max_cols_query is not null )
    then
        execute immediate p_max_cols_query into l_max_cols;
    else
        RAISE_APPLICATION_ERROR(-20001, 'Cannot figure out max cols');
    end if;
    -- Now, construct the query that can answer the question for us...
    -- start with the C1, C2, ... CX columns:
    l_query := 'select ';
    for i in 1 .. p_anchor.count
    loop
        l_query := l_query || p_anchor(i) || ',';
    end loop;
    -- Now add in the C{x+1}... CN columns to be pivoted:
    -- the format is "max(decode(rn,1,C{X+1},null)) cx+1_1"
    for i in 1 .. l_max_cols
    loop
        for j in 1 .. p_pivot.count
        loop
            l_query := l_query ||
                'max(decode(rn,'||i||','||
                           p_pivot(j)||',null)) ' ||
                            p_pivot(j) || '_' || i || ',';
        end loop;
    end loop;
    -- Now just add in the original query
    l_query := rtrim(l_query,',')||' from ( '||p_query||') group by ';
    -- and then the group by columns...
    for i in 1 .. p_anchor.count
    loop
        l_query := l_query || p_anchor(i) || ',';
    end loop;
    l_query := rtrim(l_query,',');
    -- and return it
    execute immediate 'alter session set cursor_sharing=force';
    open p_cursor for l_query;
    execute immediate 'alter session set cursor_sharing=exact';
end;
end;
/I can see from the error message that it is ignoring the x declaration, I assume it is because it does not recognise the type refcursor from the procedure.
How do I get it to recognise this?
Thank you in advance

Thank you for your help
This is the version of Oracle I am running, so this may have something to do with that.
Oracle9i Enterprise Edition Release 9.2.0.8.0 - Production
With the Partitioning, OLAP and Oracle Data Mining options
JServer Release 9.2.0.8.0 - Production
I found this on Ask Tom (http://asktom.oracle.com/pls/asktom/f?p=100:11:0::::P11_QUESTION_ID:3027089372477)
Hello, Tom.
I have one bind variable in a dynamic SQL expression.
When I open cursor for this sql, it gets me to ora-01008.
Please consider:
Connected to:
Oracle8i Enterprise Edition Release 8.1.7.4.1 - Production
JServer Release 8.1.7.4.1 - Production
SQL> declare
2    type cur is ref cursor;
3    res cur;
4 begin
5    open res for
6    'select * from (select * from dual where :p = 1) connect by 1 = 1'
7    using 1;
8 end;
9 /
declare
ERROR at line 1:
ORA-01008: not all variables bound
ORA-06512: at line 5
SQL> declare
2    type cur is ref cursor;
3    res cur;
4 begin
5    open res for
6    'select * from (select * from dual where :p = 1) connect by 1 = 1'
7    using 1, 2;
8 end;
9 /
PL/SQL procedure successfully completed.
And if I run the same thing on 10g -- all goes conversely. The first part runs ok, and the second
part reports "ORA-01006: bind variable does not exist" (as it should be, I think). Remember, there
is ONE bind variable in sql, not two. Is it a bug in 8i?
What should we do to avoid this error running the same plsql program code on different Oracle
versions?
P.S. Thank you for your invaluable work on this site.
Followup   June 9, 2005 - 6pm US/Eastern:
what is the purpose of this query really?
but it would appear to be a bug in 8i (since it should need but one). You will have to work that
via support. I changed the type to tarray to see if the reserved word was causing a problem.
variable v_refcursor refcursor;
set autoprint on;
begin
     crosstab.pivot (p_max_cols => 4,
             p_query =>
               'SELECT job, COUNT (*) cnt, deptno, ' ||
               '       ROW_NUMBER () OVER ( ' ||
               '          PARTITION BY job ' ||
               '          ORDER BY deptno) rn ' ||
               'FROM   emp ' ||
               'GROUP BY job, deptno',
               p_anchor => crosstab.tarray ('JOB'),
               p_pivot => crosstab.tarray ('DEPTNO', 'CNT'),
               p_cursor => :v_refcursor);
end;
/Was going to use this package as a stored procedure in forms but I not sure it's going to work now.

[Solved] 27.8.4 How to Create a VO on a REF CURSOR - Missing first row

Searching the forum I found: BC4J - Get one less row from view object.
Dive into BC4J related --REF CURSOR (Resultset)
The first message did not have any answers, and the second had a follow up question - still no answers though - and I thought I would try a different title.
(This is off topic, but it would be a great help if the search results also displayed the number of replys in the thread. That way, I wouldn't have to view the messages that don't have responses.)
(This will be deployed on a server that has the ADF for JDeveloper 10.1.2 installed, so using that version of JDeveloper to develop the app.)
Okay, back to the problem ==>
I created a VO from a ref cursor, using the manual as a guide. When I run a page that displays a read only table of the view object, I am missing the first row. (Always the first row!) I don't have any order set, and if I call the ref cursor in a Java program for testing, I see all rows and the count is correct.
One other point, when I call the page, I get the following validation error:
Validation Error
You must correct the following error(s) before proceeding:
* JBO-29000: Unexpected exception caught: java.lang.ClassCastException, msg=null
* null
I still see the table, it is just missing the first row of data.
In my form I have first, previous set, next set , and last
navigation buttons. If I press last then first, the error goes away. I still don't see the missing row though.
Any guidance would be appreciated! I can post my code, but it is pretty much the same code in the AdvancedViewObjectExamples.zip in the ViewObjectOnRefCursor example. I just substituted my two package calls (getRefCursor and getRefCursorCount).
Thanks, Ken

Went back to a backup copy of the source. Fixed the error. Now I'm back to just not being able to see the first row of data.
Additional Note: I had removed fields in the display. Once I truncated the ps_txn table in the schema defined by the model, the data would display (Still without the first record.)
Are there any examples that are more in depth than the few pages in the developer guide?
Here is the code for my VOImpl class:
package newslinearchive.model.datamodel;
import java.math.BigDecimal;
import java.sql.CallableStatement;
import java.sql.PreparedStatement;
import java.sql.ResultSet;
import java.sql.SQLException;
import java.sql.Timestamp;
import java.sql.Types;
import oracle.jbo.InvalidParamException;
import oracle.jbo.JboException;
import oracle.jbo.domain.Date;
import oracle.jbo.domain.Number;
import oracle.jbo.server.DBTransaction;
import oracle.jbo.server.QueryCollection;
import oracle.jbo.server.SQLBuilder;
import oracle.jbo.server.ViewObjectImpl;
import oracle.jbo.server.ViewRowImpl;
import oracle.jbo.server.ViewRowSetImpl;
import oracle.jdbc.driver.OracleCallableStatement;
import oracle.jdbc.driver.OracleTypes;
// --- File generated by Oracle ADF Business Components Design Time.
// --- Custom code may be added to this class.
// --- Warning: Do not modify method signatures of generated methods.
public class SearchRefCursorImpl extends ViewObjectImpl {
* This is the default constructor (do not remove)
public SearchRefCursorImpl() {
* Overridden framework method.
* Executed when the framework needs to issue the database query for
* the query collection based on this view object. One view object
* can produce many related result sets, each potentially the result
* of different bind variable values. If the rowset in query is involved
* in a framework-coordinated master/detail viewlink, then the params array
* will contain one or more framework-supplied bind parameters. If there
* are any user-supplied bind parameter values, they will PRECEED the
* framework-supplied bind variable values in the params array, and the
* number of user parameters will be indicated by the value of the
* numUserParams argument.
protected void executeQueryForCollection(Object qc,Object[] params,int numUserParams) {
storeNewResultSet(qc,retrieveRefCursor(qc,params));
super.executeQueryForCollection(qc, params, numUserParams);
* Overridden framework method.
* Wipe out all traces of a built-in query for this VO
protected void create() {
getViewDef().setQuery(null);
getViewDef().setSelectClause(null);
setQuery(null);
* Overridden framework method.
* The role of this method is to "fetch", populate, and return a single row
* from the datasource by calling createNewRowForCollection() and populating
* its attributes using populateAttributeForRow().
protected ViewRowImpl createRowFromResultSet(Object qc, ResultSet rs) {
* We ignore the JDBC ResultSet passed by the framework (null anyway) and
* use the resultset that we've stored in the query-collection-private
* user data storage
rs = getResultSet(qc);
* Create a new row to populate
ViewRowImpl r = createNewRowForCollection(qc);
try {
* Populate new row by attribute slot number for current row in Result Set
// populateAttributeForRow(r,0, rs.getLong(1));
// populateAttributeForRow(r,1, rs.getString(2));
// populateAttributeForRow(r,2, rs.getString(3));
// MASTERID NOT NULL NUMBER
populateAttributeForRow(r,0, rs.getBigDecimal(1));
//ID NOT NULL NUMBER
populateAttributeForRow(r,1, rs.getBigDecimal(2));
// CAID NOT NULL NUMBER
populateAttributeForRow(r,2, rs.getBigDecimal(3));
// LANGUAGE NOT NULL VARCHAR2(30)
populateAttributeForRow(r,3, rs.getString(4));
// IS_CURRENT_VERSION NOT NULL NUMBER(1)
populateAttributeForRow(r,4, rs.getBigDecimal(5));
// FOLDER_ID NOT NULL NUMBER
populateAttributeForRow(r,5, rs.getBigDecimal(6));
// FOLDER_REGION_ID NOT NULL NUMBER
populateAttributeForRow(r,6, rs.getBigDecimal(7));
// NAME NOT NULL VARCHAR2(256)
populateAttributeForRow(r,7, rs.getString(8));
// DISPLAY_NAME VARCHAR2(256)
populateAttributeForRow(r,8, rs.getString(9));
// ITEMTYPE NOT NULL VARCHAR2(30)
populateAttributeForRow(r,9, rs.getString(10));
// SUBTYPE VARCHAR2(40)
populateAttributeForRow(r,10, rs.getString(11));
// SUBTYPE_CAID NUMBER
populateAttributeForRow(r,11, rs.getBigDecimal(12));
// PARENT_ITEM_ID NUMBER
populateAttributeForRow(r,12, rs.getBigDecimal(13));
// CATEGORY_ID NUMBER
populateAttributeForRow(r,13, rs.getBigDecimal(14));
// CATEGORY_CAID NUMBER
populateAttributeForRow(r,14, rs.getBigDecimal(15));
// AUTHOR VARCHAR2(50)
populateAttributeForRow(r,15, rs.getString(16));
// DESCRIPTION VARCHAR2(2000)
populateAttributeForRow(r,16, rs.getString(17));
// PUBLISH_DATE NOT NULL DATE
populateAttributeForRow(r,17, rs.getDate(18));
// EXPIREMODE VARCHAR2(90)
populateAttributeForRow(r,18, rs.getString(19));
// EXPIRENUMBER NUMBER
populateAttributeForRow(r,19, rs.getBigDecimal(20));
// EXPIREDATE DATE
populateAttributeForRow(r,20, rs.getDate(21));
// IMAGE VARCHAR2(350)
populateAttributeForRow(r,21, rs.getString(22));
// KEYWORDS VARCHAR2(2000)
populateAttributeForRow(r,22, rs.getString(23));
// URL VARCHAR2(4000)
populateAttributeForRow(r,23, rs.getString(24));
// FILENAME VARCHAR2(350)
populateAttributeForRow(r,24, rs.getString(25));
// TEXT CLOB()
populateAttributeForRow(r,25, rs.getClob(26));
// FOLDER_LINK_ID NUMBER
populateAttributeForRow(r,26, rs.getBigDecimal(27));
// FOLDER_LINK_CAID NUMBER
populateAttributeForRow(r,27, rs.getBigDecimal(28));
// ACTIVE NOT NULL NUMBER(1)
populateAttributeForRow(r,28, rs.getBigDecimal(29));
// CAN_BE_CHECKEDOUT NUMBER(1)
populateAttributeForRow(r,29, rs.getBigDecimal(30));
// IS_ITEM_CHECKEDOUT NUMBER(1)
populateAttributeForRow(r,30, rs.getBigDecimal(31));
// CHECKER_USERNAME VARCHAR2(256)
populateAttributeForRow(r,31, rs.getString(32));
// CHECKOUT_DATE DATE
populateAttributeForRow(r,32, rs.getDate(33));
// FULLSCREEN NOT NULL NUMBER(1)
populateAttributeForRow(r,33, rs.getBigDecimal(34));
// INPLACE NOT NULL NUMBER(1)
populateAttributeForRow(r,34, rs.getBigDecimal(35));
// CREATEDATE NOT NULL DATE
populateAttributeForRow(r,35, rs.getDate(36));
// CREATOR NOT NULL VARCHAR2(256)
populateAttributeForRow(r,36, rs.getString(37));
// UPDATEDATE DATE
populateAttributeForRow(r,37, rs.getDate(38));
// UPDATOR VARCHAR2(256)
populateAttributeForRow(r,38, rs.getString(39));
// SECURITY VARCHAR2(25)
populateAttributeForRow(r,39, rs.getString(40));
// VISIBLE NOT NULL NUMBER(1)
populateAttributeForRow(r,40, rs.getBigDecimal(41));
// SEQUENCE NOT NULL NUMBER
populateAttributeForRow(r,41, rs.getBigDecimal(42));
// CATEGORY_SEQUENCE NOT NULL NUMBER
populateAttributeForRow(r,42, rs.getBigDecimal(43));
// AUTHOR_SEQUENCE NOT NULL NUMBER
populateAttributeForRow(r,43, rs.getBigDecimal(44));
// CREATE_DATE_SEQUENCE NOT NULL NUMBER
populateAttributeForRow(r,44, rs.getBigDecimal(45));
// ITEMTYPE_SEQUENCE NOT NULL NUMBER
populateAttributeForRow(r,45, rs.getBigDecimal(46));
catch (SQLException s) {
throw new JboException(s);
return r;
* Overridden framework method.
* Return true if the datasource has at least one more record to fetch.
protected boolean hasNextForCollection(Object qc) {
ResultSet rs = getResultSet(qc);
boolean nextOne = false;
try {
nextOne = rs.next();
* When were at the end of the result set, mark the query collection
* as "FetchComplete".
if (!nextOne) {
setFetchCompleteForCollection(qc, true);
* Close the result set, we're done with it
rs.close();
catch (SQLException s) {
throw new JboException(s);
return nextOne;
* Overridden framework method.
* The framework gives us a chance to clean up any resources related
* to the datasource when a query collection is done being used.
protected void releaseUserDataForCollection(Object qc, Object rs) {
* Ignore the ResultSet passed in since we've created our own.
* Fetch the ResultSet from the User-Data context instead
ResultSet userDataRS = getResultSet(qc);
if (userDataRS != null) {
try {
userDataRS.close();
catch (SQLException s) {
/* Ignore */
super.releaseUserDataForCollection(qc, rs);
* Overridden framework method
* Return the number of rows that would be returned by executing
* the query implied by the datasource. This gives the developer a
* chance to perform a fast count of the rows that would be retrieved
* if all rows were fetched from the database. In the default implementation
* the framework will perform a SELECT COUNT(*) FROM (...) wrapper query
* to let the database return the count. This count might only be an estimate
* depending on how resource-intensive it would be to actually count the rows.
public long getQueryHitCount(ViewRowSetImpl viewRowSet) {
Long result = (Long)callStoredFunction(NUMBER,
"PORTAL.SEARCH_REFCURSOR.getRefCursorCount",
viewRowSet.getParameters(true));
return result.longValue();
// ------------- PRIVATE METHODS ----------------
* Return a JDBC ResultSet representing the REF CURSOR return
* value from our stored package function.
* new Object[]{getNamedBindParamValue("Email",params)}
private ResultSet retrieveRefCursor(Object qc, Object[] params) {
ResultSet rs = (ResultSet)callStoredFunction(OracleTypes.CURSOR,
"PORTAL.SEARCH_REFCURSOR.getRefCursor",
null);
return rs ;
private Object getNamedBindParamValue(String varName, Object[] params) {
Object result = null;
if (getBindingStyle() == SQLBuilder.BINDING_STYLE_ORACLE_NAME) {
if (params != null) {
for (Object param : params) {
Object[] nameValue = (Object[])param;
String name = (String)nameValue[0];
if (name.equals(varName)) {
return (String)nameValue[1];
throw new JboException("No bind variable named '"+varName+"'");
* Store a new result set in the query-collection-private user-data context
private void storeNewResultSet(Object qc, ResultSet rs) {
ResultSet existingRs = getResultSet(qc);
// If this query collection is getting reused, close out any previous rowset
if (existingRs != null) {
try {existingRs.close();} catch (SQLException s) {}
setUserDataForCollection(qc,rs);
hasNextForCollection(qc); // Prime the pump with the first row.
* Retrieve the result set wrapper from the query-collection user-data
private ResultSet getResultSet(Object qc) {
return (ResultSet)getUserDataForCollection(qc);
* Return either null or a new oracle.jbo.domain.Date
private static Date nullOrNewDate(Timestamp t) {
return t != null ? new Date(t) : null;
* Return either null or a new oracle.jbo.domain.Number
private static Number nullOrNewNumber(BigDecimal b) {
try {
return b != null ? new Number(b) : null;
catch (SQLException s) { }
return null;
//----------------[ Begin Helper Code ]------------------------------
public static int NUMBER = Types.NUMERIC;
public static int DATE = Types.DATE;
public static int VARCHAR2 = Types.VARCHAR;
public static int CLOB = Types.CLOB;
* Simplifies calling a stored function with bind variables
* You can use the NUMBER, DATE, and VARCHAR2 constants in this
* class to indicate the function return type for these three common types,
* otherwise use one of the JDBC types in the java.sql.Types class.
* NOTE: If you want to invoke a stored procedure without any bind variables
* ==== then you can just use the basic getDBTransaction().executeCommand()
* @param sqlReturnType JDBC datatype constant of function return value
* @param stmt stored function statement
* @param bindVars Object array of parameters
* @return function return value as an Object
protected Object callStoredFunction(int sqlReturnType, String stmt,
Object[] bindVars) {
CallableStatement st = null;
try {
st = getDBTransaction().createCallableStatement("begin ? := " + stmt +
"; end;", 0);
st.registerOutParameter(1, sqlReturnType);
if (bindVars != null) {
for (int z = 0; z < bindVars.length; z++) {
st.setObject(z + 2, bindVars[z]);
st.executeUpdate();
return st.getObject(1);
catch (SQLException e) {
throw new JboException(e);
finally {
if (st != null) {
try {
st.close();
catch (SQLException e) {}
/**Gets the bind variable value for Email
public String getEmail() {
return (String)getNamedWhereClauseParam("Email");
/**Sets <code>value</code> for bind variable Email
public void setEmail(String value) {
setNamedWhereClauseParam("Email", value);
/**getEstimatedRowCount - overridden for custom java data source support.
public long getEstimatedRowCount() {
long value = super.getEstimatedRowCount();
return value;
Thanks, Ken

Error while working with ref cursor

Hi I tried the following but getting the err
DECLARE
TYPE ref_nm IS REF CURSOR;
vref REF_NM;
vemp emp%rowtype;
BEGIN
OPEN vref FOR SELECT ename ,sal FROM EMP;
LOOP
FETCH vref INTO vemp;
EXIT WHEN vref%NOTFOUND;
DBMS_OUTPUT.PUT_LINE ( vemp.ename ||','||vemp.sal );
END LOOP;
CLOSE vref;
END;Error is
ORA-06504: PL/SQL: Return types of Result Set variables or query do not match

Syntactically correct - but a horrible approach performance wise if you only need specific columns.
The basic problem here is using the wrong tool. A ref cursor. Why? There's no reason in this code for using a ref cursor. Using a standard cursor data type addresses the requirement a lot better.
As you can define the cursor:
cursor myCursor is select c1, c2 from tab1;
You can define an array data type for fetching that cursors data - and thus definition does not need to be touched when you change the cursor itself to include/exclude columns:
type TMyCursorBuffer is table of myCursor%RowType;
And finally you can define a bulk collection buffer for bulk processing the output of the cursor:
myBuffer TMyCursorBuffer;

Bind Variables in ref cursor

Version details
BANNER
Oracle Database 10g Enterprise Edition Release 10.2.0.4.0 - 64bi
PL/SQL Release 10.2.0.4.0 - Production
CORE    10.2.0.4.0      Production
TNS for Solaris: Version 10.2.0.4.0 - Production
NLSRTL Version 10.2.0.4.0 - ProductionBelow is the procedure where I'm using ref cursor
CREATE OR REPLACE PROCEDURE ref_sample
p_account_nbr     in     varchar2,
p_ref_out          out     sys_refcursor
IS
BEGIN
     OPEN p_ref_out FOR
     SELECT     account_nbr,status,(
                                        CASE
                                             WHEN status = 'Pending' THEN
                                                  req_mail_date
                                             WHEN status IN ('Rejected','Approved') THEN
                                                  NVL(verified_mail_date,req_mail_date)
                                        END
                                        )req_mail_date ,
                                             CASE
                                                  WHEN status = 'Pending' THEN
                                                       NULL
                                                  WHEN status IN ('Rejected','Approved') THEN
                                                       NVL(verified_user_id,req_user_id)
                                             END
                                        )req_user_id
     FROM     X_tbl
     WHERE     account_nbr IN p_account_nbr
               AND TRUNC(upload_date) = TRUNC(SYSDATE)
     ORDER BY upload_date DESC ;
END;
/My input parameter p_account_nbr looks like ('a1','a2','a3')
Now,after knowing the importance of bind variables I'd like to make use of them in the above ref cursor.
But,here my input parameter is a string of varying length..either I've to go for the approach suggested here
http://asktom.oracle.com/pls/asktom/f?p=100:11:3667281145899708::::P11_QUESTION_ID:110612348061
or
http://www.dba-oracle.com/t_cursor_sharing.htm
I'm not much clear with the first approach,so I'm thinking of to modify my procedure as below
CREATE OR REPLACE PROCEDURE ref_sample
p_account_nbr     in     varchar2,
p_ref_out          out     sys_refcursor
IS
BEGIN
     alter session set cursor_sharing=force;
     OPEN p_ref_out FOR
     SELECT     account_nbr,status,(
                                        CASE
                                             WHEN status = 'Pending' THEN
                                                  req_mail_date
                                             WHEN status IN ('Rejected','Approved') THEN
                                                  NVL(verified_mail_date,req_mail_date)
                                        END
                                        )req_mail_date ,
                                             CASE
                                                  WHEN status = 'Pending' THEN
                                                       NULL
                                                  WHEN status IN ('Rejected','Approved') THEN
                                                       NVL(verified_user_id,req_user_id)
                                             END
                                        )req_user_id
     FROM     X_tbl
     WHERE     account_nbr IN p_account_nbr
               AND TRUNC(upload_date) = TRUNC(SYSDATE)
     ORDER BY upload_date DESC ;
     alter session set cursor_sharing=exact;
END;
/Please let me know if the above modified code is fine or should I use bind variables??Also let me know better approach of both.

The correct way to do this is use an array type for the input values as in this example.
SQL> create or replace procedure p
2      (
3      p_values sys.odcivarchar2list,
4      c out sys_refcursor
5      ) as
6 begin
7      open c for
8         select object_name, owner, object_type
9         from all_objects
10         where object_name in (select column_value from table(p_values));
11 end;
12 /
Procedure created.
SQL> var c refcursor
SQL> exec p (sys.odcivarchar2list('DUAL','USER_VIEWS'), :c)
PL/SQL procedure successfully completed.
SQL> print c
OBJECT_NAME                    OWNER                          OBJECT_TYPE
DUAL                           SYS                            TABLE
DUAL                           PUBLIC                         SYNONYM
USER_VIEWS                     SYS                            VIEW
USER_VIEWS                     PUBLIC                         SYNONYM
SQL> exec p (sys.odcivarchar2list('DUAL','USER_VIEWS','ALL_OBJECTS','ALL_SOURCE'), :c)
PL/SQL procedure successfully completed.
SQL> print c
OBJECT_NAME                    OWNER                          OBJECT_TYPE
DUAL                           SYS                            TABLE
DUAL                           PUBLIC                         SYNONYM
ALL_OBJECTS                    SYS                            VIEW
ALL_OBJECTS                    PUBLIC                         SYNONYM
USER_VIEWS                     SYS                            VIEW
USER_VIEWS                     PUBLIC                         SYNONYM
ALL_SOURCE                     SYS                            VIEW
ALL_SOURCE                     PUBLIC                         SYNONYM
8 rows selected.
SQL>That and other methods are described here.
http://tkyte.blogspot.com/2006/06/varying-in-lists.html
You would not use dynamic SQL.

REF Cursor as Job Parameter

Hi,
is it possible to use a REF cursor as a parameter for a job? I need jobs created with DBMS_SCHEDULER.create_job to process data which has always the same format but comes from different sources. The job must be generic and thus should not know where the data comes from. The job calls a PL/SQL procdure which takes the REF cursor as a parameter.
Regards,
Pat
Edited by: Gwydion on Jun 16, 2010 7:13 AM

Gwydion wrote:
I tried a different way: Instead of a ref cursor, I tried to pass a PL/SQL table containing the data that the job should process. Bad idea in general as this is not very scalable. A PL/SQL "+table+" is nothing table-like at all. The correct term is array or collection. This structure is primitive in that it cannot be indexed or structured like a SQL table. It also resides in "expensive" memory (the PGA/private memory of a server process) - unlike a SQL table data that will reside in the SGA buffer cache when used. The db buffer cache is shared memory - usable by all processes and thus "inexpensive" memory in comparison.
So passing and processing any largish data structure that way (via PGA memory) is inherently an unscalable approach. Also, this data set needs to persist until the job process kicks off - thus it needs to be stored somewhere. Why store this data again when it is based on table and row table that already exist in the database? Why duplicate the data?
You also need to consider data concurrency and integrity. By the time that job/scheduler background process kicks off and receives this data set structure to use for processing... the row data that this set was based on may have changed.. or may not even exist anymore.
For example, let's say that the data set contains a list of invoice numbers that need to be processed. Some of those invoices may have been canceled in the meantime. Or deleted from the system. Whatever invoice numbers this background server process now needs to deal with, can now be invalid.
You therefore need to carefully consider exactly what issues and requirements you're trying to address with this approach you are attempting. There are numerous problems areas that it needs to deal with.
Which begs the question - why are you considering this approach? What problem are you trying to address? Why not simply pass the parameters to the job it needs to use to call the procedure that creates the ref cursor? This is a much simpler and elegant solution than to pass the actual data returned by the ref cursor to the job.

Ref Cursor as input parameter

Is it possible to pass a ref cursor from java as an input parameter to a stored procedure, which the procedure can then process?
(Passing Ref Cursors as output parameters from a stored procedure to a result set works fine.)

You may want to try passing Arrays to the Stored Procedure. Use Oracle.SQL.ARRAY and then in the stored procedure get the array.

Passing Ref Cursor as IN parameter

I have a situation where I have to send whole table into the stored procedure, I read from the oracle new feature for .Net documentation, It is possible to pass an ref cursor as IN parameter to an stored procedure. Please help me in this, it is urgent!
I am using Visual Studio.Net 2003, ODP.Net.

Array binding is the best method for large inserts from .NET.
The problem with the associative array is:
-you must write a PL/SQL procedure to actually do the inserts.
-while this will get the data to the PL/SQL engine on the server in a block, it still must cross the into the SQL engine one row at at time.
-Array binding is simpler (you use a normal insert).
-There's a built-in knob to control the batch size.
-The array of row data is copied all the way into the SQL engine together, so it's faster.
Remember, array binding is how the SQL*Loader's conventional-path inserts work.
According to Tom Kyte "you can achive 100's of rows per second into heavily indexed tables that are concurrently being read and updated by other processes."
David

DESCRIBING A REF CURSOR

For debugging purposes, I would like to be able to describe a REF CURSOR returned by a function in a package, using pl/sql.
Can you help me ?

ZlatKo
A brave attempt and well done for putting up with the forum software.
Alas, I can't catually run the code - I get the following error-stack. I'll have another go at it this lunchtime.
If I'm not mistaken your procedure would give us a visual output of the REF CURSOR structure, just like DESC. I think the original poster was after something they could use programmatically inside a PL/SQL procedure to interrogate and process a REF CURSOR with an unknown signature. It would be possible to walk the structure as you do, suck out the values and store them in arrays. Fine. But without knowing the column names how do we know what to do with these values?
At some point the receiving program has to know what the purpose of the REF CURSOR. We can imagine a function that generates a REF CURSOR from one of three queries depending upon the value of an input parameter. Every program which calls that function must pass in that parameter and therefore must know the signature of the REF CURSOR that it gets back.
The idea of a function returning data at its own whim is a bit laughable. Even web services, surely the ultimate expression of the encapsulation paradigm, insist on the publication of a signature - what we have to put in, what we will get back.
Cheers, APC

Ref Cursor is invalid error

Hi,
I have written a jdbc application which return a ref cursor. I use ResultSet rset = ((OracleCallableStatement)call).getCursor (1); to get the result set according to jdbc user guide. Compilation is ok but when I run it, the error is:
Exception in thread "main" java.sql.SQLException: Ref cursor is invalid
at java.lang.Throwable.fillInStackTrace(Native Method)
at java.lang.Throwable.fillInStackTrace(Compiled Code)
at java.lang.Throwable.<init>(Compiled Code)
at java.lang.Exception.<init>(Exception.java:42)
at java.sql.SQLException.<init>(SQLException.java:43)
at oracle.jdbc.dbaccess.DBError.throwSqlException(DBError.java:114)
at oracle.jdbc.dbaccess.DBError.throwSqlException(DBError.java:156)
at oracle.jdbc.dbaccess.DBError.throwSqlException(DBError.java:219)
at oracle.jdbc.driver.OracleStatement.getCursorValue(OracleStatement.java:2544)
at oracle.jdbc.driver.OracleCallableStatement.getCursor(OracleCallableStatement.java:583)
at useCursor.main(useCursor.java:41)
I attach the code for ur reference. Pls point out to me what's wrong.
import oracle.sql.*;
import oracle.jdbc.oracore.Util;
import java.sql.*;
import java.io.*;
import oracle.jdbc.driver.*;
class useCursor{
private static final String Server = "job.comp.edu.au:1521:testing";
private static final String Username = "user";
private static final String Password = "org";
public static void main (String args [])
throws SQLException
// Connect to the database
// You can put a database name after the @ sign in the connection URL.
Connection conn = getConnection ();
String elements[] = new String[1];
elements[0] = "select * from product";
// Create the stored procedure
init (conn);
ArrayDescriptor desc = ArrayDescriptor.createDescriptor("STR_VARRAY3", conn);
ARRAY newArray = new ARRAY(desc, conn, elements);
// Prepare a PL/SQL call
CallableStatement call =
conn.prepareCall ("{ ? = call java_refcursor.job_listing (?)}");
// Find out all the SALES person
call.registerOutParameter (1, OracleTypes.CURSOR);
((OraclePreparedStatement)call).setARRAY(2, newArray);
call.execute ();
ResultSet rset = ((OracleCallableStatement)call).getCursor (1);
// Dump the cursor
while (rset.next ())
System.out.println (rset.getString ("AUTHOR"));
// Utility function to create the stored procedure
static void init (Connection conn)
throws SQLException
Statement stmt = conn.createStatement ();
try
stmt.execute ("DROP TYPE str_varray3");
catch (SQLException e)
// the above drop statements will throw exceptions
// if the types and tables did not exist before. Just ingore it.
stmt.execute ("CREATE TYPE str_varray3 AS VARRAY(10) OF VARCHAR2(50)");
stmt.execute ("create or replace package java_refcursor as " +
" type myrctype is ref cursor; "+
" function job_listing (sa str_varray3) return myrctype; " +
"end java_refcursor;");
stmt.execute ("create or replace package body java_refcursor as " +
" function job_listing (sa str_varray3) return myrctype is " +
" rc myrctype; " +
" i number; " +
" begin " +
" while sa.COUNT >= i" +
" loop " +
" open rc for sa(1);" +
" exit when rc%NOTFOUND; " +
" i := i + 1; " +
" end loop; " +
" return rc; " +
" end; " +
"end java_refcursor;");
public static Connection getConnection(){
Connection con = null;
try {
DriverManager.registerDriver(new oracle.jdbc.driver.OracleDriver());
catch (SQLException sqlex ) {
System.err.println("Unable register Oracle driver");
sqlex.printStackTrace();
try {
con = DriverManager.getConnection("jdbc:oracle:thin:@" + Server, Username, Password);
catch (SQLException sqlex) {
System.err.println("Unable to connect");
sqlex.printStackTrace();
return con;
Please advise me. Thank you very much.
null

i have found out my mistake
it lacks of i := 0;
my mind was not so ok then.. :)

Dynamic Ref Cursor with Dynamic Fetch

Hi,
I'm using dynamic sql (DBMS_SQL) to define columns of ref cursor.
It works Ok but the problem is when i'm using PL/SQL CURSOR in the REF CURSOR. Then,
I'm getting :
Error at line 3
ORA-00932: inconsistent datatypes: expected NUMBER got CURSER
ORA-06512: at "SYS.DBMS_SQL", line 1830
ORA-06512: at "TW.PRINT_REF_CURSOR", line 28
ORA-06512: at line 9
Here is my code:
set serveroutput on
exec DBMS_OUTPUT.ENABLE(1000000);
declare
l_cursor sys_refcursor;
begin
OPEN l_cursor FOR
SELECT SERVICE_TABLE.SERVICE, SERVICE_TABLE.SERVICE_GROUP, SERVICE_TABLE.SERVICE_DESC,
CURSOR(SELECT SERVICE_TABLE.SERVICE_CD FROM SERVICE_TABLE) SERVICE_CD_CURSOR
FROM SERVICE_TABLE ;
print_ref_cursor( l_cursor );
end;
=========================
CREATE OR REPLACE procedure print_ref_cursor
( p_query in out sys_refcursor,
p_date_fmt in varchar2 default 'dd-mon-yyyy hh24:mi:ss' )
is
l_theCursor integer;
l_columnValue varchar2(4000);
l_descTbl dbms_sql.desc_tab2;
l_colCnt number;
l_date date;
l_cursor SYS_REFCURSOR;
begin
l_theCursor := dbms_sql.to_cursor_number( p_query );
dbms_sql.describe_columns2
( l_theCursor, l_colCnt, l_descTbl );
-- define all columns to be cast to varchar2's, we
-- are just printing them out
for i in 1 .. l_colCnt loop
if ( l_descTbl(i).col_type in ( 12, 178, 179, 180, 181, 231 ) )
then
dbms_sql.define_column
(l_theCursor, i, l_date );
else
dbms_sql.define_column
(l_theCursor, i, l_columnValue, 4000);
end if;
end loop;
while ( dbms_sql.fetch_rows(l_theCursor) > 0 )
loop
for i in 1 .. l_colCnt loop
if ( l_descTbl(i).col_type in ( 12, 178, 179, 180, 181, 231 ) )
then
dbms_sql.column_value( l_theCursor, i, l_date );
l_columnValue := to_char( l_date, p_date_fmt );
else
dbms_sql.column_value( l_theCursor, i, l_columnValue );
end if;
dbms_output.put_line
( rpad( l_descTbl(i).col_schema_name || '.' ||
l_descTbl(i).col_name, 30 ) || ': ' || l_columnValue );
end loop;
dbms_output.put_line( '-----------------' );
end loop;
dbms_sql.close_cursor( l_theCursor );
end;
Is there a solution or bypass?
Regards,
Tamir Geva

No. The problem is that one cannot use DBMS_SQL.define_column() to define that column in the SQL projection as a cursor, and then use DBMS_SQL.column_value() to read it into a ref cursor variable.
You can however detect the cursor column - the DBMS_SQL.describe_columns3() call will return a col_type value of 102. In which case you can treat it as an exception (i.e. not process that column in the projection).
As a general issue - a cursor as a SQL column projection does not make sense to me. I have never used this in any production code. Nor do I see any reasons why.
If you want that column in the projection to contain a "list" of sorts (the results of the cursor), then a nested table type can be used as projected type and the MultiSet() function used to execute the in-line SQL and provide that SQL cursor's result as an array/nested table.
But even this approach raises the question why a standard relational join is not used?

Passing Ref Cursor to Oracle Stored Procedure Via C#

Hi all,
I am new to oracle and stuck with an issue. I have three insert stored procedures for three different tables. Two of them have multiple rows to be inserted, which is currently done via iterating through each row and insert to db in C# code. My requirement is to merge these three procedures in one and instead of iterating from C# code send table rows as (ref cursor or collection) to procedure and the procedure will handle the rest.
I read that ref cursor only works if you're data is in database as it reference the memory but in my case data is build on client side.
I am using Oracle 11i and ASP.Net 2.0
Can any help me on this please?
Edited by: 929463 on Apr 23, 2012 12:38 AM

929463 wrote:
I am new to oracle and stuck with an issue. I have three insert stored procedures for three different tables. Two of them have multiple rows to be inserted, which is currently done via iterating through each row and insert to db in C# code. My requirement is to merge these three procedures in one and instead of iterating from C# code send table rows as (ref cursor or collection) to procedure and the procedure will handle the rest.Why a single procedure? How is the procedure to determine the target table to insert the data into? And please - no dynamic SQL as that is 99% of the time wrong.
A ref cursor is something that PL/SQL creates - with the purpose of passing the cursor handle to your code. This enables the actual SQL statement for that cursor to be moved from client code, into a PL/SQL stored proc. It abstracts the client from having to understand SQL, understand the data model and so on. All clients use the same PL/SQL proc and thus the same code for creating that cursor. Thus no issue of some clients getting it half right or half wrong and dealing with data inconsistencies between clients.
The PL/SQL proc can be tuned and optimised, modified for catering for data model changes and so on. Without your client code having to be even recompiled as it is isolated against these server changes.
For all other interaction (running PL/SQL code, doing insert/update/delete/etc SQL statements), you need to create the cursor yourself in your code.
Also, the SQL engine only sees cursors. There are no differences between cursors. The client (e.g. PL/SQL) can call it a reference cursor, or an implicit cursor, or a DBMS_SQL cursor.. the SQL engine does not know that and does not care.
A ref cursor is simply a special type of client interface to a SQL cursor, allowing PL/SQL to create that SQL cursor and then pass the handle of that SQL cursor to other code to consume that cursor.
Okay, so if you want to insert data, you need in your code to create a cursor. This can be a SQL INSERT cursor - the actual insert statement. Or it can be a PL/SQL call - an anonymous PL/SQL code block that calls a stored proc that performs the insert (after applying validation and business logic).
The cursor will have one or more bind variables. Your client will pass values for these variables and the server-side code (SQL or PL/SQL) will be executed using this as variable data.
You can for example create a cursor as follows:
begin
DoFunkyInsert( :1, :2, :3 );
end;
{code}
3 bind variables are expected. You can now in the client build an array for each of these variables, containing a 100 values each (total of a 100 rows to insert). Do a single execute of the cursor, and tell Oracle that the bind is actually a 100 element array.
The complete array ships to Oracle - Oracle opens a loop and execute the cursor for each element in the array.
This is called bulk binding.
An alternative approach is to define the bind variable as a collection (a non-scalar value). And then code the PL/SQL procedure to open a loop and iterate through the collection/array, inserting a row per iteration.
The binding itself is more complex as your code know needs to understand Oracle object types and be able to define an array/collection that is a valid Oracle non-scalar data type.
The +Oracle Call Interface+ (OCI) is quite flexible in this regard. However, as you work via an abstraction layer (e.g. ADO, OleDB, ODBC, etc) your code is subject to whatever functionality this abstraction layer makes available to your code. And this is seldom includes all the power, functionality and flexibility of the (more complex) OCI itself.

STORED PROCEDURE/REF CURSOR: How to output entire buffer

I wrote a Stored Procedure wherein I use a Cursor to extract multiple
rows and columns. I then write them into the buffer
(dmbs_output.put_line). But when I try to capture the entire result
into an OUT variable, I only get the last buffered line.
So how do I output the entire buffer- all rows and columns? In other words (maybe), how do I use dbms_output.get_lines() to assign value to an OUT variable?
Alternatively, using REF CURSOR as OUT variable, I added the following to "CREATE OR REPLACE PROCEDURE ... ()":
cursor_out_test OUT cursor_test
But when I tried:
DEFINE CURSOR TYPE cursor_test IS REF CURSOR RETURN table%ROWTYPE;
...or...
DECLARE TYPE cursor_test IS REF CURSOR RETURN table%ROWTYPE;
I still got syntax errors.
In one line, what I am trying to do is break the result array at the database level rather than at the application level.
Cheers, Bill
http://download.oracle.com/docs/cd/B19306_01/appdev.102/b14261/sqloperations.htm#LNPLS00605

I did the following:
OPEN CURSOR x
     LOOP
          FETCH CURSOR x INTO col1, col2
          (EXIT WHEN...)
          variable_line := col1 || col2
     END LOOP
CLOSE CURSOR
But after closing this cursor, variable_line contains only the last buffered line. I want all the looped lines (without using associative arrays, nested tables etc). So I guess I am just looking for some way to append data lines- adding chr(10) doesn't work either.
Cheers, Bill

REF CURSOR as IN parameter to stored procedure

Currently, ODP.NET supports REF CURSOR parameter as OUT parameter only.
Based on the project requirements it is necessary to pass
multiple records to the stored procedure.
In this case REF CURSOR as IN parameter is useful.
What are the plans to implement REF CURSOR as IN parameter to stored procedure?
What is the work around in case it is necessary to pass several different
record types as arrays to stored procedure?

ODP.NET does not limit REF Cursor parameters to IN parameters. This is a known PL/SQL limitation.
An excerpt from Application Developer's Guide:
"If you pass a host cursor variable to PL/SQL, you cannot fetch from it on the server side unless you also open it there on the same server call".

Array/ In Ref cursor

Similar Messages

Maybe you are looking for