Should stored procs internally commit changes or should the caller?

A debate is raging at my workplace between the app developers who write the stored procs used on their back end db and the data warehouse etl developers who are to use said stored procs. The app developers say that stored procs should never internally issue 'commit' statements and that only the calling connection should commit afterwards and that this is a general standard in the database world. The data warehouse etl group insist that it's perfectly OK for a stored proc to issue its own commit.
What are the standards where you work; do sp's commit internally or not? Is one way or the other the exception or the rule in typical Oracle practices?

>
A debate is raging at my workplace between the app developers who write the stored procs used on their back end db and the data warehouse etl developers who are to use said stored procs. The app developers say that stored procs should never internally issue 'commit' statements and that only the calling connection should commit afterwards and that this is a general standard in the database world. The data warehouse etl group insist that it's perfectly OK for a stored proc to issue its own commit.
What are the standards where you work; do sp's commit internally or not? Is one way or the other the exception or the rule in typical Oracle practices?
>
The question is not 'who should commit' it is 'when to commit'. The 'when' will generallyl dictate the 'who'.
A COMMIT should be performed when a unit of work has been completed.
Based on how your post was worded your app developers are WRONG if for no other reason than you simply can't use the word 'never' when you are talking about complex processes.
For the same reason the data warehouse group is RIGHT since they did not say it is ALWAYS ok, just that it is OK for a stored proc to issue its own commit. And
All anyone from either of your groups needs to do is come up with just ONE example where it is OK for a stored proc to issue its own commit and the dev groups position is TOAST; ETL is vindicated once again.
You can't make broad generalizations like NEVER and ALWAYS. So since we are talking COMMIT and not RECOVERY let's ignore flashback and the like.
If a stored proc is designed to perform a truncate and load of a staging table there is absolutely no reason NOT to issue a commit when the load is completed. The TRUNCATE itself will be irreversible; meaning it can't be undone using a simple ROLLBACK.
And if the subsequent load of the table is successful why shouldn't the proc issue the commit? The proc, in this case, is the transaction controller and can determine if the unit of work has been completed. If it has it should issue the COMMIT.
The process architecture needs to be designed to take restart and recovery into account. That design will dictate at what steps in the process work should be committed. The controlling process that determines that the unit of work has been completed successfully is the process that should perform the commit. If that process is part of a master stored procedure the commit should happen in the procedure. If that process is part of an ETL workflow decision the commit should be issued as part of that same workflow.

Similar Messages

BizTalk Stored Proce-passing XML as one of the Input parameter and String as another parameter

I have a requirement in BizTalk that
- I will receive XML from Source and i need to submit this XML data and two other string parameters in SQL storeprocedure as a parameters and submit data
Ex: My_SP(myID Integer INPUT,myXML xml Input,mystring OUT)
Could you please help me how call storeprocedure and submit multiple parameters in BizTalk.

you can execute stored procedure by generating schemas from WCF-SQL Adapter.
for passing parameters you will have to do the mapping to the Generated schema for Stored proc.
I would suggest to do this in Message Assignment shape, there you can easily assign all the parameters.
Integer and String parameters can be assigned from normal variables and XML parameter can be inserted as suggested by Abhishek-
xmldoc=requestMsg;
varOuterstring=xmldoc.Outerxml.Tostring();
Please refer the below article.
https://www.packtpub.com/books/content/new-soa-capabilities-biztalk-server-2009-wcf-sql-server-adapter
http://msdn.microsoft.com/en-us/library/dd787968.aspx
Thanks,
Prashant
Please mark this post accordingly if it answers your query or is helpful.

Change language for the Caller ID announcement

How can I change the Caller ID announcement from English to say German ?
I have installed the German text to speech package and that works as expected for messages text to speech so one would think that can be used for the Caller ID announcement as well.
Any thoughts, help ?
Cheers
chris

Hi MC
In the OP43 transaction (Setup group category-setup group Key SPRo ), select the setup group key for which you want the translation to be maintained, select the GOTO in the menu bar - click on the translation and select the Language for which you want to maintain the translation. Maintain your translated text here.
The other way is to Log in in the language you want to maintain the transaltion Go to OP43 and maintain the details.
You will have to maintain them in SPRO and the same will be reflected for the recipes in the respective login language in the master recipe.
If the solution works for you or is helpful, give suitable points.
Regards,
Amol Kale
Edited by: Amol Kale on Mar 1, 2012 7:09 AM

How can I make a control inside a loop in a subVI change value when the calling VI's control changes?

Hi.
I think this is a pretty basic LV question, but I have not been able to find a good solution.
I am attaching VIs that show the problem I am having, but obviously, the real application is a lot complicated, which forces me to try to do it this way.
The issue is: I have a subVI with a Boolean control inside a loop. When the control is true I want some action to take place. When the subVI is run on its own, it works fine, acting properly when I set the boolean control to true via the LV FPGA interface from the host. However, when I use it as a subVI, in which the top-level VI calls several instances of the subVI, I have the Boolean controls in the top level VI. What is happening is that the false Boolean value with which the top-level VI starts is passed into the subVIs, and not updated, even though the control is inside the loop.
Can any one suggest a good solution?
Thanks,
AlejandroZ
Attachments:
CallingVI.vi ‏7 KB
subVI.vi ‏8 KB

Hi.
I know the example I posted might seem silly, but it was just to illustrate the problem I am having. In reality this is the application:
I have some LV FPGA code which uses a few FPGA IO to implement a serial link to communicate with a device. Most of the time we are getting data from the device, so the serial link is used to send a read command, read in the data and put it into a FIFO. However, I also wanted the VI to support sending data to the device, so I added an array control to put the data you want to send, and a boolean control to tell it you want to send it.
Since sending and receiving data are done using the same FPGA IO, they cannot be independent operations, because they would garble each other. Therefore, in the subVI I have a loop in which I first read data if there is any to read, then check the boolean write control to see if there is data to write.
As I mentioned, this works perfectly for talking to a single device. However, we run into the issue of this topic when trying to replicate this for several devices.
One easy solution is to not have the loop in the subVI, and have it in the calling VI (I am favoring this simple solution right now). The only reason why I have not done this yet, is that the subVI has more than one loop, so I am going to have to create several subVIs. I just posted to see if there was an even simpler solution...
There have been some other possibly good solutions proposed here, though I am not sure if they work in LV FPGA.
Thanks for all your responses.
AlejandroZ

Iphone 5 with IOS 7.02 sensive to change facetime during the call

It was an issue when using Ipgone 5 with ios & above.
when talking it was so eassy to touch the screen and chnage to facetime or mute the call
as i remember from the previous time it need to press the button to call those manual out...
according to this new arrangement, I have more than 50% of call will drop during the call.
plesae help to how prevent this

Hi cltbox,
Thanks for visiting Apple Support Communities.
If your iPhone does not connect to a Wi-Fi network, but other devices can, try these troubleshooting steps first:
Be sure you're in range of your Wi-Fi router (access point).
Tap Settings > Wi-Fi and turn Wi-Fi off and on. If your Wi-Fi setting is dimmed, follow these steps.
Restart your iOS device.
Update your device to the latest version of software.
You can find the troubleshooting article with these steps here:
iOS: Troubleshooting Wi-Fi networks and connections
http://support.apple.com/kb/ts1398
Cheers,
Jeremy

Unable to pass the values to stored proc using presentation variable in OBI

Hi All,
Need your help regarding in resolving an issue in OBIEE 10.1.3.4.1
There is an OBIEE requirement whereby two prompts need to be defined.
1. Textbox prompt
2. Drop-down prompt
The dropdown values should be populated using textbox prompt. So, we have used presentation variable in textbox prompt and passing the same to select query of drop-down.
Until this step, the report works just fine.
Now, the value in both textbox and drop-down needs to be passed to stored proc.
While trying to pass the values by using the presentation variable, the following error comes-up saying
Session variable has no value definition.
Note:Although the corresponding session variable has been set to default value,still the error appears.
Please advise.
Regards,
Prasad

"Session variable has no value definition" I'm assuming typo error and it should be presentation variable.
Set default value for presentation variable that may work

Java Stored Proc, passing BLOB, not stored in a table.

I have checked this article, and found it will not suit my needs.
http://otn.oracle.com/sample_code/tech/java/codesnippet/jdbc/lob/LobToSP.html
I have even looked at the 4 articles on the "JSP How-To Documents" page.
I need to pass in a BLOB to my Java Stored proc, but not store it in the DB - My java code manipulates the LOB and it should never go into the DB. I know I could do the normal store, get an id and pass to my JSP... but that is a poor way to implement this. Are there any code samples on how to do this please ?
D

WOW! That seems like a very bad design.
Is it just the compiler that has the problem or is the PL/SQL code inside the stored procedure also affected? For example, if I'm doing an
EXECUTE IMMEDIATE 'insert ... INTO OtherSchema.tblTest';
Is this going to be a problem?
The problem with granting the access explicitly is the Archive Schema, which has the stored procedure that is failing, might not exist at the time when the other schemas are created. Therefore, there is no user to grant permissions to. I solved this by granting permissions to the role and when the archive user is created it is assigned that role.
This seems like the proper way to solve this problem.

Item Type Based on PL/SQL stored proc causing errors

I am getting the following error when I add this item type based on a pl/sql stored proc that generates html based on the stored proc... the stored proc returns find but the other porlets get Error Message in there headers and in the porlet content I get:
Error: The listener returned the following Message: 503 Service Unavailable....
I also get the same error if I click on a link that is generated from my stored proc for the entire page I get the error:
Error: The listener returned the following Message: 503 Service Unavailable
and get nothing else returned...

Greetings
I suggest you take a look at the following page and have a look on how to create portlets. Since the PL/SQL PDK allows you to create your own portlets inside the Oracle Database there should be no problem for creating the content inside a portlet the way you want it.
http://www.oracle.com/technology/products/ias/portal/pdk.html
Best regards
Johan

Java stored proc status

hello.
What would trigger the change of "STATUS" from valid to invalid
for a java stored procedure (status of the PROCEDURE, not the
java class/source)?
E.g.
- loaded a java source/class
- published it with CREATE PROCEDURE ...
- the procedure inserts a row into some table X
- select* from user_objects where object_type = 'PROCEDURE'; shows that
the proc is valid
- call the proc, fine
- delete a column from the table X
- select* from user_objects where object_type = 'PROCEDURE'; shows that
the proc is still VALID
- call the proc, the call actually completed without error! (but the new row didn't get inserted into the table)
When the schema of a table that a java stored proc refers
to changes, will the java stored proc automatically marked
as "INVALID"? If not, when will this status changed from
VALID to INVALID?
Thanks.
Min

Min:
A java stored procedure is marked as VALID or INVALID depending on whether Oracle had any problems loading the class file into the database. Since you had no problems loading the class file it will always be valid.
Now, after you deleted the column in the table I would assume that the java stored procedure would have a problem when it tried to do an insert and it would throw an exception. Your java stored procedure is VALID, it just isn't executing properly and is more than likely throwing an exception.
Sometimes an exception thrown from a jsp is written to a log file instead of to the console. I suspect that this is the case in your situation. You think that the call to the jsp executed with no errors, but I bet it didn't. You will probably find the errors somewhere in the oracle\ora90\admin\mydatabase directory. Look at the latest .trc file. (Note: the path I gave to the .trc files may be different depending on what version of oracle you are using).

Clearing Oracle Parameters to call diferent stored procs

Sub Main
Do While i < 3
make_excel(i)
i = i + 1
Loop
end Sub
Sub make_excel(ByVal array As Integer)
objCmd.Parameters.Add(New OracleParameter("p_cursor", OracleType.Cursor)).Direction = ParameterDirection.Output
end Sub
objCmd.Parameters.Clear()
objCmd.Parameters.Remove("p_cursor")

I agree with you. I was looking into mod_plsql, but it turns out that their development standards include not using Oracle HTTP server. The whole reason why they want to be able to execute a stored proc based on what is specified in an XML doc is that they want to avoid having to change their middle-tier configurations anytime a new stored proc or a change in stored proc parameters is made. I'm not familiar with what all is involved with .NET and the middle-tier, but supposedly this way, they can specify any stored procedure name and its parameters in an XML file. The XML is then suppose to be passed on to an Oracle stored procedure which will parse the XML and dynamically execute the stored procedure that was specified in the XML.
Here is an example of the XML:
'<Root>
<PackageName>TEST_PKG</PackageName>
<ProcedureName>TEST_PROC</ProcedureName>
    <Parameters>
    <Parameter>
        <Name>EmpID</Name>
        <Value>12345</Value>
    </Parameter>
    <Parameter>
        <Name>Org</Name>
        <Value>ABC</Value>
    </Parameter>
    </Parameters>
</Root>I basically need to parse out the pkg/proc names:
SELECT t.COLUMN_VALUE.extract('//PackageName/text()').getstringval() PkgName,
         t.COLUMN_VALUE.extract('//ProcedureName/text()').getstringval() ProcName
     INTO v_pkg_name, v_proc_name
     FROM TABLE(xmlsequence(XMLTYPE(v_XML_input) .extract('/Root'))) t;...and then execute the procedure:
    EXECUTE IMMEDIATE 'BEGIN '||v_pkg_name||'.'||v_proc_name||'(:a, :b, :c); END;'
      using in v_in_param1, v_in_param2, out v_XML_output;The problem is that this approach is very complicated since there can be any number of IN/OUT parameters and of various datatypes. I would have to create all kinds of possible bind variables!

Is there a way to call a stored proc from the web in Oracle 10g?

I've found an article about Native Oracle XML DB Web Services in 11g, but it appears to be a new feature. Is there any way of accomplishing something similar in 10g? I would like to be able to process XML documents that contain the name and parameters of the stored proc/function to execute.
Link to 11g article:
http://download.oracle.com/docs/cd/B28359_01/appdev.111/b28369/xdb_web_services.htm

I agree with you. I was looking into mod_plsql, but it turns out that their development standards include not using Oracle HTTP server. The whole reason why they want to be able to execute a stored proc based on what is specified in an XML doc is that they want to avoid having to change their middle-tier configurations anytime a new stored proc or a change in stored proc parameters is made. I'm not familiar with what all is involved with .NET and the middle-tier, but supposedly this way, they can specify any stored procedure name and its parameters in an XML file. The XML is then suppose to be passed on to an Oracle stored procedure which will parse the XML and dynamically execute the stored procedure that was specified in the XML.
Here is an example of the XML:
'<Root>
<PackageName>TEST_PKG</PackageName>
<ProcedureName>TEST_PROC</ProcedureName>
    <Parameters>
    <Parameter>
        <Name>EmpID</Name>
        <Value>12345</Value>
    </Parameter>
    <Parameter>
        <Name>Org</Name>
        <Value>ABC</Value>
    </Parameter>
    </Parameters>
</Root>I basically need to parse out the pkg/proc names:
SELECT t.COLUMN_VALUE.extract('//PackageName/text()').getstringval() PkgName,
         t.COLUMN_VALUE.extract('//ProcedureName/text()').getstringval() ProcName
     INTO v_pkg_name, v_proc_name
     FROM TABLE(xmlsequence(XMLTYPE(v_XML_input) .extract('/Root'))) t;...and then execute the procedure:
    EXECUTE IMMEDIATE 'BEGIN '||v_pkg_name||'.'||v_proc_name||'(:a, :b, :c); END;'
      using in v_in_param1, v_in_param2, out v_XML_output;The problem is that this approach is very complicated since there can be any number of IN/OUT parameters and of various datatypes. I would have to create all kinds of possible bind variables!

Delete/update statements insite stored proce

I know this is probably fairly simple for some of you, but I am brand new to Oracle-SQL. In Sybase you can have the following stored procedure and works:
create procedure foo as
begin
create table ...
create index...
delete x
where ....
update x
where ...
end
In oracle everytime I need to execute a DML comand in a stored proc do I have to use execute immediate ('delete x...')?
thanks

> the above is not an application code, it is a simple
and dirty maybe way to transfer some data from
microsoft access to oracle. so, i need a quick and
disrty way to transfer this data.
I would have been using a SQL*Plus script myself to do this and not a stored procedure.
> is it there a limit of how many tables i can cerate
inside my stored proc using:
execute immediate('CREATE TABLE xxx AS select * from
yyy.... ?
No. Of course not.
Just keep in mind that each DDL is an implicit commit.
> i have a stored proc that creates couple tables with
the method shown above but the compiler compains for
the second create table zzz as select * from .... any
idea why?
DDL statement and error number are?

Stored Proc running twice using DBMS_Scheduler

Hello all,
I have a vb front end that calls a main stored proc which submits scheduler jobs to execute several stored procs asynchronously. Everything is working, except the part that the several stored procs are running twice. In the troubleshooting, I have eliminated the front end from being the culprit and the stored procs themselves. Essentially, when I call the stored proc using dbms_scheduler.create_job, it runs twice, even manually. I am about at wits end trying to figure out why: Using Oracle 11gR2
I started off setting up the programs
begin
--create program
dbms_scheduler.create_program
( program_name => 'prog_name'
,program_type => 'STORED_PROCEDURE'
,program_action => 'usp_sub_proc_1'
,number_of_arguments => 8
,enabled => FALSE
dbms_scheduler.DEFINE_PROGRAM_ARGUMENT
( program_name=> 'prog_name'
,argument_position=>1
,argument_name => 'name'
,argument_type=>'VARCHAR2'
/*the remaining 7 arguments are in code but not display for space reasons*/
dbms_scheduler.enable('prog_name');
end;Then the main stored proc executes this code:
declare v_job_name varchar2(100);
    v_1 varchar(50) := 'All';
    v_2 varchar(50) := 'All';
    v_3 varchar(50) := 'All';
    v_4 varchar(50) := 'All';
    v_5 varchar(50) := 'TEST';
    i_6 integer := 1;
    v_7 varchar(50) := 'TEST_NE';
    ts_8 timestamp := current_timestamp;
begin
    v_job_name := 'uj_dmo_1';
dbms_scheduler.create_job (v_job_name
                                        ,program_name => 'prog_name'
                                        ,job_class => 'UCLASS_1'
                                        ,auto_drop => TRUE
--set parameters
dbms_scheduler.set_job_argument_value(v_job_name,1, v_1);
dbms_scheduler.set_job_argument_value(v_job_name,2, v_2);
dbms_scheduler.set_job_argument_value(v_job_name,3, v_3);
dbms_scheduler.set_job_argument_value(v_job_name,4, v_4);
dbms_scheduler.set_job_argument_value(v_job_name,5, v_5);
dbms_scheduler.set_job_argument_value(v_job_name,6, to_char(i_6));
dbms_scheduler.set_job_argument_value(v_job_name,7, v_7);
dbms_scheduler.set_job_argument_value(v_job_name ,8, to_char(ts_8));
--enable job
dbms_scheduler.enable(v_job_name);
--execute job
dbms_scheduler.run_job(job_name => v_job_name , use_current_session => FALSE);
end;
...And this is where I get the double execution of the job, but I am just not seeing it in my syntax, dba_scheduler_jobs, logging, etc. Any help is greatly appreciated, thanks!!

Well apparently I will not win any Captain Obvious awards;
With 34MCA2K2's response with "what doesn't work" for some reason turned the light on. After some more testing here is what I found.
This code works as expected :
Exhibit A
begin
dbms_scheduler.create_job (job_name =>'TESTER'
                               ,job_type => 'PLSQL_BLOCK'
                               ,job_action => 'declare test1 integer := 1; begin test1 := test1 + 5; end;'
                               ,auto_drop => True
   /*dbms_scheduler.enable('TESTER');   */
   dbms_scheduler.run_job(job_name => 'TESTER', use_current_session =>FALSE);
end;As does this:
Exhibit B
begin
dbms_scheduler.create_job (job_name =>'TESTER'
                               ,job_type => 'PLSQL_BLOCK'
                               ,job_action => 'declare test1 integer := 1; begin test1 := test1 + 5; end;'
                               ,auto_drop => True
   dbms_scheduler.enable('TESTER');
/*dbms_scheduler.run_job(job_name => 'TESTER', use_current_session =>FALSE);    */
end;Exhibit A will create the job and is visible in the schedulerjobs view, and the RUN_JOB will execute it even when not enabled, but the pl/sql will not drop the job.
Exhibit B will create the job and once enabled, executes the job and then drops from schedulerjobs view.
Therefore, my desired results for running the jobs once asynchronously and dropping immediately is....
begin
    v_job_name := 'uj_dmo_1';
dbms_scheduler.create_job (v_job_name
                                        ,program_name => 'prog_name'
                                        ,job_class => 'UCLASS_1'
                                        ,auto_drop => TRUE
--set parameters
dbms_scheduler.set_job_argument_value(v_job_name,1, v_1);
dbms_scheduler.set_job_argument_value(v_job_name,2, v_2);
dbms_scheduler.set_job_argument_value(v_job_name,3, v_3);
dbms_scheduler.set_job_argument_value(v_job_name,4, v_4);
dbms_scheduler.set_job_argument_value(v_job_name,5, v_5);
dbms_scheduler.set_job_argument_value(v_job_name,6, to_char(i_6));
dbms_scheduler.set_job_argument_value(v_job_name,7, v_7);
dbms_scheduler.set_job_argument_value(v_job_name ,8, to_char(ts_8));
/*enable job*/
dbms_scheduler.enable(v_job_name);
/*execute job (Do not execute the code below, it will lead to multiple executions)
dbms_scheduler.run_job(job_name => v_job_name , use_current_session => FALSE); */
end;

Calling a COBOL stored proc from Java Servlet

I am trying to call a COBOL stored proc from a Java Servlet. The stored proc is stored on a DB2 database. I need to send 6 inputs to the COBOL stored proc and the output will be the return code of the stored proc. I'm not sure if I'm going about this the right way. This is how my code looks...
public int callStoredProc(CallableStatement cstmt,
Connection con,
String sYear,
String sReportNbr,
String sSystemCode,
String sUserId,
String sModuleNbr,
String sFormId){
int iParm1 = 0;
try{
Class.forName ("sun.jdbc.odbc.JdbcOdbcDriver");
catch(ClassNotFoundException ex){
System.out.println("Failed to locate database driver: "
+ ex.toString());
return iParm1;
try{
cstmt = con.prepareCall("{? = CALL MKTPZ90C
cstmt.registerOutParameter(1, Types.INTEGER);
cstmt.setString(2, sYear);
cstmt.setString(3, sReportNbr);
cstmt.setString(4, sSystemCode);
cstmt.setString(5, sUserId);
cstmt.setString(6, sModuleNbr);
cstmt.setString(7, sFormId);
cstmt.execute();
iParm1 = cstmt.getInt(1);
CloseSQLStatement(cstmt);
catch(SQLException ex) {
CloseSQLStatement(cstmt);
System.out.println("SQL exception occurred:" +
ex.toString());
return iParm1;
return iParm1;
Could someone tell me if this is the right way to go about doing this?
Thanks!!!!!!

I didn't see the code where you create the database connection (variable "con"). However, the answer to your question "Is this the right way...", for me, is "Anything that works is the right way." So try it. That's a first approximation, but once you have something that works you can start on improving it, if that becomes necessary.

Handling result form Stored Proc in java program

Folks, I have a question on how to handle results from Stored Procedures with the java.sql API. I execute a stored proc from a java program using the statement:
statement.execute();
where 'statement' is of type Statement. Then I get the results:
ResultSet rs = query.getResultSet();
The above returns me a ResultSet object. Now, my stored proc is such that it will return an integer in case of errors (as error code), and, if no error,it'll return the result set. Because I wouldn't know if the stored proc is returning an integer or a result set, how do I get the result of the stored proc in the java program? 'query.getResultSet()' would get me only an object of type ResultSet. What if the stored proc is returning an integer (i.e. when an error occurs)?
Thanks.

GSP wrote:
Thanks to all for your replies. I do not have access to modify the stored procedure. I can just use it in my java program. The stored proc first validates its input parameters. If it finds them invalid, then it returns an appropriate error code (which is an integer) depending on which input param is found invalid. If all the input parameters are found valid, then it fetches the rows from the DB tables & returns them as result set. Now my question is: say if I give a statement as this in my java program:
ResultSet rs = query.getResultSet();
what if the stored proc returns an error code (Since the above statement gets only ResultSet object, how will it handle if the stored proc returns an int)? Is there any alternative?
Ok, so there is a piece of missing data.
Store procs, conceptuatlly can return data in a variety of ways.
So the first step is to determine how the data is being returned.
Unless you know that there is no way to determine how to use it in java.

Should stored procs internally commit changes or should the caller?

Similar Messages

Maybe you are looking for