Logical Delete in DB Polling - OSB

Hello All,
I have a question in polling. I have a logical delete column with Read value as 'P and unread value as 'N', Unlike BPEL, OSB's polling, does not make a record to 'P' until the process completes successfully. My DB polling adapter polls the same records while the previous instance is under processing or the previous instance ended in error. Is there a way to logically delete the record immediately once the record is read?
To avoid this scenario, I added an error handler to make the record to 'E' if the instance encountered any error. By the time the error handler kicks in and updates the record, OSB polls the record a few times. I increased my polling frequency from 5 seconds to 30 seconds, but no luck. Any clue how to handle this scenario?
Thanks,
Dwarak

Is there a way to logically delete the record immediately once the record is read?Dont do any logic in db adapter proxy. Instead make the proxy to just write to a jms queue and then have your processing logic in the jms proxy service which reads off the jms queue.

Similar Messages

DBAdapter polling with logical delete x distrib polling x DB rows per trans

Hi all.
I'm trying to configure a DBAdapter with "logical delete" polling strategy, distributed polling (cluster environment) and a defined number of "Database Rows per Transaction".
When I check the box "Distributed Polling", the SQL generated gets appended by "FOR UPDATE NOWAIT"
However, when I set a value for "Database Rows per Transaction" the "FOR UPDATE NOWAIT" sql clause disappear.
Is this a bug, or some limitation related to the "logical delete" strategy???
Thanks
Denis

Hi All,
Actually I'm also facing the same problem.
Step I follwed:
1) Created a job_table in database
create talbe job_table(id, job_name, job_desc, job_status)
2)created a bpel process to test the Inbound distributed polling.
3)Configure the DBAdapter for polling.
a)update a field in the job_table with logical delete.
b)select the field name form the drop down
c) change the read value-->Inprogress and unRead value--->Ready
d) dont change the value for Reserved value
e) select the check box for "distributed polling".
f) the query will be appended with "For update NoWait."
g)click next and then finish.
4) Then i followed the below steps.
To enable pessimistic locking, run through the wizard once to create an inbound polling query. In the Applications Navigator window, expand Application Sources, then TopLink, and click TopLink Mappings. In the Structure window, click the table name. In Diagram View, click the following tabs: TopLink Mappings, Queries, Named Queries, Options; then the Advanced… button, and then Pessimistic Locking and Acquire Locks. You see the message, "Set Refresh Identity Map Results?" If a query uses pessimistic locking, it must refresh the identity map results. Click OK when you see the message, "Would you like us to set Refresh Identity Map Results and Refresh Remote Identity Map Results to true?Ó Run the wizard again to regenerate everything. In the new toplink_mappings.xml file, you see something like this for the query: <lock-mode>1</lock-mode>.
5) lock-mose is not changed to 1 in toplink_mappingss.xml
Can we edit the toplink_mappings.xml manually.
If yes, what allt he values Ineed to change in toplink_mappings.xml file, so that it will not pick the same record for the multiple times in clustered environment.
Please help me out this is urgent.
Thanking you in advance.

Duplicate processing by DBAdapter when using Distributed Polling with Logical Delete Strategy

We have DBAdapter based polling services in OSB running across two Active-Active clusters (total 20 managed service across 2 clusters),
listening to the same database table. (Both clusters read from the same source DB). We want to ensure distributed polling without duplication,
hence in the DBAdapter we have selected Distributed Polling option, meaning we are using "Select For Update Skip Locking".
But we see that sometimes, same rows are processed by two different nodes and transactions are processed twice.
How do we ensure that only one managed server processes a particular row using select for update? We do not want to use the markReservedValue option which was preferred in older version of DBAdapter.
We are using following values in DB Adapter configuration, the Jdev project for DBAdapter and the OSB proxy using DBAdapter are attached.
LogicalDeletePolling Strategy
MarkReadValue = Processed
MarkUnreadValue = Initiate
MarkReservedValue = <empty as we are using Skip Locking>
PollingFrequency = 1 second
maxRaiseSize = 1
MaxTransactionSize = 10
DistributionPolling = checked (adds lock-n-wait in properties file and changes the SQL to SELECT FOR UPDATE SKIP LOCKED)
Thanks and Regards

Hi All,
Actually I'm also facing the same problem.
Step I follwed:
1) Created a job_table in database
create talbe job_table(id, job_name, job_desc, job_status)
2)created a bpel process to test the Inbound distributed polling.
3)Configure the DBAdapter for polling.
a)update a field in the job_table with logical delete.
b)select the field name form the drop down
c) change the read value-->Inprogress and unRead value--->Ready
d) dont change the value for Reserved value
e) select the check box for "distributed polling".
f) the query will be appended with "For update NoWait."
g)click next and then finish.
4) Then i followed the below steps.
To enable pessimistic locking, run through the wizard once to create an inbound polling query. In the Applications Navigator window, expand Application Sources, then TopLink, and click TopLink Mappings. In the Structure window, click the table name. In Diagram View, click the following tabs: TopLink Mappings, Queries, Named Queries, Options; then the Advanced… button, and then Pessimistic Locking and Acquire Locks. You see the message, "Set Refresh Identity Map Results?" If a query uses pessimistic locking, it must refresh the identity map results. Click OK when you see the message, "Would you like us to set Refresh Identity Map Results and Refresh Remote Identity Map Results to true?Ó Run the wizard again to regenerate everything. In the new toplink_mappings.xml file, you see something like this for the query: <lock-mode>1</lock-mode>.
5) lock-mose is not changed to 1 in toplink_mappingss.xml
Can we edit the toplink_mappings.xml manually.
If yes, what allt he values Ineed to change in toplink_mappings.xml file, so that it will not pick the same record for the multiple times in clustered environment.
Please help me out this is urgent.
Thanking you in advance.

How to specify custom SQL in polling db adapter with logical delete option

Hi all,
I am writing a SOA composite app using JDeveloper SOA Suite 11.1.1.4 connecting to a SQL Server db using a polling DB Adapter with the logical delete option to send data to a BPEL process.
I have requirements which go beyond what is supported in the JDeveloper UI for DB Adapter polling options, namely:
* update more than one column to mark each row read, and
* specify different SQL for the logical delete operation based on whether bpel processing of the data polled was successful or not.
A complicating factor is that the polling involves two tables. Here is my full use-case:
1) Polling will select data derived from two tables: e.g. 'headers' and 'details' simplified for this example:
table: headers
hid - primary key
name - data label
status - 'unprocessed', 'processed', or 'error'
processedDate - null when data is loaded, set to current datetime when row is processed
table: details
hid - foreign key pointed at header.hid
attr - data attribute name
value - value of data attribute
2) There is a many:1 relationship between detail and header rows through the hid columns. The db adapter polling SELECT shall return results from an outer join consisting of one header row and the associated detail rows where header.status = 'unprocessed' and header.hid = details.hid. (This is supported by the Jdeveloper UI)
3) The polled data will be sent to be processed by a bpel process:
3.1) If the bpel processing succeeds, the logical delete (UPDATE) operation shall set header.status = 'processed', and header.processedDate = 'getdate()'.
3.2) If bpel processing fails (e.g. hits a data error while processing the selected data) the logical delete (UPDATE) operation shall set header.status = 'failed', header.processedDate = 'getdate()', and header.errorMsg = '{some text returned from bpel}'.
Several parts of #3 are not supported by the JDeveloper UI: updating multiple columns to mark the row processed, using getdate() to populate a value of one of those column updates, doing different update operations based on the results of the BPEL processing of the data (success or error), and using data obtained from BPEL processing as a value of those column updates (error message).
I have found examples which describe specifying custom SQL using the polling delete option to create a template then modifying the toplink file(s) to specify custom select and update SQL to implement a logical delete. (e.g. http://dlimiter.wordpress.com/2009/11/05/advanced-logic-in-oracle-bpel-polling-database-adapter/ and http://myexperienceswithsoa.blogspot.com/2010/06/db-adapter-polling-tricks.html). But none of them match what I've got in my project, in the first case because maybe because I'm using a higher version of JDeveloper, and in the second I think because in my case two tables are involved.
Any suggestions would be appreciated. Thanks, John

Hi John,
You've raised a good scenario.
First of all let me say that the purpose of the DB polling transaction, is to have an option to initiate a process from a DB table/view and not to update multiple fields in a table (or have other complex manipulation on the table).
So, when choose to update a field in a record, after reading it, you are "telling" the engine not to poll this record again. Sure, i guess you can find a solution/workaround for it, but I don't think this is the way....
The question now is what to do?
You can have another DB adapter where you can update the data after finishing the process. In that case, after reading the data (on polling transaction) - update the header.status = 'processed' for example, and after processing the selected data update the rest of the fields.
Hope it make some sense to you.
Arik

Database Adapter using Logical Delete Polling Strategy not updating field

I have an ESB database adapter defined against a table. The adapter is set to use the Logical Delete Polling Strategy, and the table has an extra column defined as the "delete" field. When I register the adapter to the ESB and add a record to the table, the adapter reads the contents of the record as expected. However, 15 seconds later (that being the polling interval) it reads it again, and again 15 seconds after that, ad infinitum.
I have defined the logical delete field as a single character field and set the Read value for "T" (a record is normally inserted with that field having a null value). Results as outlined in the previous paragraph. I redefined the field as a number and set the Read value to "1" (with a record normally inserted with that field having a value of "0"). Results as outlined in the previous paragraph.
Apparently the update of the logical delete value to the Read value is not occurring. My question (obviously) is: Why?
Thanks for your time.

Here is the adapter WSDL:
<?xml version="1.0" encoding="UTF-8"?>
<definitions
name="Poll_PM_LOG"
targetNamespace="http://xmlns.oracle.com/pcbpel/adapter/db/Poll_PM_LOG/"
xmlns:tns="http://xmlns.oracle.com/pcbpel/adapter/db/Poll_PM_LOG/"
xmlns:plt="http://schemas.xmlsoap.org/ws/2003/05/partner-link/"
xmlns:jca="http://xmlns.oracle.com/pcbpel/wsdl/jca/"
xmlns:pc="http://xmlns.oracle.com/pcbpel/"
xmlns:top="http://xmlns.oracle.com/pcbpel/adapter/db/top/PollPMLOG"
xmlns="http://schemas.xmlsoap.org/wsdl/">
<types>
<schema xmlns="http://www.w3.org/2001/XMLSchema">
<import namespace="http://xmlns.oracle.com/pcbpel/adapter/db/top/PollPMLOG"
schemaLocation="PollPMLOG_table.xsd"/>
</schema>
</types>
<message name="LogCollection_msg">
<part name="LogCollection" element="top:LogCollection"/>
</message>
<portType name="Poll_PM_LOG_ptt">
<operation name="receive">
<input message="tns:LogCollection_msg"/>
</operation>
</portType>
<binding name="Poll_PM_LOG_binding" type="tns:Poll_PM_LOG_ptt">
<pc:inbound_binding/>
<operation name="receive">
<jca:operation
ActivationSpec="oracle.tip.adapter.db.DBActivationSpec"
DescriptorName="PollPMLOG.Log"
QueryName="Poll_PM_LOG"
PollingStrategyName="LogicalDeletePollingStrategy"
MarkReadFieldName="PROCESSED"
MarkReadValue="T"
SequencingFieldName="ID"
MaxRaiseSize="1"
MaxTransactionSize="unlimited"
PollingInterval="15"
NumberOfThreads="1"
UseBatchDestroy="false"
ReturnSingleResultSet="false"
MappingsMetaDataURL="PollPMLOG_toplink_mappings.xml" />
<input/>
</operation>
</binding>
<service name="Poll_PM_LOG">
<port name="Poll_PM_LOG_pt" binding="tns:Poll_PM_LOG_binding">
<jca:address location="eis/DB/PM"
UIConnectionName="PM"
ManagedConnectionFactory="oracle.tip.adapter.db.DBManagedConnectionFactory"
/>
</port>
</service>
<plt:partnerLinkType name="Poll_PM_LOG_plt" >
<plt:role name="Poll_PM_LOG_role" >
<plt:portType name="tns:Poll_PM_LOG_ptt" />
</plt:role>
</plt:partnerLinkType>
</definitions>
The field PROCESSED is defined as CHAR(1), and is normally null when a record is inserted. The value of the field is to be set to 'T' when the record is read by the polling adapter.

Databse polling without using logical delete or physical delete

Hi all,
Without using logical delete or physical delete how can i poll to the database????
Thanks
Phani

Whatever table are you interested in polling, have a duplicate table for it. Whenever an event (update, delete, insert etc) happens on the source table, fire a trigger to perform that activity in your dup table. Now, from the other tier poll a query for a change. Once the change is consumed clear the dup table.
Hope this help!
Regards,
Yousuf Baig

Logical delete in database adapter

Hello
I was wondering if someone has solution the problem with polling database. You can specify the logical delete column and you can give values for READ, UNREAD and RESERVED states. The problem is that when for example ESB project polls some specific table and starts an instance for every new row with specified logical delete field with value UNREAD, when something unexpected happens and something goes wrong the database adapter updates the row with READ value. This is problematic if we have thousands of rows, and we would like to separate the errored rows from the successfully read rows. Is there anyway (easy) way to update those rows that went wrong to some other value than READ?
I don't know if anyone understood me, but just for clarification here's a example:
I have a ESB-project which poll specific database table and parses and XML from the data. After this the ESB-project sends the data to some Web Service. The database table has column CONDITION_CODE in which value 0 means unread and value 1 means read. Now if everything goes fine there is no problems. But if the Web Service is unavailable or the data is malformed, the database adapter still updates the CONDITION_CODE to 1! We have no ways (except to listen ESB_ERROR topic and implement some error handling there) to know what rows were successfully delivered and which were not...
Hope I was able to clarify the problem... And I hope someone could be able to provide me with answer.
Best Regards Tuomas

Did you use the RESERVED value property? How about the transaction mechanism? Do you have global transactions? I gues you would have to use them!

Logical delete in a DBAdapter

Hi,
I am using a logical delete to poll a table. A simple logic of unproccesed records assigned to 'N', processed ones to 'Y' and locked once to 'X' doesnt seem to work. The process doesnt pickup any inserted records.
Are any other additional settings required for this logical delete to get working.
Thanks,
Valli.

When you have the table definition which has the primary key defined already, the database adapter directly takes the field and constructs a where condition(update tbl set col1="", col2="" where primarykey=#input) based on the primary key; in this case your input should have the primary key field mapped from the BPEL process!
If you don't have a primary key defined on that you obviously have to choose one otherwise db adapter will not proceed further and rest all the is the same logic!
You can also observe that you can't select only few fields for updation as the XML schema generated will have all the columns available in the table. You can simply assign/tranform input data to what ever the fields which require for updation and rest all will not be treated as null but they will have the same value which is already existing in the database!!
Hope it helps!! ( You can still have a question !! how do I change my primary key if required by the business !! ) - well I have no answer for this yet !!

Logical delete in DB adapter with 2 columns to update

Hi',
I am using Oracle SOA 11G.
I am polling a DB table and after reading the data I have to update 2 columns.
With logical delete we can update 1 field, is it possible to update 2 fields.
If Yes, then I have to update the first column BPEL_FLAG = R and BPEL_TIME = currentTime()
Please advice,
Thanks
Yatan

Hi Yatan,
Came across the possible solution at this link.
Could not try and validate though.
http://myexperienceswithsoa.blogspot.in/2010/06/db-adapter-polling-tricks.html
Please try and let us know if it works.
Thanks,
Deepak.

Unable to update Logical Delete field in AS400

hi all,
We have an ESB process that polls for data in AS400 DB and transfers this data into an Oracle DB. This polling is based on Logical Delete. The data transferred is in the range of some lakhs of records. So we are sending around 5000 records per transaction in ESB. When we deploy the process, the data in AS400 is being read and getting transferred. However, the status field is not changing to 'Read Value'. Due to this, we are facing a lot of issues like, when we're in middle of data transfer and the opmn restarts, the data which has already been transferred is also being read and this is throwing an error because of primary constraint violation. Even after the data transfer, if we delete any rows in Oracle DB, the data transfer resumes as the status field is not updated.
I have worked with logical delete in Oracle DB before and was working fine.
Please help me with Logical Delete functionality in case of AS400 Database and also please let me know the privilages we need to have on the AS400 DB to successfully run this process.
Thanks,
Kamal.

Hi,
According to your post, my understanding is that you was unable to update managed metadata field in designer workflow.
You need to provide the exact string for the MMS value, in the form of <id>;<value>.
The format of the value you wanted to set is uncorrectly, so you cannot update the managed metadata field.
If you want to set the managed metadata field value with correct format, you’d better to create custom action.
There is an article for your reference, although it is about the SharePoint 2010, it is similar to SharePoint 2013.
http://patrickboom.wordpress.com/2013/07/23/workflow-activity-set-managed-metadata-column/
Best Regards,
Linda Li
Linda Li
TechNet Community Support

DB Adapter for Logical Delete - selecting multiple StatusesAND

Hi,
I am using SOA 11.1.1.3 and my requirement is to use a DB Adapter to poll both ERROR & NEW Records. Once Polled, I want the status to be changed to "PICKED" (Logical Delete)
The DB Adapter only allows for a single Unread Value to be selected. I tried to edit the SQL to use a "OR" condition but the code always replaces it with an "AND":
Polling SQL:
SELECT QUEUE_ID, BATCH_ID, STATUS, DEL_ID, ORDER_NUMBER, HEADER_ID, LINE_ID, ORD_TYPE, OPERATION, ERR_MESG, CREATE_DT, INV_ORG_ID, PROCESS_DATE FROM DEV.DEV_Q WHERE ((STATUS = 'ERROR') AND (STATUS = 'NEW')) ORDER BY QUEUE_ID ASC
After Read SQL:
UPDATE DEV.DEV_Q SET STATUS = 'PICKED' WHERE (QUEUE_ID = #QUEUE_ID)
I tried to edit the SQL in the <PartnerLink>-or-mappings.xml but it isn't working.
I did read some posts http://blogs.oracle.com/soa_how_to/2009/11/a_db_adapter_sample_for_pure_sql_and_logicaldeletepollingstrategy_scenario.html that mention that sql edits can be made in DB Adapter only if a Delete is used.
Question is - Where can I get the sequel generated (it is not there in the TOPLINK mapping) in order to edit it? Can someone please provide the steps?
Thx a ton!
Suhas.
Edited by: user13616720 on Apr 28, 2011 3:42 AM

Did you tried modifying SQL to
WHERE STATUS in ('ERROR','NEW')or
where regexp_like(status,'ERROR|NEW')Thanks
AJ

DB Adapter for Logical Delete - selecting multiple Statuses

Hi,
I am using SOA 11.1.1.3 and I am using a DB Adapter to poll ERROR/NEW Records. Once Polled, I want the status to be changed to "PICKED" (Logical Delete)
The DB Adapter only allows for a single Unread Value to be selected. I tried to edit the SQL to use a "OR" condition but the code always replaces it with an "ADD":
Polling SQL:
SELECT QUEUE_ID, BATCH_ID, STATUS, DEL_ID, ORDER_NUMBER, HEADER_ID, LINE_ID, ORD_TYPE, OPERATION, ERR_MESG, CREATE_DT, INV_ORG_ID, PROCESS_DATE FROM DEV.DEV_Q WHERE ((STATUS = 'ERROR') AND (STATUS = 'NEW')) ORDER BY QUEUE_ID ASC
After Read SQL:
UPDATE DEV.DEV_Q SET STATUS = 'PICKED' WHERE (QUEUE_ID = #QUEUE_ID)
I tried to edit the SQL in the +<PartnerLink>-or-mappings.xml+ but it isn't working.
I did read some posts http://blogs.oracle.com/soa_how_to/2009/11/a_db_adapter_sample_for_pure_sql_and_logicaldeletepollingstrategy_scenario.html that mention that sql edits can be made in DB Adapter only if a Delete is used.
Question is - Where can I get the sequel generated (it is not there in the TOPLINK mapping) in order to edit it? Can someone please provide the steps?
Thx a ton!
Suhas.
Edited by: user13616720 on Apr 28, 2011 3:37 AM

Hi,
wrong form. Use the SOA forum here: SOA Suite
Frank

How to hide row from table after logical delete

Hello.
I am using Jdeveloper 11.1.1.3.0, ADF BC and ADF Faces.
I want to implement Logical delete in my application.
In my Entity object I have Deleted attribute and I override the remove() method in my EntityImpl class.
    @Override
    public void remove()
       setDeleted("Y");
    }and I added this condition to my view object
WHERE NVL(Deleted,'N') <> 'Y'in my page I have a table. this table has a column to delete each row. I dragged and drop RemoveRowWithKey action from the data control
and set the parameter to *#{row.rowKeyStr}* .
I what I need is this:
when the user click the delete button I want to hide the roe from the table. I tried to re-execute the query after the delete but the row is still on the page. Why execute query does not hide the row from the screen.
here is the code I used for delete and execute query
    public String deleteLogically()
        BindingContainer bindings = getBindings();
        OperationBinding operationBinding = bindings.getOperationBinding("removeRowWithKey");
        Object result = operationBinding.execute();
        DCBindingContainer dc=(DCBindingContainer) bindings;
        DCIteratorBinding iter=dc.findIteratorBinding("TakenMaterialsView4Iterator");
        iter.getCurrentRow().setAttribute("Deleted", "Y");
        //iter.getViewObject().executeQuery();
        iter.executeQuery();
        return null;
    }as you see I used two method iter.getViewObject().executeQuery(); and iter.executeQuery(); but the result is same.

Thank you Jobinesh.
I used this method.
    @Override
    protected boolean rowQualifies(ViewRowImpl viewRowImpl)
      Object attrValue =viewRowImpl.getAttribute("Deleted");
        if (attrValue != null) {
         if ("Y".equals(attrValue))
            return false;
         else
            return true;
        return super.rowQualifies(viewRowImpl);
    }But I have one drawback for using it, and here is the case:
If the user clicks the delete button *(no commit)* the row will be hidden in the table, but when the user click cancel changes the row is not returned since it is not returned due to the rowQualifies(ViewRowImpl viewRowImpl) (the Deleted attribute is set to "N" now).
here is the code for delete and cancel change buttons
    public String deleteLogically()
        BindingContainer bindings = getBindings();
        OperationBinding operationBinding =
            bindings.getOperationBinding("removeRowWithKey");
        Object result = operationBinding.execute();
        DCBindingContainer dc = (DCBindingContainer)bindings;
        DCIteratorBinding iter =
            dc.findIteratorBinding("TakenMaterialsView4Iterator");
        iter.getCurrentRow().setAttribute("Deleted", "Y");
         iter.executeQuery();
        AdfFacesContext adfFacesContext = AdfFacesContext.getCurrentInstance();
        adfFacesContext.addPartialTarget(this.getTakenMaterialsTable());
        return null;
    public String cancelChanges(String iteratorName)
        System.out.println("begin cancel change");
        BindingContainer bindings =
            BindingContext.getCurrent().getCurrentBindingsEntry();
        DCBindingContainer dc = (DCBindingContainer)bindings;
        DCIteratorBinding iter =
            (DCIteratorBinding)dc.findIteratorBinding(iteratorName);
        ViewObject vo = iter.getViewObject();
        //create a secondary RowSetIterator to avoid disturbing row currency
        RowSetIterator rsi = vo.createRowSetIterator(null);
        //move the currency to the slot before the first row.
        rsi.reset();
        while (rsi.hasNext())
                currentRow = rsi.next();
                currentRow.setAttribute("Deleted", "N");
        rsi.closeRowSetIterator();
        iter.executeQuery();
        AdfFacesContext adfFacesContext = AdfFacesContext.getCurrentInstance();
        adfFacesContext.addPartialTarget(this.getTakenMaterialsTable());
        return null;
    }as example, if the user initially has 8 rows, then deleted 2 rows, in cancelChanges only 6 rows appears. and the deleted rows are not there??
any suggestion?

Issue in workflow - Logically Deleted Scenario

Hi,
We have an issue in Journal Entry Workflow. The user has tried to upload a document for Journal Entry Posting. Document is in Parked Status. In the workflow, when it has come to the required step of workflow approval and checks for the document type, it has got logically deleted. Thereby the workflow got terminated.
Kindly let us know as how to proceed further. What would have been the reason for this automatic logical delete? Mail us the solution regarding the same.
Regards,
Veera

Automatic Logical deletion ! I think this is not possible check with the log of the workitem whose user id is appearing then in the code check weter you have assigned any FM to locially delete the work item which is being trigered by mistake if not just start another WF and proceed with your work.
Regards
Bikas

Send step in BPM is "logical deleted"

Hi everybody,
we are sending IDOCs from BPM to receiver determination (R/3 Backend).
The BPM protocol marks the send step as "logically deleted".
In SXMB_MONI strange!
Sender = Receiver = BPM
Any ideas?
Regards Mario

The required number in the fork was wrong!

Logical Delete in DB Polling - OSB

Similar Messages

Maybe you are looking for