Interface TransactionChaining
Transaction Chaining
Transaction Chaining allowsqueues
of
DSRequests to be "chained" together, such that later
DSRequests in
the queue can use the results of previous requests in the queue. This allows you to
declaratively handle various situations where information only becomes available during the
processing of a queue.
To see an example of server-driven transaction chaining used to commit a mixed transaction of
an Order and related OrderItems, see the
Master/Detail Add sample.
Transaction Chaining is only available with Power Edition licenses or better. See the Editions & Pricing page for details.
As an example of Transaction Chaining, consider an application that needs to do a
master-detail add, which involves saving a new Record representing a sales order to an
order DataSource, and also saving several related Records representing
individual items in the order to an orderItem DataSource. The Records for the
individual orderItems need to set up foreign keys referencing the primary key
assigned to the Record for the order, but the primary key of the
order record is assigned only when the Record is inserted into the database; it
cannot be known up-front.
You could resolve this programmatically - for example, you could use DMIs to store and
retrieve the PK value using servletRequest attributes - but Transaction Chaining
gives you an elegant, declarative, code-free alternative, giving you a way to declare that the
foreignKey value for the orderItem records should use the primary key value
resulting from the creation of the order record earlier in the same queue.
As another example, consider an application that allows a user to submit a free-form question which must be persisted to the database like a normal update, but which should initially show the user a list of previously-provided answers that appear to be relevant. The operation that handles the add of the question categorizes it by analyzing the text, and the category is added to the record inserted into the database, and thus to the record returned in the response. Now, via transaction chaining, a "fetch" operation later in the queue can pick up the newly assigned category and use it in criteria to fetch the list of related answers.
Transaction Chaining is implemented by specifying DSRequestModifiers in
OperationBinding.values and
OperationBinding.criteria.
These two properties
provide a general means of declaratively modifying DSRequests server-side, and transaction
chaining is only one of their uses. They can also be used, for example, to implement security
rules, by adding the currently-authorized user to the criteria of all fetch requests.
Specifically for transaction chaining, you specify criteria and/or
values entries on the operationBinding where the
value property references the
$responseData
Velocity context variable - see the "value" link for more details.
Alternatively, you can use the RPCManager APIs getFirstResponse()
and getLastResponse() to get access to the same information, either
programmatically from DMI or custom DataSource Java code, or in
JSR 223 scripts, or in Velocity expressions via
the $rpc
context variable.
Client-driven Transaction Chaining
A limited form of transaction chaining (limited for security reasons) is possible without server-side operationBinding configuration, usingfieldValueExpressions. The
primary intended use
case is a master-detail add, where the detail records require the master's primary key for
use as foreign keys, but that value is not known until the master record has been inserted.
In such a case, you create a queue of
requests, with the
add of the master record first, followed by the detail records, each of which has
fieldValueExpressions set up to use $masterId like so:
DSRequest properties = new DSRequest();
Map fve = new HashMap();
fve.put("fkField", "$masterId");
properties.setFieldValueExpressions(fve);
myDataSource.addData(record, callback, properties);
It is also possible to achieve the same thing in a less compact but more flexible way by
using the $responseData context variable (note that client-driven usages of
$responseData are limited for security reasons - see
fieldValueExpressions for
details). This approach
allows you to reference values where there is no declared foreignKey relationship, and it
allows you to reference responses other than the most recent one. For example:
DSRequest properties = new DSRequest();
Map fve = new HashMap();
fve.put("anyField", "$responseData.first.anyOtherField");
properties.setFieldValueExpressions(fve);
myDataSource.addData(record, callback, properties);
To see an example of client-driven transaction chaining used to commit a mixed transaction
of an Order and related OrderItems, see the
Master/Detail Add sample.
As with Server-Side Transaction Chaining, fieldValueExpressions can be used to
affect
dsRequest.criteria as well. For example, perhaps you are fetching the most recent
Order
and its related OrderItems, by sorting the returned Orders by orderDate and asking
for
only one record. Since you are fetching the most recent Order, you don't have the Order.orderId
value
available to fetch related OrderItems by passing the Order.orderId value as
criteria for
OrderItem.orderId. You need the primaryKey of whichever Order turns
out
to be most recent. Use fieldValueExpressions like so to solve this case:
RPCManager.startQueue();
DSRequest orderRequestProperties = new DSRequest(DSOperationType.FETCH);
orderRequestProperties.setStartRow(0);
orderRequestProperties.setEndRow(1);
orderRequestProperties.setSortBy(new SortSpecifier("orderDate", SortDirection.DESCENDING));
DataSource.get("Order").fetchData(null, null, orderRequestProperties);
DSRequest orderItemRequestProperties = new DSRequest(DSOperationType.FETCH);
orderItemRequestProperties.setFieldValueExpressions(new HashMap<String, String>() {{
put("orderId", "$responseData.first.orderId");
}});
DataSource.get("OrderItem").fetchData(null, null, orderItemRequestProperties);
RPCManager.sendQueue();
Stand-alone Application Transaction Chaining
Transaction chaining is supported when using transactions standalone. Every request within the same transaction will be eligible during the chaining. It works in the same way as it would if your application was a full blown Smart GWT application. SeeStandalone DataSource Usage for
examples on how to
do this.