A question from Mark Cade case study

[Srinivas Bitla]

Hi All,

Can some one elaborate the following statement from Mark Cade case study, third paragraph page no. 170?

"Because the Shipping, Payment and Accounting systems have Java technology APIs and the data is not persisted, you have direct access from the OrderProcessor as opposed to encapsulating the requests in a DAO."

I couldn't understand that statement completly. I am not sure what does he mean by "the data is not persisted". Does he mean that the data is not persisted in the system being developed?

In the following situations, when can I go for a DAO and when can I use a Processor to talk to the subsystem?

1) The subsystem has a Java technology APIs and the data is not persisted?(Cade says we should go for Processor)

2) The subsystem has a Java technology APIs and the data is persisted?

3) The subsystem does not have a Java technology APIs and the data is not persisted?

4) The subsystem does not have a Java technology APIs and the data is persisted?

Thanks in advance,
Srinivas.

[Srinivas Bitla]

Can some one answer this question?

Thanks in advance,
Srinivas.

[Ramon Gill]

Srinivas, I'll try to help.

If you look at pages 166, 167, you'll see that the new system consists of new packages Order, Customer and Catalog. Data for these packages is persisted in the new system (i.e. DAO's required in component diagrams).

These new packages have dependencies on some existing packages (Inventory, Accounting, Shipping, Payment and ContentMgmt). These existing packages persist their own data (i.e. not part of the new system). Therefore in component diagrams interfaces are shown as the way to access data in these systems.

Hope this helps,
Ray

[Ajith Kallambella]

I couldn't understand that statement completly. I am not sure what does he mean by "the data is not persisted". Does he mean that the data is not persisted in the system being developed?

In the following situations, when can I go for a DAO and when can I use a Processor to talk to the subsystem?

Authors are of the opinion that an additional DAO layer may not be necessary and hence direct data access constitutes a good design when the mode of access is 'read-only'.

Personally I beg to differ. DAOs are not about persistence, but about encapsulating data access. Hence the mode of data access( read-only versus read-write) should not be the criteria driving your decision to use( or not ) DAOs.

Well, anyways...