330 likes | 494 Views
Chapter 38. Persistence Framework with Patterns. Persistent Objects. Object databases Relational databases Other: Flat files, XML, hierarchical databases, etc. Persistence Service.
E N D
Chapter 38 Persistence Framework with Patterns CS6359 Fall 2011 John Cole
Persistent Objects • Object databases • Relational databases • Other: Flat files, XML, hierarchical databases, etc. CS6359 Fall 2011 John Cole
Persistence Service • A Persistence Framework is a general-purpose, reusable, and extendable set of types that provides functionality to support persistent objects. • Also called an O-R Mapping service if it works with an RDB. CS6359 Fall 2012 John Cole
Frameworks • A Framework is an extensible set of objects for related functions. For example Swing, or LINQ. • Provides an implementation of core and unvarying functions and allows a programmer to plug in the varying functions or to extend them. CS6359 Fall 2011 John Cole
Frameworks • Is a cohesive set of interfaces and classes to provide services for the core, unvarying part of a logical subsystem • Contain concrete and abstract classes that define interfaces to conform to, object interactions to participate in, and other invariants • Usually require the user to define subclasses of framework class to use, customize, and extend framework services CS6359 Fall 2011 John Cole
Frameworks • Have abstract classes that may contain both abstract and concrete methods • Rely on the Hollywood Principle: Don’t call us, we’ll call you. User-defined classes will receive messages from predefined framework classes • Frameworks are reusable CS6359 Fall 2011 John Cole
Persistence Service and Framework • Requirements: • Store and retrieve objects in a persistent storage mechanism • Commit and rollback transactions • Extensible to support different storage mechanisms CS6359 Fall 2011 John Cole
Key Ideas • Mapping between a class and its persistent store and between object attributes and the fields in a record • Object identity – to easily relate records to objects and ensure there are no inappropriate duplicates, records and objects have a unique identifier • Database mapper – pure fabrication object responsible for materialization and dematerialization CS6359 Fall 2011 John Cole
Key Ideas • Caches – persistence services cache materialized for performance • Transaction state of object – State of objects in relation to the current transaction. Which objects are dirty? • Transaction operations – Commit and Rollback CS6359 Fall 2011 John Cole
Key Ideas • Lazy materialization – not all objects are materialized at once; materialization on demand • Virtual proxies – Lazy materialization can be implemented using a smart reference known as a virtual proxy CS6359 Fall 2011 John Cole
Pattern: Representing Objects as Tables • Define a table in the RDB for each persistent object class • Attributes containing primitive data types map to columns • Relational model requires that values be atomic (first normal form) CS6359 Fall 2011 John Cole
Pattern: Object Identifier • Since it is desirable to have a consistent way to relate objects to records and no duplicate rows, use an Object Identifier for each record. • Every table will have an OID as its key field and every record in the table will have a unique value for this key • OID provides a consistent key type CS6359 Fall 2011 John Cole
Accessing Persistence Service with a Facade • Define a façade for the various services • Retrieve an object given its OID; system needs to know what kind of object, so provide the class type, too CS6359 Fall 2011 John Cole
Mapping Objects • The Persistence Façade delegates the work • If a persistent object (such as ProductDescription) persists itself, this is a direct mapping design • This works if a compiler tool generates it • Problems: strong coupling of the object to storage; complex responsibilities in an unrelated area. CS6359 Fall 2011 John Cole
Database Broker • Class that is responsible for materialization, dematerialization, and object caching • Pattern is also called Database Mapper • Define a different mapper class for each persistent object CS6359 Fall 2011 John Cole
Mapper Factory • Better not to name each mapper with a different operation. For example: class MapperFactory { public IMapper getProductDescriptionMapper() {...} public IMapper getSaleMapper() {...} ... } CS6359 Fall 2011 John Cole
Mapper Factory • Better: class PersistenceFacade { private java.util.Map mappers = MapperFactory.getInstance().getAllMappers(); ... } CS6359 Fall 2011 John Cole
Metadata-based Mappers • Instead of creating individual mapper classes for persistent types, these generate the mapping using metadata that describes the mapping • Thus you can change the schema mapping in an external store and it will be reflected in the running system • Pattern is Protected Variations WRT schema variations CS6359 Fall 2011 John Cole
Template Method Pattern • Define a method (the Template Method) in a superclass that defines the skeleton of an algorithm. This invokes other methods, some of which may be overridden by a subclass • Subclasses can override varying methods to provide unique behavior at points of variability CS6359 Fall 2011 John Cole
Template Pattern • The basic algorithm structure is: • If the object is in cache, return it • If not, create it from its representation in storage, save it in cache, and return it • The point of variation is how the object is created from storage CS6359 Fall 2011 John Cole
Template Pattern • Create the get method as the template in an abstract superclass AbstractPersistenceMapper that defines the template, and use a hook method in subclasses for the varying part. • The get method contains critical section code that must be made thread-safe CS6359 Fall 2011 John Cole
Synchronized Methods in UML • +get(OID): Object {guarded} CS6359 Fall 2011 John Cole
Transactional States and the State Pattern • Persistent objects can be inserted, deleted, or modified; • Operating on a persistent object does not cause an immediate database update; an explicit commit operation must be done • The response to an operation depends upon the transactional state of the object CS6359 Fall 2011 John Cole
Persistent Object Statechart CS6359 Fall 2011 John Cole
PersistentObject • Implements the following methods: • Commit() • Delete() • Rollback() • Save() • Extending a domain class with a technical services class mixes architectural concerns CS6359 Fall 2011 John Cole
State Pattern • Commit and Rollback have different logic but similar structure • Problem: If an object’s behavior is dependent upon its state, is there an alternative to conditional logic? • Solution: Create state classes for each state, implementing a common interface. Delegate state-dependent operations from the context object to its current state object. Ensure the context object always points to a state object reflecting its current state. CS6359 Fall 2011 John Cole
Transaction and the Command Pattern • Transaction is a set of tasks that either must all be completed or none completed. Thus it is atomic • The order of database tasks within a transaction can influence its success and performance CS6359 Fall 2011 John Cole
Transaction examples • Suppose the database has a referential integrity constraint such that when a record is updated in table A that contains a foreign key reference to table B, the record in B must exist • Suppose a transaction contains an INSERT task to add the B record and an update for A. If the update occurs before the insert, error • Ordering the tasks helps CS6359 Fall 2011 John Cole
Command Pattern • Problem: How to handle requests or tasks that need functions such as sorting, queuing, delaying, logging, or undoing? • Solution: Make each task a class that implements a common interface CS6359 Fall 2011 John Cole
Command Pattern • Actions become objects and can be sorted, logged, etc. • Each task or action is an object with an Execute method • GUI actions such as cut and paste are examples of Command. The CutCommand’s execute method does a cut and its undo method reverses the cut. These objects can be stacked CS6359 Fall 2011 John Cole
Lazy Materialization with a Virtual Proxy • Sometimes useful not to materialize an object until absolutely necessary. This may be because the object is rarely referenced. • Example: Manufacturer object referenced by ProductDescription • Virtual Proxy gets the real object on first reference, thus standing in for the object. • Created by the object that references the other object CS6359 Fall 2011 John Cole
Representing Relationships • One-to-one: place an OID foreign key in one or both tables or create an associative table with the OIDs of each object in the relationship. • One-to-many: Create an associative table with keys of both objects • Many-to-many: Create an associative table with keys of both objects CS6359 Fall 2011 John Cole
Separation of Concerns • Weakness of creating a PersistentObject class but it couples domain classes to technical services classes • Separation is not an absolute rule, but be careful how you break it CS6359 Fall 2011 John Cole