INTRODUCTION

Summary

Data Access Patterns

The following data access patterns are covered:

Pattern Purpose
Data Accessor Encapsulates physical data access in a separate component, exposing only logical operations. Application code is decoupled from data access operations.
Active Domain Object Encapsulates data model and data access details within a relevant domain object. In other words, an Active Domain Object abstracts the semantics of the underlying data store (i.e., SQL Server) and the underlying data access technology (i.e., ADO.NET) and provides a simple programmatic interface for retrieving and operating on data.
Object-Relational Map Decouples active domain objects from the underlying data model and data access details. An Object-Relational Map object is responsible for mapping relational data to object-oriented concepts, allowing it to be changed independently of the application and its domain objects.
Layers Stack orthogonal application features that access data with increasing levels of abstraction.

Resource Patterns

Logically, a resource is an abstraction that simplifies the low-level complexity of working with input and output. Physically, a resource is an entity that represents storage or devices reserved for use by application. A file handle is a simple example of a resource. A file handle represents a channel that allows writing and reading from a file (the resource). 

Resources and Context

In addition to encapsulating input/output details, a resource also serves a semantic purpose by storing contextual information and enabling controlled concurrent access to the underlying storage or device. For example, a file handle stores contextual information about how the file was opened, and whether it the file was opened for reading, writing or appending. In general, resources that store contextual information save programmers from typing lots of redundant code and managing state information from one operation to the other.

Resources and Concurrency

Resources often represent data or objects that are available to multiple applications distributed across a network. For example, a database table is available to a variety of applications including reporting, management tools and client-side applications. When multiple users/tools access the same table in incompatible ways, unpredictable errors occur. Resources play an important role in robust concurrency solutions. Resources offer some level of synchronization that restricts concurrent access. 

Resources and Data Access

The following general-purpose low-level resources provide unstructured data access and are useful in a range of applications:

Resources and Management

Resources usually consume significant amount of storage as long as they remain open. For example, a database connection requires client memory to store contextual information and it allocates server memory to maintain server-side context. In addition, a database connection keeps a socket open on both sides to enable fast communication between both sockets. Resources also implement synchronization to restrict concurrent access to one or more objects.

Since open resources consume memory and reduce concurrency, it is important to understand how applications use and manage resources. This list describes a few ideas that apply to most scenarios:

This and the following resource patterns define a common design strategy for managing resources at the application or middleware level.

Patterns

The following resource patterns are covered:

Pattern Purpose
Resource Pool A Resource Pool recycles resources to minimize resource initialization overhead. A resource pool manages resources efficiently while allowing application code to freely allocate them.
Resource Decorator A resource decorator dynamically attaches behavior to an existing resource with minimal disruption to application code. A resource decorator extends a resource's functionality without sub-classing or changing functionality.
Resource Timer Automatically releases inactive resources. This pattern solves the problem of resources being allocated indefinitely.
Resource Descriptor Isolates platform- and data-source-dependent behavior within a single component. A resource descriptor exposes platform and datasource specifics as generic logical operations. This allows the majority of data access code to remain independent of its physical environment.
Resource Retriever Automatically retries operations whose failure is expected under certain defined conditions. This pattern enables fault-tolerance for data access operations.

Input-Output Patterns

Domain objects directly model application or business concepts rather than relational database entities, and enable you to decouple the physical data model and data access details from the application logic. When you design domain objects, you must also design their domain object mapping. A domain object mapping describes the translation between domain objects and corresponding relational data. For example, in the Active Domain Object pattern, each object is responsible for defining and encapsulating its own mapping.

Input-output patterns are used to define and implement database input and output operations. Database input and output are a primary function of domain object mapping:

Input / Output Operations

Here are some example of input and output operations expressed in term of domain objects:

Identity Objects

In addition to translating data between objects and tables, another important factor for domain object mapping is the issue of identity objects. When an application invokes input/output operations using domain objects, it must identify target data. 

To illustrate the issue of identity objects, assume that table [Product] contains all products available for sale. For example, consider these two scenarios for looking up information:

Scenario SQL
Look up a product by category select * from Product where category = 'Baby Food'
Look up a product's price select price from Product where ProductCode = 1234

These expressions should never appear in application code that otherwise uses domain objects. One reason is that it explicitly mentions the names of data entity models such as Product and ProductCode, and second it includes specific SQL syntax that may change if you decide to move your database to another platform.

Identity objects solve this problem by using domain concepts to identify the target relational data. An identity object identifies a domain object, just as a set of primary key values uniquely identifies a row of table data. In fact, it is common for identity objects to correspond to a table's primary key.

In the [Product] table, the primary key is likely to be the unique ProductCode column. The analogous identity object is simple a string representation of the ProductCode value. The application code can then find any product using its ProductCode like this:

Product product = ProductInventory.Find( 1234 );

Identity objects do not always correspond directly to a table's primary key, especially in cases where applications may search on columns other than those included in the primary key. An alternative identity object could define multiple attributes that correspond to search criteria. In this respect, a single identity object does not uniquely define a single domain object, but rather a set of domain objects that matches its criteria. For example, if you want to find all products whose category is vegetable and price is < 1, you would designate this information using a ProductCriteria object:

ProductCriteria criteria = new ProductCriteria();
criteria.Category = Categories.Vegetable;
criteria.Price = " < 1.00";
Product[] products = ProductInventory.Find( criteria );

While the above examples using Identity objects to query and read target data, identity objects are also important for output operations since they identify specific database rows to update/insert/delete. For example, the following code contains code to read and update an identity object:

Product product = ProductInventory.Find( 1234 );
product.Price = product.Price * 1.1;
ProductInventory.Update( product )

Note that the application's code does not explicitly indicate an identity object when it calls the Update operation. The Update operation's implementation implicitly extracts the product's ID to issue the appropriate SQL UPDATE statement.

Patterns

The following resource patterns are covered:

Pattern Purpose
Selection Factory Generates query selections based on identity object attributes
Domain Object Factory Populates domain objects based on query result data.
Update Factory Generates update selections based on modified domain object attributes.
Domain Object Assembler Populates, persists, and deletes domain objects using uniform factory framework.

These patterns are often combined to build a robust domain object mapping framework that can decouple generic mapping logic from the customized conversion details for specific types. This separation allows you to introduce additional domain objects as your application requires them. It also allows you to apply common optimizations and enhancements that apply immediately to operations on all domain objects.

Cache Patterns

Cache patterns define strategies for integrating caching into your applications and middleware components. These patterns concentrate on improving data access performance and resource utilizations by eliminating redundant data access operations. Data access operations are a common source of bottlenecks as they consume a significant portion of a system's memory. While recycling database resources and using indices goes a long way to achieve this, one of the most effective strategies is to eliminate redundant data access operations altogether. Caching enables applications to avoid issues multiple database read operations for the same data item. Caches usually reside in memory and enable fast access to their components. Applications do not need to issue subsequent database operations to access the cached data.

Cache Operations and Transparency

A cache starts empty, and at some point during application startup or initialization, applications or middleware components read data from the database to store in the cache. Strategies for populating caches vary, ranging from simple copying of entire ADO.NET DataSets to using strategic and selective decisions to populate only the most accessed data. Application code is the ultimate consumer of cached data. It is common for applications to access cached data using primary keys or identity objects, but some applications may require other semantics such as statement handles or a query language.

The semantics that cache operations define help achieve cache transparency. Cache transparency refers to the visibility of a cache to applications and middleware code. Consider the cases of non-transparent and transparent caches:

Cached Data

With caches you also need to consider the form of the cached data. You can store data in its physical database format using software representations of tables, rows, columns, and relationships . In ADO.NET this would correspond to using DataSets and DataTables. Another option would be to convert cached data into the domain form that your application expects, in other words, the cache stores domain objects rather than raw data.

Patterns

The first group of cache patterns describe strategies for integrating cache storage and retrieval operations in applications and middleware components with various degrees of transparency. These patterns address how to utilize caches rather than how to implement caches directly:

Pattern Purpose
Cache Accessor Decouples caching logic from the data model and the data access details
Demand Cache Populates a cache on-demand as applications request data. This is useful for data that is read frequently but unpredictably.
Primed Cache Populates a cache with a predicted set of data. This is useful for data that is read frequently and  predictably.
Cache Search Sequence Inserts short cuts into a cache to optimize the number of operations that future searches require.

The second group of cache patterns describe strategies for efficient caching implementations

Pattern Purpose
Cache Collector The equivalent of a .NET Garbage Collector - purges unneeded entries.
Cache Replicator Replicates operations across multiple caches.
Cache Statistics Record/Publish cache and pool statistics.

These caching patterns are independent of each others and can be mixed and matched to build a comprehensive caching solution.