Differences between revisions 1 and 2
Revision 1 as of 2017-01-05 17:24:56
Size: 37
Comment:
Revision 2 as of 2017-01-05 17:25:20
Size: 2024
Comment:
Deletions are marked like this. Additions are marked like this.
Line 1: Line 1:
Describe primer_arch_patterns here. == Introduction ==
Often a task’s complexity is daunting. Often aspects of a task can be used to guide design decisions in a direction which leads to elegance, robustness, and flexibility. This can lead to more readable code and easier maintenance.

This primer is meant to inspire better design, not to guide it. Please explore patterns of interest to consider implementation. Any feedback or suggestions are welcomed.

== Some Architectural Patterns ==
=== Singleton and Registry ===
If an object needs only one instance, it’s a Singleton. Restrict to one instance if you need to coordinate actions or keep track of other classes. For Registry, use a singleton base class to keep a reference to all items.

=== Loose Coupling and Chain of Responsibility ===
Keep the steps of the process loosely coupled to keep the codebase modular. Call the next function until the process is done.

=== Proxy ===
Much like proxies in other cases, use an object to interface to another. This can be used to simplify usage of remote resources, to simplify a resource intensive object or process, or to control access to something.

=== Observer and Publish-subscribe ===
To pass information, either have a subject notify dependents upon state change directly, or use another service, like a Message Service.

=== Blackboard ===
If you have a data-heavy problem, like object or sound identification, focus on the data. When designing, consider all memory as a blackboard which reads a control component, which uses knowledge sources to operate on the blackboard memory.

=== Stateful ===
Consider the application as a relation of states, and think about how every action can be made to be a change of state. Essentially, implement a finite state machine.

== Other Design Considerations ==
=== Async ===
If a method does not need to be done in order, try to separate the steps so that the kernel can perform the task, along with others, without keeping similarly-parallel tasks waiting.

=== Mapreduce ===

Introduction

Often a task’s complexity is daunting. Often aspects of a task can be used to guide design decisions in a direction which leads to elegance, robustness, and flexibility. This can lead to more readable code and easier maintenance.

This primer is meant to inspire better design, not to guide it. Please explore patterns of interest to consider implementation. Any feedback or suggestions are welcomed.

Some Architectural Patterns

Singleton and Registry

If an object needs only one instance, it’s a Singleton. Restrict to one instance if you need to coordinate actions or keep track of other classes. For Registry, use a singleton base class to keep a reference to all items.

Loose Coupling and Chain of Responsibility

Keep the steps of the process loosely coupled to keep the codebase modular. Call the next function until the process is done.

Proxy

Much like proxies in other cases, use an object to interface to another. This can be used to simplify usage of remote resources, to simplify a resource intensive object or process, or to control access to something.

Observer and Publish-subscribe

To pass information, either have a subject notify dependents upon state change directly, or use another service, like a Message Service.

Blackboard

If you have a data-heavy problem, like object or sound identification, focus on the data. When designing, consider all memory as a blackboard which reads a control component, which uses knowledge sources to operate on the blackboard memory.

Stateful

Consider the application as a relation of states, and think about how every action can be made to be a change of state. Essentially, implement a finite state machine.

Other Design Considerations

Async

If a method does not need to be done in order, try to separate the steps so that the kernel can perform the task, along with others, without keeping similarly-parallel tasks waiting.

Mapreduce

primer_arch_patterns (last edited 2017-01-05 17:25:20 by RyanBirmingham)

Unable to edit the page? See the FrontPage for instructions.