openmct/docs/src/guide/index.md

# Open MCT Web Developer Guide
Victor Woeltjen

[victor.woeltjen@nasa.gov](mailto:victor.woeltjen@nasa.gov)

September 23, 2015  
Document Version 1.1  

Date                | Version   | Summary of Changes      | Author
------------------- | --------- | ----------------------- | ---------------
April 29, 2015      | 0         | Initial Draft           | Victor Woeltjen
May 12, 2015        | 0.1       |                         | Victor Woeltjen
June 4, 2015        | 1.0       | Name Changes            | Victor Woeltjen
October 4, 2015     | 1.1       | Conversion to MarkDown  | Andrew Henry

# Introduction
The purpose of this guide is to familiarize software developers with the Open
MCT Web platform.

## What is Open MCT Web
Open MCT Web is a platform for building user interface and display tools,
developed at the NASA Ames Research Center in collaboration with teams at the
Jet Propulsion Laboratory. It is written in HTML5, CSS3, and JavaScript, using 
[AngularJS](http://www.angularjs.org) as a framework. Its intended use is to 
create single-page web applications which integrate data and behavior from a 
variety of sources and domains.

Open MCT Web has been developed to support the remote operation of space
vehicles, so some of its features are specific to that task; however, it is
flexible enough to be adapted to a variety of other application domains where a
display tool oriented toward browsing, composing, and visualizing would be
useful.

Open MCT Web provides:

*  A common user interface paradigm which can be applied to a variety of domains
and tasks. Open MCT Web is more than a widget toolkit - it provides a standard
tree-on-the-left, view-on-the-right browsing environment which you customize by
adding new browsable object types, visualizations, and back-end adapters.  
*  A plugin framework and an extensible API for introducing new application
features of a variety of types.  
*  A set of general-purpose object types and visualizations, as well as some
visualizations and infrastructure specific to telemetry display.

## Client-Server Relationship
Open MCT Web is client software - it runs entirely in the user's web browser. As
such, it is largely 'server agnostic'; any web server capable of serving files
from paths is capable of providing Open MCT Web.

While Open MCT Web can be configured to run as a standalone client, this is
rarely very useful. Instead, it is intended to be used as a display and
interaction layer for information obtained from a variety of back-end services.
Doing so requires authoring or utilizing adapter plugins which allow Open MCT
Web to interact with these services.

Typically, the pattern here is to provide a known interface that Open MCT Web
can utilize, and implement it such that it interacts with whatever back-end
provides the relevant information. Examples of back-ends that can be utilized in
this fashion include databases for the persistence of user-created objects, or
sources of telemetry data.

See the [Architecture Guide](../architecture/index.md#Overview) for information
on the client-server relationship.

## Developing with Open MCT Web
Building applications with Open MCT Web typically means authoring and utilizing
a set of plugins which provide application-specific details about how Open MCT
Web should behave.

### Technologies

Open MCT Web sources are written in JavaScript, with a number of configuration
files written in JSON. Displayable components are written in HTML5 and CSS3.  
Open MCT Web is built using [AngularJS](http://www.angularjs.org)  from Google. A
good understanding of Angular is recommended for developers working with Open
MCT Web.

### Forking
Open MCT Web does not currently have a single stand-alone artifact that can be
used as a library. Instead, the recommended approach for creating a new
application is to start by forking/branching Open MCT Web, and then adding new
features from there. Put another way, Open MCT Web's source structure is built
to serve as a template for specific applications.

Forking in this manner should not require that you edit Open MCT Web's sources.
The preferred approach is to create a new directory (peer to `index.html`) for
the new application, then add new bundles (as described in the Framework
chapter) within that directory.

To initially clone the Open MCT Web repository:
`git clone <repository URL> <local repo directory> -b open-master`

To create a fork to begin working on a new application using Open MCT Web:  

    cd <local repo directory>  
    git checkout open-master
    git checkout -b <new branch name>

As a convention used internally, applications built using Open MCT Web have
master branch names with an identifying prefix. For instance, if building an
application called 'Foo', the last statement above would look like:

    git checkout -b foo-master

This convention is not enforced or understood by Open MCT Web in any way; it is
mentioned here as a more general recommendation.

# Overview

Open MCT Web is implemented as a framework component which manages a set of
other components. These components, called _bundles_, act as containers to group
sets of related functionality; individual units of functionality are expressed
within these bundles as _extensions_.

Extensions declare dependencies on other extensions (either individually or
categorically), and the framework provides actual extension instances at
run-time to satisfy these declared dependency. This dependency injection
approach allows software components which have been authored separately (e.g. as
plugins) but to collaborate at run-time.

Open MCT Web's framework layer is implemented on top of AngularJS's [dependency
injection mechanism](https://docs.angularjs.org/guide/di)  and is modelled after
[OSGi](hhttp://www.osgi.org/)  and its [Declarative Services component model](http://wiki.osgi.org/wiki/Declarative_Services). 
In particular, this is where the term _bundle_ comes from.

## Framework Overview

The framework's role in the application is to manage connections between
bundles. All application-specific behavior is provided by individual bundles, or
as the result of their collaboration.

The framework is described in more detail in the [Framework Overview](../architecture/Framework.md#Overview) of the 
architecture guide.

### Tiers
While all bundles in a running Open MCT Web instance are effectively peers, it
is useful to think of them as a tiered architecture, where each tier adds more
specificity to the application.
```nomnoml
#direction: down
[Plugins (Features external to OpenMCTWeb) *Bundle]->[<frame>OpenMCTWeb |
[Application (Plots, layouts, ElasticSearch wrapper) *Bundle]->[Platform (Core API, common UI, infrastructure) *Bundle]
[Platform (Core API, common UI, infrastructure) *Bundle]->[Framework (RequireJS, AngularJS, bundle loader)]]
```

* __Framework__ : This tier is responsible for wiring together the set of
configured components (called _bundles_) together to instantiate the running
application. It is responsible for mediating between AngularJS (in particular,
its dependency injection mechanism) and RequireJS (to load scripts at run-time.)
It additionally interprets bundle definitions (see explanation below, as well as
further detail in the Framework chapter.) At this tier, we are at our most
general: We know only that we are a plugin-based application.
* __Platform__: Components in the Platform tier describe both the general user
interface and corresponding developer-facing interfaces of Open MCT Web. This
tier provides the general infrastructure for applications. It is less general
than the framework tier, insofar as this tier introduces a specific user
interface paradigm, but it is still non-specific as to what useful features
will be provided. Although they can be removed or replaced easily, bundles
provided by the Platform tier generally should not be thought of as optional.
* __Application__: The application tier consists of components which utilize the
infrastructure provided by the Platform to provide functionality which will (or
could) be useful to specific applications built using Open MCT Web. These
include adapters to specific persistence back-ends (such as ElasticSearch or
CouchDB) as well as bundles which describe more user-facing features (such as
_Plot_ views for visualizing time series data, or _Layout_ objects for
display-building.) Bundles from this tier can be added or removed without
compromising basic application functionality, with the caveat that at least one
persistence adapter needs to be present.
* __Plugins__: Conceptually, this tier is not so different from the application
tier; it consists of bundles describing new features, back-end adapters, that
are specific to the application being built on Open MCT Web. It is described as
a separate tier here because it has one important distinction from the
application tier: It consists of bundles that are not included with the platform
(either authored anew for the specific application, or obtained from elsewhere.)

Note that bundles in any tier can go off and consult back-end services. In
practice, this responsibility is handled at the Application and/or Plugin tiers;
Open MCT Web is built to be server-agnostic, so any back-end is considered an
application-specific detail.

## Platform Overview

The "tiered" architecture described in the preceding text describes a way of
thinking of and categorizing software components of a Open MCT Web application,
as well as the framework layer's role in mediating between these components.
Once the framework layer has wired these software components together, however,
the application's logical architecture emerges.

An overview of the logical architecture of the platform is given in the [Platform Architecture](../architecture/Platform.md#PlatformArchitecture) 
section of the Platform guide

### Web Services

As mentioned in the Introduction, Open MCT Web is a platform single-page 
applications which runs entirely in the browser. Most applications will want to 
additionally interact with server-side resources, to (for example) read 
telemetry data or store user-created objects. This interaction is handled by 
individual bundles using APIs which are supported in browser (such as 
`XMLHttpRequest`, typically wrapped by Angular's `$http`.)

```nomnoml
#direction: right
[Web Service #1] <- [Web Browser]
[Web Service #2] <- [Web Browser]
[Web Service #3] <- [Web Browser]
[<package> Web Browser |
  [<package> Open MCT Web |
    [Plugin Bundle #1]-->[Core API]
    [Core API]<--[Plugin Bundle #2]
    [Platform Bundle #1]-->[Core API]
    [Platform Bundle #2]-->[Core API]
    [Platform Bundle #3]-->[Core API]
    [Core API]<--[Platform Bundle #4]
    [Core API]<--[Platform Bundle #5]
    [Core API]<--[Plugin Bundle #3]
  ]
  [Open MCT Web] ->[Browser APIs]
]
```

This architectural approach ensures a loose coupling between applications built 
using Open MCT Web and the backends which support them. 
 
### Glossary 
 
Certain terms are used throughout Open MCT Web with consistent meanings or 
conventions. Other developer documentation, particularly in-line documentation, 
may presume an understanding of these terms.

* __bundle__: A bundle is a removable, reusable grouping of software elements. 
The application is composed of bundles. Plug-ins are bundles.
* __capability__: A JavaScript object which exposes dynamic behavior or 
non-persistent state associated with a domain object.
* __category__: A machine-readable identifier for a group that something may 
belong to.
* __composition __: In the context of a domain object, this refers to the set of 
other domain objects that compose or are contained by that object. A domain 
object's composition is the set of domain objects that should appear immediately
 beneath it in a tree hierarchy. A domain object's composition is described in 
its model as an array of identifiers; its composition capability provides a 
means to retrieve the actual domain object instances associated with these 
identifiers asynchronously. 
* __description__: When used as an object property, this refers to the human-
readable description of a thing; usually a single sentence or short paragraph. 
(Most often used in the context of extensions, domain object models, or other 
similar application-specific objects.) 
* __domain object __: A meaningful object to the user; a distinct thing in the 
work support by Open MCT Web. Anything that appears in the left-hand tree is a 
domain object. 
* __extension __: An extension is a unit of functionality exposed to the platform 
in a declarative fashion by a bundle. The term 'extension category' is used to 
distinguish types of extensions from specific extension instances. 
* __id__: A string which uniquely identifies a domain object. 
* __key__: When used as an object property, this refers to the machine-readable 
identifier for a specific thing in a set of things. (Most often used in the 
context of extensions or other similar application-specific object sets.) This 
term is chosen to avoid attaching ambiguous meanings to 'id'. 
* __model__: The persistent state associated with a domain object. A domain 
object's model is a JavaScript object which can be converted to JSON without 
losing information (that is, it contains no methods.) 
* __name__: When used as an object property, this refers to the human-readable 
name for a thing. (Most often used in the context of extensions, domain object 
models, or other similar application-specific objects.) 
* __navigation__: Refers to the current state of the application with respect to 
the user's expressed interest in a specific domain object; e.g. when a user 
clicks on a domain object in the tree, they are navigating to it, and it is 
thereafter considered the navigated object (until the user makes another such 
choice.) This term is used to distinguish navigation from selection, which 
occurs in an editing context. 
* __space__: A machine-readable name used to identify a persistence store. 
Interactions with persistence with generally involve a space parameter in some 
form, to distinguish multiple persistence stores from one another (for cases 
where there are multiple valid persistence locations available.) 
* __source__: A machine-readable name used to identify a source of telemetry 
data. Similar to "space", this allows multiple telemetry sources to operate 
side-by-side without conflicting. 

# Framework
  
Open MCT Web is built on the [AngularJS framework]( http://www.angularjs.org ). A 
good understanding of that framework is recommended. 

Open MCT Web adds an extra layer on top of AngularJS to (a) generalize its 
dependency injection mechanism slightly, particularly to handle many-to-one 
relationships; and (b) handle script loading. Combined, these features become a 
plugin mechanism. 
  
This framework layer operates on two key concepts: 

* __Bundle:__  A bundle is a collection of related functionality that can be 
added to the application as a group. More concretely, a bundle is a directory 
containing a JSON file declaring its contents, as well as JavaScript sources, 
HTML templates, and other resources used to support that functionality. (The 
term bundle is borrowed from [OSGi](http://www.osgi.org/)  - which has also 
inspired many of the concepts used in the framework layer. A familiarity with 
OSGi, particularly Declarative Services, may be useful when working with Open 
MCT Web.)
* __Extension:__  An extension is an individual unit of functionality. Extensions 
are collected together in bundles, and may interact with other extensions. 

The framework layer, loaded and initiated from `index.html`, is the main point 
of entry for an application built on Open MCT Web. It is responsible for wiring 
together the application at run time (much of this responsibility is actually 
delegated to Angular); at a high-level, the framework does this by proceeding 
through four stages:

1. __Loading definitions:__ JSON declarations are loaded for all bundles which 
will constitute the application, and wrapped in a useful API for subsequent 
stages. 
2. __Resolving extensions:__ Any scripts which provide implementations for 
extensions exposed by bundles are loaded, using Require. 
3. __Registering extensions__ Resolved extensions are registered with Angular, 
such that they can be used by the application at run-time. This stage includes 
both registration of Angular built-ins (directives, controllers, routes, 
constants, and services) as well as registration of non-Angular extensions. 
4. __Bootstrapping__ The Angular application is bootstrapped; at that point, 
Angular takes over and populates the body of the page using the extensions that 
have been registered. 

## Bundles 

The basic configurable unit of Open MCT Web is the _bundle_. This term has been 
used a bit already; now we'll get to a more formal definition. 

A bundle is a directory which contains:

* A bundle definition; a file named `bundle.json`.
* Subdirectories for sources, resources, and tests. 
* Optionally, a `README.md` Markdown file describing its contents (this is not 
used by Open MCT Web in any way, but it's a helpful convention to follow.)

The bundle definition is the main point of entry for the bundle. The framework 
looks at this to determine which components need to be loaded and how they 
interact.

A plugin in Open MCT Web is a bundle. The platform itself is also decomposed 
into bundles, each of which provides some category of functionality. The 
difference between a _bundle_ and a _plugin_ is purely a matter of the intended 
use; a plugin is just a bundle that is meant to be easily added or removed. When 
developing, it is typically more useful to think in terms of bundles. 
  
### Configuring Active Bundles 
 
To decide which bundles should be loaded, the framework loads a file named 
`bundles.json` (peer to the `index.html` file which serves the application) to 
determine which bundles should be loaded. This file should contain a single JSON 
array of strings, where each is the path to a bundle. These paths should not 
include bundle.json  (this is implicit) or a trailing slash. 

For instance, if `bundles.json` contained: 

    [ 
        "example/builtins", 
        "example/extensions" 
    ] 
   
...then the Open MCT Web framework would look for bundle definitions at 
`example/builtins/bundle.json` and `example/extensions/bundle.json`, relative 
to the path of `index.html`. No other bundles would be loaded.  

### Bundle Definition 
 
A bundle definition (the `bundle.json` file located within a bundle) contains a 
description of the bundle itself, as well as the information exposed by the 
bundle. 
 
This definition is expressed as a single JSON object with the following 
properties (all of which are optional, falling back to reasonable defaults):

* `key`: A machine-readable name for the bundle. (Currently used only in 
logging.) 
* `name`: A human-readable name for the bundle. (Also only used in logging.) 
* `sources`: Names a directory in which source scripts (which will implement 
extensions) are located. Defaults to 'src' 
* `resources`: Names a directory in which resource files (such as HTML templates, 
images, CS files, and other non-JavaScript files needed by this bundle) are 
located. Defaults to 'res'  
* `libraries`: Names a directory in which third-party libraries are located. 
Defaults to 'lib' 
* `configuration`: A bundle's configuration object, which should be formatted as 
would be passed to require.config (see [RequireJS documentation](http://requirejs.org/docs/api.html ) ); 
note that only paths and shim have been tested. 
* `extensions`: An object containing key-value pairs, where keys are extension 
categories, and values are extension definitions. See the section on Extensions 
for more information.  

For example, the bundle definition for example/policy looks like:  

    {
        "name": "Example Policy", 
        "description": "Provides an example of using policies.", 
        "sources": "src", 
        "extensions": { 
            "policies": [ 
                { 
                    "implementation": "ExamplePolicy.js", 
                    "category": "action" 
                }
            ] 
        } 
    }

### Bundle Directory Structure 
 
In addition to the directories defined in the bundle definition, a bundle will 
typically contain other directories not used at run-time. Additionally, some 
useful development scripts (such as the command line build and the test suite) 
expect this directory structure to be in use, and may ignore options chosen by 
`b undle.json`. It is recommended that the directory structure described below be 
used for new bundles.

* `src`: Contains JavaScript sources for this bundle. May contain additional 
subdirectories to organize these sources; typically, these subdirectories are 
named to correspond to the extension categories they contain and/or support, but 
this is only a convention. 
* `res`: Contains other files needed by this bundle, such as HTML templates. May 
contain additional subdirectories to organize these sources. 
* `lib`: Contains JavaScript sources from third-party libraries. These are 
separated from bundle sources in order to ignore them during code style checking 
from the command line build.
* `test`: Contains JavaScript sources implementing [Jasmine](http://jasmine.github.io/) 
tests, as well as a file named `suite.json` describing which files to test. 
Should have the same folder structure as the `src` directory; see the section on 
automated testing for more information. 
 
For example, the directory structure for bundle `platform/commonUI/about`  looks 
like: 

    Platform
      |
      |-commonUI
      |
      +-about
        |
        |-res
        |
        |-src
        |
        |-test
        |
        |-bundle.json
        |
        +-README.md

## Extensions 

While bundles provide groupings of related behaviors, the individual units of 
behavior are called extensions. 

Extensions belong to categories; an extension category is the machine-readable 
identifier used to identify groups of extensions. In the `extensions` property 
of a bundle definition, the keys are extension categories and the values are 
arrays of extension definitions. 
 
### General Extensions 

Extensions are intended as a general-purpose mechanism for adding new types of 
functionality to Open MCT Web. 

An extension category is registered with Angular under the name of the 
extension, plus a suffix of two square brackets; so, an Angular service (or, 
generally, any other extension) can access the full set of registered 
extensions, from all bundles, by including this string (e.g. `types[]` to get 
all type definitions) in a dependency declaration. 

As a convention, extension categories are given single-word, plural nouns for 
names within Open MCT Web (e.g. `types`.) This convention is not enforced by the 
platform in any way. For extension categories introduced by external plugins, it 
is recommended to prefix the extension category with a vendor identifier (or 
similar) followed by a dot, to avoid collisions. 
 
### Extension Definitions 

The properties used in extension definitions are typically unique to each 
category of extension; a few properties have standard interpretations by the 
platform. 

* `implementation`: Identifies a JavaScript source file (in the sources 
folder) which implements this extension. This JavaScript file is expected to 
contain an AMD module (see http://requirejs.org/docs/whyamd.html#amd ) which 
gives as its result a single constructor function. 
* `depends`: An array of dependencies needed by this extension; these will be 
passed on to Angular's [dependency injector](https://docs.angularjs.org/guide/di ) . 
By default, this is treated as an empty array. Note that depends does not make 
sense without `implementation` (since these dependencies will be passed to the 
implementation when it is instantiated.) 
* `priority`: A number or string indicating the priority order (see below) of 
this extension instance. Before an extension category is registered with 
AngularJS, the extensions of this category from all bundles will be concatenated 
into a single array, and then sorted by priority. 

Extensions do not need to have an implementation. If no implementation is 
provided, consumers of the extension category will receive the extension 
definition as a plain JavaScript object. Otherwise, they will receive the 
partialized (see below) constructor for that implementation, which will 
additionally have all properties from the extension definition attached. 

#### Partial Construction 

In general, extensions are intended to be implemented as constructor functions, 
which will be used elsewhere to instantiate new objects of that type. However, 
the Angular-supported method for dependency injection is (effectively) 
constructor-style injection; so, both declared dependencies and run-time 
arguments are competing for space in a constructor's arguments. 

To resolve this, the Open MCT Web framework registers extension instances in a 
partially constructed form. That is, the constructor exposed by the extension's 
implementation is effectively decomposed into two calls; the first takes the 
dependencies, and returns the constructor in its second form, which takes the 
remaining arguments. 

This means that, when writing implementations, the constructor function should 
be written to include all declared dependencies, followed by all run-time 
arguments. When using extensions, only the run-time arguments need to be 
provided. 
 
#### Priority 

Within each extension category, registration occurs in priority order. An 
extension's priority may be specified as a `priority` property in its extension 
definition; this may be a number, or a symbolic string. Extensions are 
registered in reverse order (highest-priority first), and symbolic strings are 
mapped to the numeric values as follows: 

* `fallback`: Negative infinity. Used for extensions that are not intended for 
use (that is, they are meant to be overridden) but are present as an option of 
last resort. 
* `default`: `-100`. Used for extensions that are expected to be overridden, but 
need a useful default. 
* `none`: `0`. Also used if no priority is specified, or if an unknown or 
malformed priority is specified. 
* `optional`: `100`. Used for extensions that are meant to be used, but may be 
overridden. 
* `preferred`: `1000`. Used for extensions that are specifically intended to be 
used, but still may be overridden in principle. 
* `mandatory`: Positive infinity. Used when an extension should definitely not 
be overridden. 

These symbolic names are chosen to support usage where many extensions may 
satisfy a given need, but only one may be used; in this case, as a convention it 
should be the lowest-ordered (highest-priority) extensions available. In other 
cases, a full set (or multi-element subset) of extensions may be desired, with a 
specific ordering; in these cases, it is preferable to specify priority 
numerically when declaring extensions, and to understand that extensions will be 
sorted according to these conventions when using them. 
   
### Angular Built-ins 

Several entities supported Angular are expressed and managed as extensions in 
Open MCT Web. Specifically, these extension categories are _directives_, 
_controllers_, _services_, _constants_, _runs_, and _routes_. 
 
#### Angular Directives

New [directives]( https://docs.angularjs.org/guide/directive ) may be 
registered as extensions of the directives category. Implementations of 
directives in this category should take only dependencies as arguments, and 
should return a directive definition object. 

The directive's name should be provided as a key property of its extension 
definition, in camel-case format. 
 
#### Angular Controllers

New [controllers]( https://docs.angularjs.org/guide/controller ) may be registered 
as extensions of the controllers category. The implementation is registered 
directly as the controller; its only constructor arguments are its declared 
dependencies. 

The directive's identifier should be provided as a key property of its extension 
definition. 
   
 
#### Angular Services

New [services](https://docs.angularjs.org/guide/services ) may be registered as 
extensions of the services category. The implementation is registered via a 
[service call]( https://docs.angularjs.org/api/auto/service/$provide#service ), so 
it will be instantiated with the new operator. 

#### Angular Constants 

Constant values may be registered as extensions of the [ constants category](https://docs.angularjs.org/api/ng/type/angular.Module#constant ). 
These extensions have no implementation; instead, they should contain a property 
 key , which is the name under which the constant will be registered, and a 
property value , which is the constant value that will be registered.

#### Angular Runs 

In some cases, you want to register code to run as soon as the application 
starts; these can be registered as extensions of the [ runs category](https://docs.angularjs.org/api/ng/type/angular.Module#run ). 
Implementations registered in this category will be invoked (with their declared 
dependencies) when the Open MCT Web application first starts. (Note that, in 
this case, the implementation is better thought of as just a function, as 
opposed to a constructor function.)

#### Angular Routes

Extensions of category `routes` will be registered with Angular's [route provider](https://docs.angularjs.org/api/ngRoute/provider/$routeProvider ). 
Extensions of this category have no implementations, and need only two 
properties in their definition: 

* `when`: The value that will be passed as the path argument to `$routeProvider.when`; 
specifically, the string that will appear in the trailing 
part of the URL corresponding to this route. This property may be omitted, in 
which case this extension instance will be treated as the default route. 
* `templateUrl`: A path to the template to render for this route. Specified as a 
path relative to the bundle's resource directory (`res` by default.) 

### Composite Services

Composite services are described in the [relevant section](../architecture/Framework.md#Composite-Services) 
of the framework guide.

A component should include the following properties in its extension definition:

* `provides`: The symbolic identifier for the service that will be composed. The 
  fully-composed service will be registered with Angular under this name.
* `type`: One of `provider`, `aggregator` or `decorator` (as above) 

In addition to any declared dependencies, _aggregators_ and _decorators_ both 
receive one more argument (immediately following declared dependencies) that is 
provided by the framework. For an aggregator, this will be an array of all 
providers of the same service (that is, with matching `provides` properties); 
for a decorator, this will be whichever provider, decorator, or aggregator is 
next in the sequence of decorators. 

Services exposed by the Open MCT Web platform are often declared as composite 
services, as this form is open for a variety of common modifications. 

# Core API 

Most of Open MCT Web's relevant API is provided and/or mediated by the 
framework; that is, much of developing for Open MCT Web is a matter of adding 
extensions which access other parts of the platform by means of dependency 
injection. 

The core bundle (`platform/core`) introduces a few additional object types meant 
to be passed along by other services. 

## Domain Objects 

Domain objects are the most fundamental component of Open MCT Web's information 
model. A domain object is some distinct thing relevant to a user's work flow, 
such as a telemetry channel, display, or similar. Open MCT Web is a tool for 
viewing, browsing, manipulating, and otherwise interacting with a graph of 
domain objects. 

A domain object should be conceived of as the union of the following:

* __Identifier__: A machine-readable string that uniquely identifies the domain 
object within this application instance. 
* __Model__: The persistent state of the domain object. A domain object's model 
is a JavaScript object that can be losslessly converted to JSON. 
* __Capabilities__: Dynamic behavior associated with the domain object. 
Capabilities are JavaScript objects which provide additional methods for 
interacting with the domain objects which expose those capabilities. Not all 
domain objects expose all capabilities. 

At run-time, a domain object has the following interface:

* `getId()`: Get the identifier for this domain object. 
* `getModel()`: Get the plain state associated with this domain object. This 
will return a JavaScript object that can be losslessly converted to JSON. Note 
that the model returned here can be modified directly but should not be; 
instead, use the mutation capability. 
* `getCapability(key)`: Get the specified capability associated with this domain 
object. This will return a JavaScript object whose interface is specific to the 
type of capability being requested. If the requested capability is not exposed 
by this domain object, this will return undefined .
* `hasCapability(key)`: Shorthand for checking if a domain object exposes the 
requested capability.
* `useCapability(key, arguments  )`: Shorthand for 
`getCapability(key).invoke(arguments)`, with additional checking between calls. 
If the provided capability has no invoke method, the return value here functions 
as `getCapability` including returning `undefined` if the capability is not 
exposed.

## Domain Object Actions

An `Action` is behavior that can be performed upon/using a `DomainObject`. An 
Action has the following interface:

* `perform()`: Do this action. For example, if one had an instance of a 
`RemoveAction` invoking its perform method would cause the domain object which 
exposed it to be removed from its container.
* `getMetadata()`: Get metadata associated with this action. Returns an object 
containing: 
    * `name`: Human-readable name.
    * `description`: Human-readable summary of this action. 
    * `glyph`: Single character to be displayed in Open MCT Web's icon font set. 
    * `context`: The context in which this action is being performed (see below)

Action instances are typically obtained via a domain object's `action` 
capability. 
 
### Action Contexts 

An action context is a JavaScript object with the following properties: 

* `domainObject`: The domain object being acted upon. 
* `selectedObject`: Optional; the selection at the time of action (e.g. the 
dragged object in a drag-and-drop operation.)

## Telemetry 

Telemetry series data in Open MCT Web is represented by a common interface, and 
packaged in a consistent manner to facilitate passing telemetry updates around 
multiple visualizations. 
 
### Telemetry Requests 

A telemetry request is a JavaScript object containing the following properties: 

* `source`: A machine-readable identifier for the source of this telemetry. This 
is useful when multiple distinct data sources are in use side-by-side. 
* `key`: A machine-readable identifier for a unique series of telemetry within 
that source. 
* _Note: This API is still under development; additional properties, such as 
start and end time, should be present in future versions of Open MCT Web._ 

Additional properties may be included in telemetry requests which have specific 
interpretations for specific sources.

### Telemetry Responses

When returned from the `telemetryService` (see [Services](#Services) section), 
telemetry series data will be packaged in a `source -> key -> TelemetrySeries` 
fashion. That is, telemetry is passed in an object containing key-value pairs. 
Keys identify telemetry sources; values are objects containing additional 
key-value pairs. In this object, keys identify individual telemetry series (and 
match they `key` property from corresponding requests) and values are 
`TelemetrySeries` objects (see below.) 

### Telemetry Series

A telemetry series is a specific sequence of data, typically associated with a 
specific instrument. Telemetry is modeled as an ordered sequence of domain and 
range values, where domain values must be non-decreasing but range values do 
not. (Typically, domain values are interpreted as UTC timestamps in milliseconds 
relative to the UNIX epoch.) A series must have at least one domain and one 
range, and may have more than one.

Telemetry series data in Open MCT Web is expressed via the following 
`TelemetrySeries` interface: 

* `getPointCount()`: Returns the number of unique points/samples in this series. 
* `getDomainValue(index, [domain])`:  Get the domain value at the specified index . 
If a second domain argument is provided, this is taken as a string identifier 
indicating which domain option (of, presumably, multiple) should be returned. 
* `getRangeValue(index, [range])`:  Get the domain value at the specified index . 
If a second range argument is provided, this is taken as a string identifier 
indicating which range option (of, presumably, multiple) should be returned. 
 
### Telemetry Metadata 

Domain objects which have associated telemetry also expose metadata about that 
telemetry; this is retrievable via the `getMetadata()` of the telemetry 
capability. This will return a single JavaScript object containing the following 
properties: 

* `source`: The machine-readable identifier for the source of telemetry data for 
this object. 
* `key`: The machine-readable identifier for the individual telemetry series. 
* `domains`: An array of supported domains (see TelemetrySeries above.) Each 
domain should be expressed as an object which includes: 
    * `key`: Machine-readable identifier for this domain, as will be passed into 
    a getDomainValue(index, domain)  call. 
    * `name`: Human-readable name for this domain. 
* `ranges`: An array of supported ranges; same format as domains . 

Note that this metadata is also used as the prototype for telemetry requests 
made using this capability. 

## Types 
A domain object's type is represented as a Type object, which has the following 
interface:

* `getKey()`: Get the machine-readable identifier for this type. 
* `getName()`: Get the human-readable name for this type. 
* `getDescription()`: Get a human-readable summary of this type. 
* `getGlyph()`: Get the single character to be rendered as an icon for this type 
in Open MCT Web's custom font set. 
* `getInitialModel()`: Get a domain object model that represents the initial 
state (before user specification of properties) for domain objects of this type. 
* `getDefinition()`: Get the extension definition for this type, as a JavaScript 
object. 
* `instanceOf(type)`: Check if this type is (or inherits from) a specified type . 
This type can be either a string, in which case it is taken to be that type's 
 key , or it may be a  `Type` instance. 
* `hasFeature(feature)`: Returns a boolean value indicating whether or not this 
type supports the specified feature, which is a symbolic string. 
* `getProperties()`: Get all properties associated with this type, expressed as 
an array of `TypeProperty` instances. 
 
### Type Features 

Features of a domain object type are expressed as symbolic string identifiers. 
They are defined in practice by usage; currently, the Open MCT Web platform only 
uses the creation feature to determine which domain object types should appear 
in the Create menu. 
 
### Type Properties 

Types declare the user-editable properties of their domain object instances in 
order to allow the forms which appear in the __Create__ and __Edit Properties__ 
dialogs to be generated by the platform. A `TypeProperty` has the following interface:

* `getValue(model)`: Get the current value for this property, as it appears in 
the provided domain object model. 
* `setValue(model, value)`: Set a new value for this property in the provided 
domain object model . 
* `getDefinition()`: Get the raw definition for this property as a JavaScript 
object (as it was declared in this type's extension definition.) 

# Extension Categories 

The information in this section is focused on registering new extensions of 
specific types; it does not contain a catalog of the extension instances of 
these categories provided by the platform. Relevant summaries there are provided 
in subsequent sections.
 
## Actions Category

An action is a thing that can be done to or using a domain object, typically as 
initiated by the user. 

An action's implementation:

* Should take a single `context` argument in its constructor. (See Action 
Contexts, under Core API.)
* Should provide a method `perform` which causes the behavior associated with 
the action to occur.
* May provide a method `getMetadata` which provides metadata associated with 
the action. If omitted, one will be provided by the platform which includes 
metadata from the action's extension definition.
* May provide a static method `appliesTo(context)` (that is, a function 
available as a property of the implementation's constructor itself), which will 
be used by the platform to filter out actions from contexts in which they are 
inherently inapplicable.

An action's bundle definition (and/or `getMetadata()` return value) may include:

* `category`: A string or array of strings identifying which category or 
categories an action falls into; used to determine when an action is displayed. 
Categories supported by the platform include: 
    * `contextual`: Actions in a context menu. 
    * `view-control`: Actions triggered by buttons in the top-right of Browse 
    view. 
* `key`: A machine-readable identifier for this action. 
* `name`: A human-readable name for this action (e.g. to show in a menu) 
* `description`: A human-readable summary of the behavior of this action. 
* `glyph`: A single character which will be rendered in Open MCT Web's custom 
font set as an icon for this action.

## Capabilities Category

Capabilities are exposed by domain objects (e.g. via the `getCapability` method) 
but most commonly originate as extensions of this category.

Extension definitions for capabilities should include both an implementation, 
and a property named key whose value should be a string used as a 
machine-readable identifier for that capability, e.g. when passed as the 
argument to a domain object's `getCapability(key)` call.
 
A capability's implementation should have methods specific to that capability; 
that is, there is no common format for capability implementations, aside from 
support for invocation via the `useCapability` shorthand.

A capability's implementation will take a single argument (in addition to any 
declared dependencies), which is the domain object that will expose that 
capability.

A capability's implementation may also expose a static method `appliesTo(model)` 
which should return a boolean value, and will be used by the platform to filter 
down capabilities to those which should be exposed by specific domain objects, 
based on their domain object models. 
 
## Controls Category

Controls provide options for the `mct-control` directive. 
 
Six standard control types are included in the forms bundle:

* `textfield`: An area to enter plain text.
* `select`: A drop-down list of options.
* `checkbox`: A box which may be checked/unchecked.
* `color`: A color picker.
* `button`: A button.
* `datetime`: An input for UTC date/time entry; gives result as a UNIX 
timestamp, in milliseconds since start of 1970, UTC. 

New controls may be added as extensions of the controls category. Extensions of 
this category have two properties: 

* `key`: The symbolic name for this control (matched against the control field 
in rows of the form structure).
* `templateUrl`: The URL to the control's Angular template, relative to the 
resources directory of the bundle which exposes the extension. 

Within the template for a control, the following variables will be included in 
scope:

* `ngModel`: The model where form input will be stored. Notably we also need to 
look at field  (see below) to determine which field in the model should be 
modified. 
* `ngRequired`: True if input is required.
* `ngPattern`: The pattern to match against (for text entry)
* `options`: The options for this control, as passed from the `options` property 
of an individual row definition. 
* `field`: Name of the field in `ngModel` which will hold the value for this 
control. 

## Gestures Category

A _gesture_ is a user action which can be taken upon a representation of a 
domain object. 

Examples of gestures included in the platform are:

* `drag`: For representations that can be used to initiate drag-and-drop 
composition.
* `drop`: For representations that can be drop targets for drag-and-drop 
composition. 
* `menu`: For representations that can be used to pop up a context menu. 
 
Gesture definitions have a property `key` which is used as a machine-readable 
identifier for the gesture (e.g. `drag`, `drop`, `menu` above.) 
 
A gesture's implementation is instantiated once per representation that uses the 
gesture. This class will receive the jqLite-wrapped `mct-representation` element 
and the domain object being represented as arguments, and should do any 
necessary "wiring" (e.g. listening for events) during its constructor call. The 
gesture's implementation may also expose an optional `destroy()`  method which 
will be called when the gesture should be removed, to avoid memory leaks by way 
of unremoved listeners.

## Indicators Category

An indicator is an element that should appear in the status area at the bottom 
of a running Open MCT Web client instance. 

### Standard Indicators 
 
Indicators which wish to appear in the common form of an icon-text pair should 
provide implementations with the following methods:

* `getText()`: Provides the human-readable text that will be displayed for this 
indicator. 
* `getGlyph()`: Provides a single-character string that will be displayed as an 
icon in Open MCT Web's custom font set. 
* `getDescription()`: Provides a human-readable summary of the current state of 
this indicator; will be displayed in a tooltip on hover. 
* `getClass()`: Get a CSS class that will be applied to this indicator. 
* `getTextClass()`: Get a CSS class that will be applied to this indicator's 
text portion. 
* `getGlyphClass()`: Get a CSS class that will be applied to this indicator's 
icon portion. 
* `configure()`: If present, a configuration icon will appear to the right of 
this indicator, and clicking it will invoke this method. 
 
Note that all methods are optional, and are called directly from an Angular 
template, so they should be appropriate to run during digest cycles. 

### Custom Indicators 

Indicators which wish to have an arbitrary appearance (instead of following the 
icon-text convention commonly used) may specify a `template` property in their 
extension definition. The value of this property will be used as the `key` for 
an `mct-include` directive (so should refer to an extension of category 
 templates .) This template will be rendered to the status area. Indicators of 
this variety do not need to provide an implementation. 

## Licenses Category

The extension category `licenses` can be used to add entries into the 'Licensing 
information' page, reachable from Open MCT Web's About dialog. 

Licenses may have the following properties, all of which are strings:

* `name`: Human-readable name of the licensed component. (e.g. 'AngularJS'.)
* `version`: Human-readable version of the licensed component. (e.g. '1.2.26'.)
* `description`: Human-readable summary of the component.
* `author`: Name or names of entities to which authorship should be attributed.
* `copyright`: Copyright text to display for this component.
* `link`: URL to full license text. 

## Policies Category

Policies are used to handle decisions made using Open MCT Web's `policyService`; 
examples of these decisions are determining the applicability of certain 
actions, or checking whether or not a domain object of one type can contain a 
domain object of a different type. See the section on the Policies for an 
overview of Open MCT Web's policy model.

A policy's extension definition should include:

* `category`: The machine-readable identifier for the type of policy decision 
being supported here. For a list of categories supported by the platform, see 
the section on Policies. Plugins may introduce and utilize additional policy 
categories not in that list. 
* `message`: Optional; a human-readable message describing the policy, intended 
for display in situations where this specific policy has disallowed something. 
 
A policy's implementation should include a single method, `allow(candidate,
context)`. The specific types used for `candidate` and `context` vary by policy 
category; in general, what is being asked is 'is this candidate allowed in this 
context?' This method should return a boolean value. 

Open MCT Web's policy model requires consensus; a policy decision is allowed 
when and only when all policies choose to allow it. As such, policies should 
generally be written to reject a certain case, and allow (by returning `true`) 
anything else. 
 
## Representations Category

A representation is an Angular template used to display a domain object. The 
`representations` extension category is used to add options for the 
`mct-representation` directive. 
 
A representation definition should include the following properties:

* `key`: The machine-readable name which identifies the representation. 
* `templateUrl`: The path to the representation's Angular template. This path is 
relative to the bundle's resources directory. 
* `uses`: Optional; an array of capability names. Indicates that this 
representation intends to use those capabilities of a domain object (via a 
`useCapability` call), and expects to find the latest results of that 
`useCapability` call in the scope of the presented template (under the same name 
as the capability itself.) Note that, if `useCapability` returns a promise, this 
will be resolved before being placed in the representation's scope. 
* `gestures`: An array of keys identifying gestures (see the `gestures` 
extension category) which should be available upon this representation. Examples 
of gestures include `drag` (for representations that should act as draggable 
sources for drag-drop operations) and `menu` (for representations which should 
show a domain-object-specific context menu on right-click.) 

### Representation Scope

While  _representations_ do not have implementations, per se, they do refer to 
Angular templates which need to interact with information (e.g. the domain 
object being represented) provided by the platform. This information is passed 
in through the template's scope, such that simple representations may be created 
by providing only templates. (More complex representations will need controllers 
which are referenced from templates. See [https://docs.angularjs.org/guide/controller ]()
for more information on controllers in Angular.) 
 
A representation's scope will contain:

* `domainObject`: The represented domain object.
* `model`: The domain object's model.
* `configuration`: An object containing configuration information for this 
representation (an empty object if there is no saved configuration.) The 
contents of this object are managed entirely by the view/representation which 
receives it. 
* `representation`: An empty object, useful as a 'scratch pad' for 
representation state. 
* `ngModel`: An object passed through the ng-model attribute of the 
`mct-representation` , if any. 
* `parameters`: An object passed through the parameters attribute of the 
`mct-representation`, if any. 
* Any capabilities requested by the uses property of the representation 
definition.
 
## Representers Category

The `representers` extension category is used to add additional behavior to the 
`mct-representation` directive. This extension category is intended primarily 
for use internal to the platform. 

Unlike _representations_, which describe specific ways to represent domain 
objects, _representers_ are used to modify or augment the process of 
representing domain objects in general. For example, support for the  _gestures_ 
extension category is added by a _representer_.

A representer needs only provide an implementation. When an `mct-representation` 
is linked (see  [https://docs.angularjs.org/guide/directive ]() or when the 
domain object being represented changes, a new _representer_ of each declared 
type is instantiated. The constructor arguments for a _representer_ are the same 
as the arguments to the link function in an Angular directive: `scope` the 
Angular scope for this representation; `element` the jqLite-wrapped 
`mct-representation` element, and `attrs` a set of key-value pairs of that 
element's attributes. _Representers_ may wish to populate the scope, attach 
event listeners to the element, etc.

This implementation must provide a single method, `destroy()`, which will be 
invoked when the representer is no longer needed. 

## Roots Category

The extension category `roots` is used to provide root-level domain object 
models. Root-level domain objects appear at the top-level of the tree hierarchy. 
For example, the _My Items_ folder is added as an extension of this category. 

Extensions of this category should have the following properties:

* `id`: The machine-readable identifier for the domaiwn object being exposed.
* `model`: The model, as a JSON object, for the domain object being exposed. 

## Stylesheets Category

The stylesheets extension category is used to add CSS files to style the 
application. Extension definitions for this category should include one 
property:

* `stylesheetUrl`: Path and filename, including extension, for the stylesheet to 
include. This path is relative to the bundle's resources folder (by default,  
`res`) 
* `theme`: Optional; if present, this stylesheet will only be included if this
value matches the `THEME` constant.
 
To control the order of CSS files, use priority  (see the section on Extension 
Definitions above.) 

## Templates Category

The `templates` extension category is used to expose Angular templates under 
symbolic identifiers. These can then be utilized using the `mct-include` 
directive, which behaves similarly to `ng-include` except that it uses these 
symbolic identifiers instead of paths.

A template's extension definition should include the following properties:

* `key`: The machine-readable name which identifies this template, matched 
against the value given to the key attribute of the `mct-include` directive.
* `templateUrl`: The path to the relevant Angular template. This path is 
relative to the bundle's resources directory. 

Note that, when multiple templates are present with the same key , the one with 
the highest priority will be used from `mct-include`. This behavior can be used 
to override templates exposed by the platform (to change the logo which appears 
in the bottom right, for instance.)

Templates do not have implementations. 

## Types Category

The types extension category describes types of domain objects which may 
appear within Open MCT Web.

A type's extension definition should have the following properties:

* `key`: The machine-readable identifier for this domain object type. Will be 
stored to and matched against the type property of domain object models.
* `name`: The human-readable name for this domain object type.
* `description`: A human-readable summary of this domain object type.
* `glyph`: A single character to be rendered as an icon in Open MCT Web's custom 
font set. 
* `model`: A domain object model, used as the initial state for created domain 
objects of this type (before any properties are specified.)
* `features`: Optional; an array of strings describing features of this domain 
object type. Currently, only creation is recognized by the platform; this is 
used to determine that this type should appear in the Create menu. More 
generally, this is used to support the `hasFeature(...)`  method of the type  
capability. 
* `properties`: An array describing individual properties of this domain object
(as should appear in the _Create_ or the _Edit Properties_ dialog.) Each 
property is described by an object containing the following properties:
    * `control`: The key of the control (see `mct-control` and the `controls` 
    [extension category](#Controls)) to use for editing this property. 
    * `property`: A string which will be used as the name of the property in the 
    domain object's model that the value for this property should be stored 
    under. If this value should be stored in an object nested within the domain 
    object model, then property should be specified as an array of strings 
    identifying these nested objects and, finally, the property itself. 
    * other properties as appropriate for a control of this type (each 
    property's definition will also be passed in as the structure for its 
    control.) See documentation of mct-form for more detail on these 
    properties.
    
Types do not have implementations. 
 
## Versions Category
The versions extension category is used to introduce line items in Open MCT 
Web's About dialog. These should have the following properties: 

* `name`: The name of this line item, as should appear in the left-hand side of 
the list of version information in the About dialog.
* `value`: The value which should appear to the right of the name in the About 
dialog.

To control the ordering of line items within the About dialog, use `priority`. 
(See section on [Extension Definitions](#ExtensionDefinitions) above.) 

This extension category does not have implementations. 
 
## Views Category

The views extension category is used to determine which options appear to the 
user as available views of domain objects of specific types. A view's extension 
definition has the same properties as a representation (and views can be 
utilized via `mct-representation`); additionally:

* `name`: The human-readable name for this view type.
* description : A human-readable summary of this view type.
* `glyph`: A single character to be rendered as an icon in Open MCT Web's custom 
font set.
* `type`: Optional; if present, this representation is only applicable for 
domain object's of this type.
* `needs`: Optional array of strings; if present, this representation is only 
applicable for domain objects which have the capabilities identified by these 
strings. 
* `delegation`: Optional boolean, intended to be used in conjunction with  
`needs`;  if present, allow required capabilities to be satisfied by means of 
capability delegation. (See [Delegation](#Delegation))
* `toolbar`: Optional; a definition for the toolbar which may appear in a 
toolbar when using this view in Edit mode. This should be specified as a 
structure for mct-toolbar , with additional properties available for each item in 
that toolbar: 
    * `property`: A property name. This will refer to a property in the view's 
    current selection; that property on the selected object will be modifiable 
    as the `ng-model` of the displayed control in the toolbar. If the value of 
    the property is a function, it will be used as a getter-setter (called with 
    no arguments to use as a getter, called with a value to use as a setter.) 
    * `method`: A method to invoke (again, on the selected object) from the 
    toolbar control. Useful particularly for buttons (which don't edit a single 
    property, necessarily.)

### View Scope 

Views do not have implementations, but do get the same properties in scope that 
are provided for `representations`. 

When a view is in Edit mode, this scope will additionally contain:

* `commit()`: A function which can be invoked to mark any changes to the view's 
  configuration as ready to persist.
* `selection`: An object representing the current selection state. 

#### Selection State 

A view's selection state is, conceptually, a set of JavaScript objects. The 
presence of methods/properties on these objects determine which toolbar controls 
are visible, and what state they manage and/or behavior they invoke. 

This set may contain up to two different objects: The  _view proxy _, which is 
used to make changes to the view as a whole, and the _ selected object _, which is 
used to represent some state within the view. (Future versions of Open MCT Web 
may support multiple selected objects.) 

The `selection` object made available during Edit mode has the following 
methods: 

* `proxy([object])`: Get (or set, if called with an argument) the current view 
proxy.  
* `select(object)`: Make this object the selected object. 
* `deselect()`: Clear the currently selected object. 
* `get()`: Get the currently selected object. Returns undefined if there is no 
currently selected object.
* `selected(object)`: Check if the JavaScript object is currently in the 
selection set. Returns true if the object is either the currently selected 
object, or the current view proxy. 
* `all()`: Get an array of all objects in the selection state. Will include 
either or both of the view proxy and selected object. 

# Directives

Open MCT Web defines several Angular directives that are intended for use both 
internally within the platform, and by plugins. 

## Before Unload 

The `mct-before-unload` directive is used to listen for (and prompt for user 
confirmation) of navigation changes in the browser. This includes reloading, 
following links out of Open MCT Web, or changing routes. It is used to hook into 
both `onbeforeunload` event handling as well as route changes from within 
Angular.

This directive is useable as an attribute. Its value should be an Angular 
expression. When an action that would trigger an unload and/or route change 
occurs, this Angular expression is evaluated. Its result should be a message to 
display to the user to confirm their navigation change; if this expression 
evaluates to a falsy value, no message will be displayed. 
 
## Chart 

The `mct-chart` directive is used to support drawing of simple charts. It is 
present to support the Plot view, and its functionality is limited to the 
functionality that is relevant for that view.

This directive is used at the element level and takes one attribute, `draw` 
which is an Angular expression which will should evaluate to a drawing object. 
This drawing object should contain the following properties:

* `dimensions`: The size, in logical coordinates, of the chart area. A 
two-element array or numbers. 
* `origin`: The position, in logical coordinates, of the lower-left corner of 
the chart area. A two-element array or numbers. 
* `lines`: An array of lines (e.g. as a plot line) to draw, where each line is 
expressed as an object containing: 
    * `buffer`: A Float32Array containing points in the line, in logical 
    coordinates, in sequential x,y pairs. 
    * `color`: The color of the line, as a four-element RGBA array, where 
    each element is a number in the range of 0.0-1.0. 
    * `points`: The number of points in the line. 
* `boxes`: An array of rectangles to draw in the chart area. Each is an object 
containing: 
    * `start`: The first corner of the rectangle, as a two-element array of 
    numbers, in logical coordinates. 
    * `end`: The opposite corner of the rectangle, as a two-element array of 
    numbers, in logical coordinates. color : The color of the line, as a 
    four-element RGBA array, where each element is a number in the range of 
    0.0-1.0. 

While `mct-chart` is intended to support plots specifically, it does perform 
some useful management of canvas objects (e.g. choosing between WebGL and Canvas 
2D APIs for drawing based on browser support) so its usage is recommended when 
its supported drawing primitives are sufficient for other charting tasks. 
 
## Container 

The `mct-container` is similar to the `mct-include` directive insofar as it allows 
templates to be referenced by symbolic keys instead of by URL. Unlike 
`mct-include` it supports transclusion.

Unlike `mct-include` `mct-container` accepts a key as a plain string attribute, 
instead of as an Angular expression.

## Control

The `mct-control` directive is used to display user input elements. Several 
controls are included with the platform to wrap default input types. This 
directive is primarily intended for internal use by the `mct-form` and 
`mct-toolbar` directives. 

When using `mct-control` the attributes `ng-model` `ng-disabled` 
`ng-required` and `ng-pattern` may also be used. These have the usual meaning 
(as they would for an input element) except for `ng-model`; when used, it will 
actually be `ngModel[field]` (see below) that is two-way bound by this control. 
This allows `mct-control` elements to more easily delegate to other 
`mct-control` instances, and also facilitates usage for generated forms. 

This directive supports the following additional attributes, all specified as 
Angular expressions:

* `key`: A machine-readable identifier for the specific type of control to 
display.
* `options`: A set of options to display in this control.
* `structure`: In practice, contains the definition object which describes this 
form row or toolbar item. Used to pass additional control-specific parameters. 
* `field`: The field in the `ngModel` under which to read/store the property 
associated with this control. 

## Drag

The `mct-drag` directive is used to support drag-based gestures on HTML 
elements. Note that this is not 'drag' in the 'drag-and-drop' sense, but 'drag' 
in the more general 'mouse down, mouse move, mouse up' sense. 

This takes the form of three attributes: 

* `mct-drag`: An Angular expression to evaluate during drag movement.
* `mct-drag-down`: An Angular expression to evaluate when the drag starts.
* `mct-drag-up`: An Angular expression to evaluate when the drag ends.

In each case, a variable `delta` will be provided to the expression; this is a 
two-element array or the horizontal and vertical pixel offset of the current 
mouse position relative to the mouse position where dragging began. 

## Form 

The `mct-form` directive is used to generate forms using a declarative structure, 
and to gather back user input. It is applicable at the element level and 
supports the following attributes: 

* `ng-model`: The object which should contain the full form input. Individual 
fields in this model are bound to individual controls; the names used for these 
fields are provided in the form structure (see below).
* `structure`: The structure of the form; e.g. sections, rows, their names, and 
so forth. The value of this attribute should be an Angular expression. 
* `name`: The name in the containing scope under which to publish form 
"meta-state", e.g. `$valid` `$dirty` etc. This is as the behavior of `ng-form`. 
Passed as plain text in the attribute. 

### Form Structure 

Forms in Open MCT Web have a common structure to permit consistent display. A 
form is broken down into sections, which will be displayed in groups; each 
section is broken down into rows, each of which provides a control for a single 
property. Input from this form is two-way bound to the object passed via 
`ng-model`. 

A form's structure is represented by a JavaScript object in the following form:
 
    { 
        "name": ... title to display for the form, as a string ..., 
        "sections": [
        { 
            "name": ... title to display for the section ..., 
            "rows": [ 
                { 
                    "name": ... title to display for this row ...,
                    "control": ... symbolic key for the control ..., 
                    "key": ... field name in ng-model ... 
                    "pattern": ... optional, reg exp to match against ... 
                    "required": ... optional boolean ... 
                    "options": [ 
                        "name": ... name to display (e.g. in a select) ..., 
                        "value": ... value to store in the model ... 
                    ] 
                    }, 
                ... and other rows ... 
            ] 
        }, 
        ... and other sections ... 
        ] 
    } 

Note that `pattern` may be specified as a string, to simplify storing for 
structures as JSON when necessary. The string should be given in a form 
appropriate to pass to a `RegExp` constructor. 

### Form Controls 

A few standard control types are included in the platform/forms bundle: 

* `textfield`: An area to enter plain text. 
* `select`: A drop-down list of options. 
* `checkbox`: A box which may be checked/unchecked. 
* `color`: A color picker. 
* `button`: A button. 
* `datetime`: An input for UTC date/time entry; gives result as a UNIX 
timestamp, in milliseconds since start of 1970, UTC. 

## Include 

The `mct-include` directive is similar to ng-include , except that it takes a 
symbolic identifier for a template instead of a URL. Additionally, templates 
included via mct-include will have an isolated scope. 

The directive should be used at the element level and supports the following 
attributes, all of which are specified as Angular expressions: 

* `key`: Machine-readable identifier for the template (of extension category  
templates ) to be displayed. 
* `ng-model`: _Optional_; will be passed into the template's scope as `ngModel`. 
Intended usage is for two-way bound user input.
* `parameters`: _Optional_; will be passed into the template's scope as 
parameters. Intended usage is for template-specific display parameters. 

## Representation 

The `mct-representation` directive is used to include templates which 
specifically represent domain objects. Usage is similar to `mct-include`. 

The directive should be used at the element level and supports the following 
attributes, all of which are specified as Angular expressions:

* `key`: Machine-readable identifier for the representation (of extension 
category _representations_ or _views_ ) to be displayed. 
* `mct-object`: The domain object being represented. 
* `ng-model`: Optional; will be passed into the template's scope as `ngModel`. 
Intended usage is for two-way bound user input. 
* `parameters`: Optional; will be passed into the template's scope as  
parameters . Intended usage is for template-specific display parameters. 

## Resize 

The `mct-resize` directive is used to monitor the size of an HTML element. It is 
specified as an attribute whose value is an Angular expression that will be 
evaluated when the size of the HTML element changes. This expression will be 
provided a single variable, `bounds` which is an object containing two 
properties, `width` and `height` describing the size in pixels of the element.

When using this directive, an attribute `mct-resize-interval` may optionally be 
provided. Its value is an Angular expression describing the number of 
milliseconds to wait before next checking the size of the HTML element; this 
expression is evaluated when the directive is linked and reevaluated whenever 
the size is checked.

## Scroll 

The `mct-scroll-x` and `mct-scroll-y` directives are used to both monitor and 
control the horizontal and vertical scroll bar state of an element, 
respectively. They are intended to be used as attributes whose values are 
assignable Angular expressions which two-way bind to the scroll bar state.

## Toolbar

The `mct-toolbar` directive is used to generate toolbars using a declarative 
structure, and to gather back user input. It is applicable at the element level 
and supports the following attributes: 

* `ng-model`: The object which should contain the full toolbar input. Individual 
fields in this model are bound to individual controls; the names used for these 
fields are provided in the form structure (see below). 
* `structure`: The structure of the toolbar; e.g. sections, rows, their names, and 
so forth. The value of this attribute should be an Angular expression.
* `name`: The name in the containing scope under which to publish form 
"meta-state", e.g. `$valid`, `$dirty` etc. This is as the behavior of 
`ng-form`. Passed as plain text in the attribute. 

Toolbars support the same control options as forms.  

### Toolbar Structure 

A toolbar's structure is defined similarly to forms, except instead of rows  
there are items . 

    { 
        "name": ... title to display for the form, as a string ..., 
        "sections": [ 
            { 
                "name": ... title to display for the section ..., 
                "items": [ 
                    { 
                        "name": ... title to display for this row ..., 
                        "control": ... symbolic key for the control ..., 
                        "key": ... field name in ng-model ... 
                        "pattern": ... optional, reg exp to match against ... 
                        "required": ... optional boolean ... 
                        "options": [ 
                            "name": ... name to display (e.g. in a select) ..., 
                            "value": ... value to store in the model ... 
                        ], 
                        "disabled": ... true if control should be disabled ... 
                        "size": ... size of the control (for textfields) ... 
                        "click": ... function to invoke (for buttons) ... 
                        "glyph": ... glyph to display (for buttons) ... 
                        "text": ... text within control (for buttons) ... 
                    }, 
                    ... and other rows ... 
                ] 
            }, 
            ... and other sections ... 
        ] 
    }

# Services 

The Open MCT Web platform provides a variety of services which can be retrieved 
and utilized via dependency injection. These services fall into two categories: 

* _Composite Services_ are defined by a set of components extensions; plugins may 
introduce additional components with matching interfaces to extend or augment 
the functionality of the composed service. (See the Framework section on 
Composite Services.) 
* _Other services_ which are defined as standalone service objects; these can be 
utilized by plugins but are not intended to be modified or augmented. 

## Composite Type Services

This section describes the composite services exposed by Open MCT Web, 
specifically focusing on their interface and contract. 
    
In many cases, the platform will include a provider for a service which consumes 
a specific extension category; for instance, the `actionService` depends on 
`actions[]` and will expose available actions based on the rules defined for 
that extension category.  

In these cases, it will usually be simpler to add a new extension of a given 
category (e.g. of category `actions`) even when the same behavior could be 
introduced by a service component (e.g. an extension of category `components` 
where `provides` is `actionService` and `type` is `provider`.)  

Occasionally, the extension category does not provide enough expressive power to 
achieve a desired result. For instance, the Create menu is populated with 
`create` actions, where one such action exists for each creatable type. Since 
the framework does not provide a declarative means to introduce a new action per 
type declaratively, the platform implements this explicitly in an `actionService` 
component of type `provider`. Plugins may use a similar approach when the normal 
extension mechanism is insufficient to achieve a desired result. 
 
### Action Service

The [Action Service](../architecture/platform#action-service) (`actionService`) 
provides `Action` instances which are applicable in specific contexts. See Core 
API for additional notes on the interface for actions. The `actionService` has 
the following interface: 

* `getActions(context)`: Returns an array of Action objects which are applicable 
in the specified action context.   

### Capability Service 

The [Capability Service](../architecture/platform#capability-service) (`capabilityService`) 
provides constructors for capabilities which will be exposed for a given domain 
object. 

The capabilityService has the following interface: 

* `getCapabilities(model)`: Returns a an object containing key-value pairs, 
representing capabilities which should be exposed by the domain object with this 
model. Keys in this object are the capability keys (as used in a 
`getCapability(...)` call) and values are either: 
    * Functions, in which case they will be used as constructors, which will 
    receive the domain object instance to which the capability applies as their 
    sole argument.The resulting object will be provided as the result of a 
    domain object's `getCapability(...)`  call. Note that these instances are cached 
    by each object, but may be recreated when an object is mutated. 
    * Other objects, which will be used directly as the result of a domain 
    object's `getCapability(...)` call. 

### Dialog Service 

The `dialogService` provides a means for requesting user input via a modal 
dialog. It has the following interface: 

* `getUserInput(formStructure, formState)`: Prompt the user to fill out a form. 
The first argument describes the form's structure (as will be passed to 
 mct-form ) while the second argument contains the initial state of that form. 
This returns a Promise for the state of the form after the user has filled it 
in; this promise will be rejected if the user cancels input. 
* `getUserChoice(dialogStructure)`: Prompt the user to make a single choice from 
a set of options, which (in the platform implementation) will be expressed as 
buttons in the displayed dialog. Returns a Promise for the user's choice, which 
will be rejected if the user cancels input. 

### Dialog Structure 

The object passed as the `dialogStructure` to `getUserChoice` should have the 
following properties:

* `title`: The title to display at the top of the dialog. 
* `hint`: Short message to display below the title. 
* `template`: Identifying key  (as will be passed to mct-include ) for the 
template which will be used to populate the inner area of the dialog. 
* `model`: Model to pass in the ng-model attribute of mct-include . 
* `parameters`: Parameters to pass in the parameters attribute of mct-include . 
* `options`: An array of options describing each button at the bottom. Each 
option may have the following properties:
    * `name`: Human-readable name to display in the button. 
    * `key`: Machine-readable key, to pass as the result of the resolved promise 
    when clicked. 
    * `description`: Description to show in tooltip on hover.

### Domain Object Service 

The [Object Service](../architecture/platform.md#object-service) (`objectService`)
provides domain object instances. It has the following interface:

* `getObjects(ids)`: For the provided array of domain object identifiers, 
returns a Promise for an object containing key-value pairs, where keys are 
domain object identifiers and values are corresponding DomainObject instances. 
Note that the result may contain a superset or subset of the objects requested. 

### Gesture Service 

The `gestureService` is used to attach gestures (see extension category gestures) 
to representations. It has the following interface:

* `attachGestures(element, domainObject, keys)`: Attach gestures specified by 
the provided gesture keys  (an array of strings) to this jqLite-wrapped HTML 
element , which represents the specified domainObject . Returns an object with a 
single method `destroy()`, to be invoked when it is time to detach these 
gestures. 

### Model Service 

The [Model Service](../architecture/platform.md#model-service) (`modelService`) 
provides domain object models. It has the following interface: 

* `getModels(ids)`: For the provided array of domain object identifiers, returns 
a Promise for an object containing key-value pairs, where keys are domain object 
identifiers and values are corresponding domain object models. Note that the 
result may contain a superset or subset of the models requested. 

### Persistence Service 

The [Persistence Service](../architecture/platform.md#persistence-service) (`persistenceService`)
provides the ability to load/store JavaScript objects 
(presumably serializing/deserializing to JSON in the process.) This is used 
primarily to store domain object models. It has the following interface: 

* `listSpaces()`: Returns a Promise for an array of strings identifying the 
different persistence spaces this service supports. Spaces are intended to be 
used to distinguish between different underlying persistence stores, to allow 
these to live side by side. 
* `listObjects()`: Returns a Promise for an array of strings identifying all 
documents stored in this persistence service. 
* `createObject(space, key, value)`: Create a new document in the specified 
persistence space , identified by the specified key , the contents of which shall 
match the specified value . Returns a promise that will be rejected if creation 
fails. 
* `readObject(space, key)`: Read an existing document in the specified 
persistence space , identified by the specified key . Returns a promise for the 
specified document; this promise will resolve to undefined if the document does 
not exist. 
* `updateObject(space, key, value)`: Update an existing document in the 
specified persistence space , identified by the specified key , such that its 
contents match the specified value . Returns a promise that will be rejected if 
the update fails. 
* `deleteObject(space, key)`: Delete an existing document from the specified 
persistence space , identified by the specified key . Returns a promise which will 
be rejected if deletion fails. 

### Policy Service 

The [Policy Service](../architecture/platform.md#policy-service) (`policyService`) 
may be used to determine whether or not certain behaviors are 
allowed within the application. It has the following interface: 
 
* `allow(category, candidate, context, [callback])`: Check if this decision 
should be allowed. Returns a boolean. Its arguments are interpreted as: 
    * `category`: A string identifying which kind of decision is being made. See 
    the [section on Categories](#PolicyCategories) for categories supported by 
    the platform; plugins may define and utilize policies of additional 
    categories, as well. 
    * `candidate`: An object representing the thing which shall or shall not be 
    allowed. Usually, this will be an instance of an extension of the category 
    defined above. This does need to be the case; additional policies which are 
    not specific to any extension may also be defined and consulted using unique 
    category identifiers. In this case, the type of the object delivered for the 
    candidate may be unique to the policy type. 
    * `context`: An object representing the context in which the decision is 
    occurring. Its contents are specific to each policy category. 
    * `callback`: Optional; a function to call if the policy decision is rejected. 
    This function will be called with the message string (which may be 
    undefined) of whichever individual policy caused the operation to fail. 

### Telemetry Service 

The [Telemetry Service](../architecture/platform.md#telemetry-service) (`telemetryService`)
is used to acquire telemetry data. See the section on 
Telemetry in Core API for more information on how both the arguments and 
responses of this service are structured. 

When acquiring telemetry for display, it is recommended that the 
`telemetryHandler` service be used instead of this service. The 
`telemetryHandler` has additional support for subscribing to and requesting 
telemetry data associated with domain objects or groups of domain objects. See 
the [Other Services](#Other-Services) section for more information. 

The `telemetryService` has the following interface:

* `requestTelemetry(requests)`: Issue a request for telemetry, matching the 
specified telemetry requests . Returns a _ Promise _ for a telemetry response 
object.
* `subscribe(callback, requests)`: Subscribe to real-time updates for telemetry, 
matching the specified `requests`. The specified `callback` will be invoked with 
telemetry response objects as they become available. This method returns a 
function which can be invoked to terminate the subscription. 

### Type Service 

The [Type Service](../architecture/platform.md#type-service) (`typeService`) exposes 
domain object types. It has the following interface:

* `listTypes()`: Returns all domain object types supported in the application, 
as an array of `Type` instances.
* `getType(key)`: Returns the `Type` instance identified by the provided key, or 
undefined if no such type exists. 

### View Service 

The [View Service](../architecture/platform.md#view-service) (`viewService`) exposes 
definitions for views of domain objects. It has the following interface:
 
* `getViews(domainObject)`:  Get an array of extension definitions of category 
`views` which are valid and applicable to the specified `domainObject`. 
 
## Other Services

### Drag and Drop 

The `dndService` provides information about the content of an active 
drag-and-drop gesture within the application. It is intended to complement the 
`DataTransfer` API of HTML5 drag-and-drop, by providing access to non-serialized 
JavaScript objects being dragged, as well as by permitting inspection during 
drag (which is normally prohibited by browsers for security reasons.) 

The `dndService` has the following methods: 

* `setData(key, value)`: Set drag data associated with a given type, specified 
by the `key` argument. 
* `getData(key)`: Get drag data associated with a given type, specified by the  
`key` argument. 
* `removeData(key)`: Clear drag data associated with a given type, specified by 
the `key` argument. 

### Navigation 
 
The _Navigation_ service provides information about the current navigation state 
of the application; that is, which object is the user currently viewing? This 
service merely tracks this state and notifies listeners; it does not take 
immediate action when navigation changes, although its listeners might. 

The `navigationService` has the following methods:

* `getNavigation()`: Get the current navigation state. Returns a `DomainObject`. 
* `setNavigation(domainObject)`: Set the current navigation state. Returns a 
`DomainObject`. 
* `addListener(callback)`: Listen for changes in navigation state. The provided 
`callback` should be a `Function` which takes a single `DomainObject` as an 
argument. 
* `removeListener(callback)`: Stop listening for changes in navigation state. 
The provided `callback` should be a `Function` which has previously been passed 
to addListener .

### Now 

The service now is a function which acts as a simple wrapper for `Date.now()`. 
It is present mainly so that this functionality may be more easily mocked in 
tests for scripts which use the current time. 

### Telemetry Formatter 

The _Telemetry Formatter_ is a utility for formatting domain and range values 
read from a telemetry series. 

`telemetryFormatter` has the following methods: 

* `formatDomainValue(value)`: Format the provided domain value (which will be 
assumed to be a timestamp) for display; returns a string. 
* `formatRangeValue(value)`: Format the provided range value (a number) for 
display; returns a string.

### Telemetry Handler 

The _Telemetry Handler_ is a utility for retrieving telemetry data associated 
with domain objects; it is particularly useful for dealing with cases where the 
telemetry capability is delegated to contained objects (as occurs 
in _Telemetry Panels_.) 

The `telemetryHandler` has the following methods: 

* `handle(domainObject, callback, [lossless])`: Subscribe to and issue future 
requests for telemetry associated with the provided `domainObject`, invoking the 
provided callback function when streaming data becomes available. Returns a 
`TelemetryHandle` (see below.) 

#### Telemetry Handle 

A TelemetryHandle has the following methods: 

* `getTelemetryObjects()`: Get the domain objects (as a `DomainObject[]`) that 
have a telemetry capability and are being handled here. Note that these are 
looked up asynchronously, so this method may return an empty array if the 
initial lookup is not yet completed. 
* `promiseTelemetryObjects()`: As `getTelemetryObjects()`, but returns a Promise  
that will be fulfilled when the lookup is complete. 
* `unsubscribe()`: Unsubscribe to streaming telemetry updates associated with 
this handle. 
* `getDomainValue(domainObject)`: Get the most recent domain value received via 
a streaming update for the specified `domainObject`. 
* `getRangeValue(domainObject)`: Get the most recent range value received via a 
streaming update for the specified `domainObject`. 
* `getMetadata()`: Get metadata (as reported by the `getMetadata()` method of a 
telemetry capability) associated with telemetry-providing domain objects. 
Returns an array, which is in the same order as getTelemetryObjects() . 
* `request(request, callback)`: Issue a new request for historical telemetry 
data. The provided callback will be invoked when new data becomes available, 
which may occur multiple times (e.g. if there are multiple domain objects.) It 
will be invoked with the DomainObject for which a new series is available, and 
the TelemetrySeries itself, in that order. 
* `getSeries(domainObject)`: Get the latest `TelemetrySeries` (as resulted from 
a previous `request(...)` call) available for this domain object.


# Models
Domain object models in Open MCT Web are JavaScript objects describing the 
persistent state of the domain objects they describe. Their contents include a 
mix of commonly understood metadata attributes; attributes which are recognized 
by and/or determine the applicability of specific extensions; and properties 
specific to given types. 

## General Metadata 

Some properties of domain object models have a ubiquitous meaning through Open 
MCT Web and can be utilized directly: 

* `name`: The human-readable name of the domain object.

## Extension-specific Properties 

Other properties of domain object models have specific meaning imposed by other 
extensions within the Open MCT Web platform. 

### Capability-specific Properties 

Some properties either trigger the presence/absence of certain capabilities, or 
are managed by specific capabilities:

* `composition`: An array of domain object identifiers that represents the 
contents of this domain object (e.g. as will appear in the tree hierarchy.) 
Understood by the composition capability; the presence or absence of this 
property determines the presence or absence of that capability. 
* `modified`: The timestamp (in milliseconds since the UNIX epoch) of the last 
modification made to this domain object. Managed by the mutation capability. 
* `persisted`: The timestamp (in milliseconds since the UNIX epoch) of the last 
time when changes to this domain object were persisted. Managed by the 
 persistence capability. 
* `relationships`: An object containing key-value pairs, where keys are symbolic 
identifiers for relationship types, and values are arrays of domain object 
identifiers. Used by the relationship capability; the presence or absence of 
this property determines the presence or absence of that capability. 
* `telemetry`: An object which serves as a template for telemetry requests 
associated with this domain object (e.g. specifying `source` and `key`; see 
Telemetry Requests under Core API.) Used by the telemetry capability; the 
presence or absence of this property determines the presence or absence of that 
capability.
* `type`: A string identifying the type of this domain object. Used by the `type` 
capability. 
 
### View Configurations 

Persistent configurations for specific views of domain objects are stored in the 
domain object model under the property configurations . This is an object 
containing key-value pairs, where keys identify the view, and values are objects 
containing view-specific (and view-managed) configuration properties. 
 
## Modifying Models 
When interacting with a domain object's model, it is possible to make 
modifications to it directly. __Don't!__  These changes may not be properly detected 
by the platform, meaning that other representations of the domain object may not 
be updated, changes may not be saved at the expected times, and generally, that 
unexpected behavior may occur. Instead, use the `mutation` capability. 

# Capabilities 

Dynamic behavior associated with a domain object is expressed as capabilities. A 
capability is a JavaScript object with an interface that is specific to the type 
of capability in use. 

Often, there is a relationship between capabilities and services. For instance, 
there is an action capability and an actionService , and there is a telemetry  
capability as well as a `telemetryService`. Typically, the pattern here is that 
the capability will utilize the service for the specific domain object.  

When interacting with domain objects, it is generally preferable to use a 
capability instead of a service when the option is available. Capability 
interfaces are typically easier to use and/or more powerful in these situations. 
Additionally, this usage provides a more robust substitutability mechanism; for 
instance, one could configure a plugin such that it provided a totally new 
implementation of a given capability which might not invoke the underlying 
service, while user code which interacts with capabilities remains indifferent 
to this detail. 
 
## Action Capability

The `action` capability is present for all domain objects. It allows applicable 
`Action` instances to be retrieved and performed for specific domain objects. 

For example: 
    `domainObject.getCapability("action").perform("navigate"); `
    ...will initiate a navigate action upon the domain object, if an action with 
    key "navigate" is defined. 
   
This capability has the following interface: 
* `getActions(context)`: Get the actions that are applicable in the specified 
action `context`; the capability will fill in the `domainObject` field of this 
context if necessary. If context is specified as a string, they will instead be 
used as the `key` of the action context. Returns an array of `Action` instances.
* `perform(context)`: Perform an action. This will find and perform the first 
matching action available for the specified action context , filling in the 
`domainObject` field as necessary. If `context` is specified as a string, they 
will instead be used as the `key` of the action context. Returns a `Promise` for 
the result of the action that was performed, or `undefined` if no matching action 
was found. 

## Composition Capability

The `composition` capability provides access to domain objects that are 
contained by this domain object. While the `composition` property of a domain 
object's model describes these contents (by their identifiers), the  
`composition` capability provides a means to load the corresponding 
`DomainObject` instances in the same order. The absence of this property in the 
model will result in the absence of this capability in the domain object. 

This capability has the following interface: 

* `invoke()`: Returns a `Promise` for an array of `DomainObject` instances.

## Delegation Capability

The delegation capability is used to communicate the intent of a domain object 
to delegate responsibilities, which would normally handled by other 
capabilities, to the domain objects in its composition. 

This capability has the following interface: 

* `getDelegates(key)`: Returns a Promise for an array of DomainObject instances, 
to which this domain object wishes to delegate the capability with the specified  
key . 
* `invoke(key)`: Alias of getDelegates(key) . 
* `doesDelegate(key)`: Returns true if the domain object does delegate the 
capability with the specified key .  

The platform implementation of the delegation capability inspects the domain 
object's type definition for a property delegates , whose value is an array of 
strings describing which capabilities domain objects of that type wish to 
delegate. If this property is not present, the delegation capability will not be 
present in domain objects of that type.  

## Editor Capability

The editor capability is meant primarily for internal use by Edit mode, and 
helps to manage the behavior associated with exiting _Edit_ mode via _Save_ or 
_Cancel_. Its interface is not intended for general use. However, 
`domainObject.hasCapability(editor)` is a useful way of determining whether or 
not we are looking at an object in _Edit_ mode.
 
## Mutation Capability

The `mutation` capability provides a means by which the contents of a domain 
object's model can be modified. This capability is provided by the platform for 
all domain objects, and has the following interface:

* `mutate(mutator, [timestamp])`: Modify the domain object's model using the 
specified `mutator` function. After changes are made, the `modified` property of 
the model will be updated with the specified `timestamp` if one was provided, 
or with the current system time. 
* `invoke(...)`: Alias of `mutate`.

Changes to domain object models should only be made via the `mutation` 
capability; other platform behavior is likely to break (either by exhibiting 
undesired behavior, or failing to exhibit desired behavior) if models are 
modified by other means.
 
### Mutator Function

The mutator argument above is a function which will receive a cloned copy of the 
domain object's model as a single argument. It may return: 

* A `Promise` in which case the resolved value of the promise will be used to 
determine which of the following forms is used. 
* Boolean `false` in which case the mutation is cancelled. 
* A JavaScript object, in which case this object will be used as the new model 
for this domain object.
* No value (or, equivalently, `undefined`), in which case the cloned copy 
(including any changes made in place by the mutator function) will be used as 
the new domain object model. 

## Persistence Capability

The persistence capability provides a mean for interacting with the underlying 
persistence service which stores this domain object's model. It has the 
following interface:

* `persist()`: Store the local version of this domain object, including any 
changes, to the persistence store. Returns a Promise for a boolean value, which 
will be true when the object was successfully persisted.
* `refresh()`: Replace this domain object's model with the most recent version 
from persistence. Returns a Promise which will resolve when the change has 
completed.
* `getSpace()`: Return the string which identifies the persistence space which 
stores this domain object.

## Relationship Capability

The relationship capability provides a means for accessing other domain objects 
with which this domain object has some typed relationship. It has the following 
interface:

* `listRelationships()`: List all types of relationships exposed by this object. 
Returns an array of strings identifying the types of relationships.
* `getRelatedObjects(relationship)`: Get all domain objects to which this domain 
object has the specified type of relationship, which is a string identifier 
(as above.) Returns a `Promise` for an array of `DomainObject` instances.

The platform implementation of the `relationship` capability is present for domain 
objects which has a `relationships` property in their model, whose value is an 
object containing key-value pairs, where keys are strings identifying 
relationship types, and values are arrays of domain object identifiers.

## Telemetry Capability

The telemetry capability provides a means for accessing telemetry data 
associated with a domain object. It has the following interface:

* `requestData([request])`: Request telemetry data for this specific domain 
object, using telemetry request parameters from the specified request if 
provided. This capability will fill in telemetry request properties as-needed 
for this domain object. Returns a `Promise` for a `TelemetrySeries`.
* `subscribe(callback, [request])`:  Subscribe to telemetry data updates for 
this specific domain object, using telemetry request parameters from the 
specified request if provided. This capability will fill in telemetry request 
properties as-needed for this domain object. The specified callback will be 
invoked with TelemetrySeries instances as they arrive. Returns a function which 
can be invoked to terminate the subscription, or undefined if no subscription 
could be obtained.
* `getMetadata()`: Get metadata associated with this domain object's telemetry.

The platform implementation of the `telemetry` capability is present for domain 
objects which has a `telemetry` property in their model and/or type definition; 
this object will serve as a template for telemetry requests made using this 
object, and will also be returned by `getMetadata()` above.

## Type Capability
The `type` capability exposes information about the domain object's type. It has 
the same interface as `Type`; see Core API.

## View Capability

The `view` capability exposes views which are applicable to a given domain 
object. It has the following interface:

* `invoke()`: Returns an array of extension definitions for views which are 
applicable for this domain object.

# Actions

Actions are reusable processes/behaviors performed by users within the system, 
typically upon domain objects.

## Action Categories

The platform understands the following action categories (specifiable as the 
`category` parameter of an action's extension definition.)

* `contextual`: Appears in context menus.
* `view-control`: Appears in top-right area of view (as buttons) in Browse mode

## Platform Actions
The platform defines certain actions which can be utilized by way of a domain 
object's `action` capability. Unless otherwise specified, these act upon (and 
modify) the object described by the `domainObject` property of the action's 
context.

* `cancel`: Cancel the current editing action (invoked from Edit mode.)
* `compose`: Place an object in another object's composition. The object to be 
added should be provided as the `selectedObject` of the action context.
* `edit`: Start editing an object (enter Edit mode.)
* `fullscreen`: Enter full screen mode.
* `navigate`: Make this object the focus of navigation (e.g. highlight it within 
the tree, display a view of it to the right.)
* `properties`: Show the 'Edit Properties' dialog.
* `remove`: Remove this domain object from its parent's composition. (The 
parent, in this case, is whichever other domain object exposed this object by 
way of its `composition` capability.)
* `save`: Save changes (invoked from Edit mode.)
* `window`: Open this object in a new window.

# Policies

Policies are consulted to determine when certain behavior in Open MCT Web is 
allowed. Policy questions are assigned to certain categories, which broadly 
describe the type of decision being made; within each category, policies have a 
candidate (the thing which may or may not be allowed) and, optionally, a context 
(describing, generally, the context in which the decision is occurring.)

The types of objects passed for 'candidate' and 'context' vary by category; 
these types are documented below.

## Policy Categories

The platform understands the following policy categories (specifiable as the 
`category` parameter of an policy's extension definition.)

* `action`: Determines whether or not a given action is allowable. The candidate 
argument here is an Action; the context is its action context object.
* `composition`: Determines whether or not domain objects of a given type are 
allowed to contain domain objects of another type. The candidate argument here 
is the container's `Type`; the context argument is the `Type` of the object to be 
contained.
* `view`: Determines whether or not a view is applicable for a domain object. 
The candidate argument is the view's extension definition; the context argument 
is the `DomainObject` to be viewed.

# Build-Test-Deploy
Open MCT Web is designed to support a broad variety of build and deployment 
options. The sources can be deployed in the same directory structure used during 
development. A few utilities are included to support development processes.

## Command-line Build
Open MCT Web includes a script for building via command line using Maven 3.0.4 
[https://maven.apache.org/]().
        
Invoking mvn clean install will:

* Check code style using JSLint. The build will fail if JSLint raises any warnings.
* Run the test suite (see below.) The build will fail if any tests fail.
* Populate version info (e.g. commit hash, build time.)
* Produce a web archive (`.war`) artifact in the `target` directory.

The produced artifact contains a subset of the repository's own folder 
hierarchy, omitting tests and example bundles.        

Note that an internet connection is required to run this build, in order to 
download build dependencies.

## Test Suite

Open MCT Web uses Jasmine [http://jasmine.github.io/]() for automated testing. 
The file `test.html` included at the top level of the source repository, can be 
run from the browser to perform tests for all active bundles, as defined in 
`bundle.json`.

To define tests for a bundle:

* Include a directory named `test` within that bundle.
* In the `test` directory, include a file named `suite.json`. This will identify 
which scripts will be tested.
* The file `suite.json` must contain a JSON array of strings, where each string 
is the name of a script to be tested. These names should include any directory 
paths to the script after (but not including) the `src` folder, and should not 
include the file's `.js` extension. (Note that while Open MCT Web's framework 
allows a different name to be chosen for the src directory, the test runner 
does not: This directory must be named `src` for the test runner to find it.)
* For each script to be tested, a corresponding test script should be located in 
the bundle's `test` directory. This should include the suffix Spec at the end of 
the filename (but before the `.js` extension.) This test script should be an AMD 
module which uses the Jasmine API to declare its test behavior. It should 
declare an AMD dependency on the script to be tested, using a relative path.

For example, if writing tests for a bundle at example/foo with two scripts:
* `example/foo/src/controllers/FooController.js`
* `example/foo/src/directives/FooDirective.js`

First, these scripts should be identified in `example/foo/test/suite.json` e.g. 
with contents:`[ "controllers/FooController", "directives/FooDirective" ]`

Then, scripts which describe these tests should be written. For example, test 
`example/foo/test/controllers/FooControllerSpec.js` could look like:

    /*global define,Promise,describe,it,expect,beforeEach*/

    define(
        ["../../src/controllers/FooController"],
        function (FooController) {
            "use strict";
    
    
            describe("The foo controller", function () {
                it("does something", function () {
                    var controller = new FooController();
                    expect(controller.foo()).toEqual("foo");
                });
            });
        }
    );


## Code Coverage

In addition to running tests, the test runner will also capture code coverage 
information using [Blanket.JS](http://blanketjs.org/) and display this at the 
bottom of the screen. Currently, only statement coverage is displayed.

## Deployment
Open MCT Web is built to be flexible in terms of the deployment strategies it 
supports. In order to run in the browser, Open MCT Web needs:

1. HTTP access to sources/resources for the framework, platform, and all active 
bundles.
2. Access to any external services utilized by active bundles. (This means that 
external services need to support HTTP or some other web-accessible interface, 
like WebSockets.)

Any HTTP server capable of serving flat files is sufficient for the first point. 
The command-line build also packages Open MCT Web into a `.war` file for easier 
deployment on containers such as Apache Tomcat.

The second point may be less flexible, as it depends upon the specific services 
to be utilized by Open MCT Web. Because of this, it is often the set of external 
services (and the manner in which they are exposed) that determine how to deploy 
Open MCT Web.

One important constraint to consider in this context is the browser's same 
origin policy. If external services are not on the same apparent host and port 
as the client (from the perspective of the browser) then access may be 
disallowed. There are two workarounds if this occurs:

* Make the external service appear to be on the same host/port, either by 
actually deploying it there, or by proxying requests to it.
* Enable CORS (cross-origin resource sharing) on the external service. This is 
only possible if the external service can be configured to support CORS. Care 
should be exercised if choosing this option to ensure that the chosen 
configuration does not create a security vulnerability.

Examples of deployment strategies (and the conditions under which they make the 
most sense) include:

* If the external services that Open MCT Web will utilize are all running on 
Apache Tomcat [https://tomcat.apache.org/](), then it makes sense to run Open 
MCT Web from the same Tomcat instance as a separate web application. The 
`.war` artifact produced by the command line build facilitates this deployment 
option. (See [https://tomcat.apache.org/tomcat-8.0-doc/deployer-howto.html() for 
general information on deploying in Tomcat.)
* If a variety of external services will be running from a variety of 
hosts/ports, then it may make sense to use a web server that supports proxying, 
such as the Apache HTTP Server [http://httpd.apache.org/](). In this 
configuration, the HTTP server would be configured to proxy (or reverse proxy) 
requests at specific paths to the various external services, while providing 
Open MCT Web as flat files from a different path.
* If a single server component is being developed to handle all server-side 
needs of an Open MCT Web instance, it can make sense to serve Open MCT Web (as 
flat files) from the same component using an embedded HTTP server such as Nancy 
[http://nancyfx.org/]().
* If no external services are needed (or if the 'external services' will just 
be generating flat files to read) it makes sense to utilize a lightweight flat 
file HTTP server such as Lighttpd [http://www.lighttpd.net/](). In this 
configuration, Open MCT Web sources/resources would be placed at one path, while 
the files generated by the external service are placed at another path.
* If all external services support CORS, it may make sense to have an HTTP 
server that is solely responsible for making Open MCT Web sources/resources 
available, and to have Open MCT Web contact these external services directly. 
Again, lightweight HTTP servers such as Lighttpd [http://www.lighttpd.net/]() 
are useful in this circumstance. The downside of this option is that additional 
configuration effort is required, both to enable CORS on the external services, 
and to ensure that Open MCT Web can correctly locate these services.

Another important consideration is authentication. By design, Open MCT Web does 
not handle user authentication. Instead, this should typically be treated as a 
deployment-time concern, where authentication is handled by the HTTP server 
which provides Open MCT Web, or an external access management system.

### Configuration
In most of the deployment options above, some level of configuration is likely 
to be needed or desirable to make sure that bundles can reach the external 
services they need to reach. Most commonly this means providing the path or URL 
to an external service.

Configurable parameters within Open MCT Web are specified via constants 
(literally, as extensions of the `constants` category) and accessed via 
dependency injection by the scripts which need them. Reasonable defaults for 
these constants are provided in the bundle where they are used. Plugins are 
encouraged to follow the same pattern.

Constants may be specified in any bundle; if multiple constants are specified 
with the same `key` the highest-priority one will be used. This allows default 
values to be overridden by specifying constants with higher priority.

This permits at least three configuration approaches:

* Modify the constants defined in their original bundles when deploying. This is 
generally undesirable due to the amount of manual work required and potential 
for error, but is viable if there are a small number of constants to change.
* Add a separate configuration bundle which overrides the values of these 
constants. This is particularly appropriate when multiple configurations (e.g. 
development, test, production) need to be managed easily; these can be swapped 
quickly by changing the set of active bundles in bundles.json.
* Deploy Open MCT Web and its external services in such a fashion that the 
default paths to reach external services are all correct.

### Configuration Constants


The following configuration constants are recognized by Open MCT Web bundles:
* Common UI elements - `platform/commonUI/general`
    * `THEME`: A string identifying the current theme symbolically. Individual
    stylesheets (the `stylesheets` extension category) may specify an optional
    `theme` property which will be matched against this before inclusion.
* CouchDB adapter - `platform/persistence/couch`
    * `COUCHDB_PATH`: URL or path to the CouchDB database to be used for domain 
    object persistence. Should not include a trailing slash.
* ElasticSearch adapter - `platform/persistence/elastic`
    * `ELASTIC_ROOT`: URL or path to the ElasticSearch instance to be used for 
    domain object persistence. Should not include a trailing slash.
    * `ELASTIC_PATH`: Path relative to the ElasticSearch instance where domain 
    object models should be persisted. Should take the form `<index>/<type>`.