Skip to content

Latest commit

 

History

History
1374 lines (1076 loc) · 59.4 KB

API.md

File metadata and controls

1374 lines (1076 loc) · 59.4 KB

Table of Contents

Developing Applications With Open MCT

Scope and purpose of this document

This document is intended to serve as a reference for developing an application based on Open MCT. It will provide details of the API functions necessary to extend the Open MCT platform meet common use cases such as integrating with a telemetry source.

The best place to start is with the Open MCT Tutorials. These will walk you through the process of getting up and running with Open MCT, as well as addressing some common developer use cases.

Building From Source

The latest version of Open MCT is available from our GitHub repository. If you have git, and node installed, you can build Open MCT with the commands

git clone https://github.com/nasa/openmct.git
cd openmct
npm install

These commands will fetch the Open MCT source from our GitHub repository, and build a minified version that can be included in your application. The output of the build process is placed in a dist folder under the openmct source directory, which can be copied out to another location as needed. The contents of this folder will include a minified javascript file named openmct.js as well as assets such as html, css, and images necessary for the UI.

Starting an Open MCT application

Warning

Open MCT provides a development server via webpack-dev-server (npm start). This should be used for development purposes only and should never be deployed to a production environment.

To start a minimally functional Open MCT application, it is necessary to include the Open MCT distributable, enable some basic plugins, and bootstrap the application. The tutorials walk through the process of getting Open MCT up and running from scratch, but provided below is a minimal HTML template that includes Open MCT, installs some basic plugins, and bootstraps the application. It assumes that Open MCT is installed under an openmct subdirectory, as described in Building From Source.

This approach includes openmct using a simple script tag, resulting in a global variable named openmct. This openmct object is used subsequently to make API calls.

Open MCT is packaged as a UMD (Universal Module Definition) module, so common script loaders are also supported.

<!DOCTYPE html>
<html>
<head>
    <title>Open MCT</title>
    <script src="dist/openmct.js"></script>
</head>
<body>
    <script>
        openmct.install(openmct.plugins.LocalStorage());
        openmct.install(openmct.plugins.MyItems());
        openmct.install(openmct.plugins.UTCTimeSystem());
        openmct.start();
    </script>
</body>
</html>

The Open MCT library included above requires certain assets such as html templates, images, and css. If you installed Open MCT from GitHub as described in the section on Building from Source then these assets will have been downloaded along with the Open MCT javascript library.

There are some plugins bundled with the application that provide UI, persistence, and other default configuration which are necessary to be able to do anything with the application initially. Any of these plugins can, in principle, be replaced with a custom plugin. The included plugins are documented in the Included Plugins section.

Types

The Open MCT library includes its own TypeScript declaration files which can be used to provide code hints and typechecking in your own Open MCT application.

Open MCT's type declarations are generated via tsc from JSDoc-style comment blocks. For more information on this, check out TypeScript's documentation.

Using Types

In order to use Open MCT's provided types in your own application, create a jsconfig.js at the root of your project with this minimal configuration:

{
  "compilerOptions": {
    "baseUrl": "./",
      "target": "es6",
      "checkJs": true,
      "moduleResolution": "node",
      "paths": {
        "openmct": ["node_modules/openmct/dist/openmct.d.ts"]
      }
  }
}

Then, simply import and use openmct in your application:

import openmct from "openmct";

Limitations

The effort to add types for Open MCT's public API is ongoing, and the provided type declarations may be incomplete.

If you would like to contribute types to Open MCT, please check out TypeScript's documentation on generating type declarations from JSDoc-style comment blocks. Then read through our contributing guide and open a PR!

Plugins

Defining and Installing a New Plugin

openmct.install(function install(openmctAPI) {
    // Do things here
    // ...
});

New plugins are installed in Open MCT by calling openmct.install, and providing a plugin installation function. This function will be invoked on application startup with one parameter - the openmct API object. A common approach used in the Open MCT codebase is to define a plugin as a function that returns this installation function. This allows configuration to be specified when the plugin is included.

eg.

openmct.install(openmct.plugins.Elasticsearch("http://localhost:8002/openmct"));

This approach can be seen in all of the plugins provided with Open MCT.

Domain Objects and Identifiers

Domain Objects are the basic entities that represent domain knowledge in Open MCT. The temperature sensor on a solar panel, an overlay plot comparing the results of all temperature sensors, the command dictionary for a spacecraft, the individual commands in that dictionary, the "My Items" folder: All of these things are domain objects.

A Domain Object is simply a javascript object with some standard attributes.
An example of a Domain Object is the "My Items" object which is a folder in which a user can persist any objects that they create. The My Items object looks like this:

{
    identifier: {
        namespace: ""
        key: "mine"
    }
    name:"My Items",
    type:"folder",
    location:"ROOT",
    composition: []
}

Object Attributes

The main attributes to note are the identifier, and type attributes.

  • identifier: A composite key that provides a universally unique identifier for this object. The namespace and key are used to identify the object. The key must be unique within the namespace.
  • type: All objects in Open MCT have a type. Types allow you to form an ontology of knowledge and provide an abstraction for grouping, visualizing, and interpreting data. Details on how to define a new object type are provided below.

Open MCT uses a number of builtin types. Typically you are going to want to define your own when extending Open MCT.

Domain Object Types

Custom types may be registered via the addType function on the Open MCT Type registry.

eg.

openmct.types.addType('example.my-type', {
    name: "My Type",
    description: "This is a type that I added!",
    creatable: true
});

The addType function accepts two arguments:

  • A string key identifying the type. This key is used when specifying a type for an object. We recommend prefixing your types with a namespace to avoid conflicts with other plugins.
  • An object type specification. An object type definition supports the following attributes
    • name: a string naming this object type
    • description: a string specifying a longer-form description of this type
    • initialize: a function which initializes the model for new domain objects of this type. This can be used for setting default values on an object when it is instantiated.
    • creatable: A boolean indicating whether users should be allowed to create this type (default: false). This will determine whether the type appears in the Create menu.
    • cssClass: A string specifying a CSS class to apply to each representation of this object. This is used for specifying an icon to appear next to each object of this type.

The Open MCT Tutorials provide a step-by-step examples of writing code for Open MCT that includes a section on defining a new object type.

Root Objects

In many cases, you'd like a certain object (or a certain hierarchy of objects) to be accessible from the top level of the application (the tree on the left-hand side of Open MCT.) For example, it is typical to expose a telemetry dictionary as a hierarchy of telemetry-providing domain objects in this fashion.

To do so, use the addRoot method of the object API.

eg.

openmct.objects.addRoot({
    namespace: "example.namespace",
    key: "my-key"
},
openmct.priority.HIGH);

The addRoot function takes a two arguments, the first can be an object identifier for a root level object, or an array of identifiers for root level objects, or a function that returns a promise for an identifier or an array of root level objects, the second is a priority or numeric value.

When using the getAll method of the object API, they will be returned in order of priority.

eg.

openmct.objects.addRoot(identifier, openmct.priority.LOW); // low = -1000, will appear last in composition or tree
openmct.objects.addRoot(otherIdentifier, openmct.priority.HIGH); // high = 1000, will appear first in composition or tree

Root objects are loaded just like any other objects, i.e. via an object provider.

Object Providers

An Object Provider is used to build Domain Objects, typically retrieved from some source such as a persistence store or telemetry dictionary. In order to integrate telemetry from a new source an object provider will need to be created that can build objects representing telemetry points exposed by the telemetry source. The API call to define a new object provider is fairly straightforward. Here's a very simple example:

openmct.objects.addProvider('example.namespace', {
    get: function (identifier) {
        return Promise.resolve({
            identifier: identifier,
            name: 'Example Object',
            type: 'example-object-type'
        });
    }
});

The addProvider function takes two arguments:

  • namespace: A string representing the namespace that this object provider will provide objects for.
  • provider: An object with a single function, get. This function accepts an Identifier for the object to be provided. It is expected that the get function will return a Promise that resolves with the object being requested.

In future, object providers will support other methods to enable other operations with persistence stores, such as creating, updating, and deleting objects.

Composition Providers

The composition of a domain object is the list of objects it contains, as shown (for example) in the tree for browsing. Open MCT provides a default solution for composition, but there may be cases where you want to provide the composition of a certain object (or type of object) dynamically.

Adding Composition Providers

You may want to populate a hierarchy under a custom root-level object based on the contents of a telemetry dictionary. To do this, you can add a new Composition Provider:

openmct.composition.addProvider({
    appliesTo: function (domainObject) {
        return domainObject.type === 'example.my-type';
    },
    load: function (domainObject) {
        return Promise.resolve(myDomainObjects);
    }
});

The addProvider function accepts a Composition Provider object as its sole argument. A Composition Provider is a javascript object exposing two functions:

  • appliesTo: A function that accepts a domainObject argument, and returns a boolean value indicating whether this composition provider applies to the given object.
  • load: A function that accepts a domainObject as an argument, and returns a Promise that resolves with an array of Identifier. These identifiers will be used to fetch Domain Objects from an Object Provider

Default Composition Provider

The default composition provider applies to any domain object with a composition property. The value of composition should be an array of identifiers, e.g.:

var domainObject = {
    name: "My Object",
    type: 'folder',
    composition: [
        {
            id: '412229c3-922c-444b-8624-736d85516247',
            namespace: 'foo'
        },
        {
            key: 'd6e0ce02-5b85-4e55-8006-a8a505b64c75',
            namespace: 'foo'
        }
    ]
};

Telemetry API

The Open MCT telemetry API provides two main sets of interfaces-- one for integrating telemetry data into Open MCT, and another for developing Open MCT visualization plugins utilizing the telemetry API.

The APIs for visualization plugins are still a work in progress and docs may change at any time. However, the APIs for integrating telemetry metadata into Open MCT are stable and documentation is included below.

Integrating Telemetry Sources

There are two main tasks for integrating telemetry sources-- describing telemetry objects with relevant metadata, and then providing telemetry data for those objects. You'll use an Object Provider to provide objects with the necessary Telemetry Metadata, and then register a Telemetry Provider to retrieve telemetry data for those objects. Alternatively, you can register a telemetry metadata provider to provide the necessary telemetry metadata.

For a step-by-step guide to building a telemetry adapter, please see the Open MCT Tutorials.

Telemetry Metadata

A telemetry object is a domain object with a telemetry property. To take an example from the tutorial, here is the telemetry object for the "fuel" measurement of the spacecraft:

{
    "identifier": {
        "namespace": "example.taxonomy",
        "key": "prop.fuel"
    },
    "name": "Fuel",
    "type": "example.telemetry",
    "telemetry": {
        "values": [
            {
                "key": "value",
                "name": "Value",
                "unit": "kilograms",
                "format": "float",
                "min": 0,
                "max": 100,
                "hints": {
                    "range": 1
                }
            },
            {
                "key": "utc",
                "source": "timestamp",
                "name": "Timestamp",
                "format": "utc",
                "hints": {
                    "domain": 1
                }
            }
        ]
    }
}

The most important part of the telemetry metadata is the values property-- this describes the attributes of telemetry datums (objects) that a telemetry provider returns. These descriptions must be provided for telemetry views to work properly.

Values

telemetry.values is an array of value description objects, which have the following fields:

attribute type flags notes
key string required unique identifier for this field.
hints object required Hints allow views to intelligently select relevant attributes for display, and are required for most views to function. See section on "Value Hints" below.
name string optional a human readable label for this field. If omitted, defaults to key.
source string optional identifies the property of a datum where this value is stored. If omitted, defaults to key.
format string optional a specific format identifier, mapping to a formatter. If omitted, uses a default formatter. For enumerations, use enum. For timestamps, use utc if you are using utc dates, otherwise use a key mapping to your custom date format.
unit string optional the unit of this value, e.g. km, seconds, parsecs
min number optional the minimum possible value of this measurement. Will be used by plots, gauges, etc to automatically set a min value.
max number optional the maximum possible value of this measurement. Will be used by plots, gauges, etc to automatically set a max value.
enumerations array optional for objects where format is "enum", this array tracks all possible enumerations of the value. Each entry in this array is an object, with a value property that is the numerical value of the enumeration, and a string property that is the text value of the enumeration. ex: {"value": 0, "string": "OFF"}. If you use an enumerations array, min and max will be set automatically for you.
Value Hints

Each telemetry value description has an object defining hints. Keys in this object represent the hint itself, and the value represents the weight of that hint. A lower weight means the hint has a higher priority. For example, multiple values could be hinted for use as the y-axis of a plot (raw, engineering), but the highest priority would be the default choice. Likewise, a table will use hints to determine the default order of columns.

Known hints:

  • domain: Values with a domain hint will be used for the x-axis of a plot, and tables will render columns for these values first.
  • range: Values with a range hint will be used as the y-axis on a plot, and tables will render columns for these values after the domain values.
  • image: Indicates that the value may be interpreted as the URL to an image file, in which case appropriate views will be made available.
  • imageDownloadName: Indicates that the value may be interpreted as the name of the image file.
The Time Conductor and Telemetry

Open MCT provides a number of ways to pivot through data and link data via time. The Time Conductor helps synchronize multiple views around the same time.

In order for the time conductor to work, there will always be an active "time system". All telemetry metadata must have a telemetry value with a key that matches the key of the active time system. You can use the source attribute on the value metadata to remap this to a different field in the telemetry datum-- especially useful if you are working with disparate datasources that have different field mappings.

Telemetry Providers

Telemetry providers are responsible for providing historical and real-time telemetry data for telemetry objects. Each telemetry provider determines which objects it can provide telemetry for, and then must implement methods to provide telemetry for those objects.

A telemetry provider is a javascript object with up to four methods:

  • supportsSubscribe(domainObject, callback, options) optional. Must be implemented to provide realtime telemetry. Should return true if the provider supports subscriptions for the given domain object (and request options).
  • subscribe(domainObject, callback, options) required if supportsSubscribe is implemented. Establish a subscription for realtime data for the given domain object. Should invoke callback with a single telemetry datum every time data is received. Must return an unsubscribe function. Multiple views can subscribe to the same telemetry object, so it should always return a new unsubscribe function.
  • supportsRequest(domainObject, options) optional. Must be implemented to provide historical telemetry. Should return true if the provider supports historical requests for the given domain object.
  • request(domainObject, options) required if supportsRequest is implemented. Must return a promise for an array of telemetry datums that fulfills the request. The options argument will include a start, end, and domain attribute representing the query bounds. See Telemetry Requests and Responses for more info on how to respond to requests.
  • supportsMetadata(domainObject) optional. Implement and return true for objects that you want to provide dynamic metadata for.
  • getMetadata(domainObject) required if supportsMetadata is implemented. Must return a valid telemetry metadata definition that includes at least one valueMetadata definition.
  • supportsLimits(domainObject) optional. Implement and return true for domain objects that you want to provide a limit evaluator for.
  • getLimitEvaluator(domainObject) required if supportsLimits is implemented. Must return a valid LimitEvaluator for a given domain object.

Telemetry providers are registered by calling openmct.telemetry.addProvider(provider), e.g.

openmct.telemetry.addProvider({
    supportsRequest: function (domainObject, options) { /*...*/ },
    request: function (domainObject, options) { /*...*/ },
})

Note: it is not required to implement all of the methods on every provider. Depending on the complexity of your implementation, it may be helpful to instantiate and register your realtime, historical, and metadata providers separately.

Telemetry Requests and Responses

Telemetry requests support time bounded queries. A call to a Telemetry Provider's request function will include an options argument. These are simply javascript objects with attributes for the request parameters. An example of a telemetry request object with a start and end time is included below:

{
    start: 1487981997240,
    end: 1487982897240,
    domain: 'utc'
}

In this case, the domain is the currently selected time-system, and the start and end dates are valid dates in that time system.

A telemetry provider's request method should return a promise for an array of telemetry datums. These datums must be sorted by domain in ascending order.

The telemetry provider's request method will also return an object signal with an aborted property with a value true if the request has been aborted by user navigation. This can be used to trigger actions when a request has been aborted.

Request Strategies draft

To improve performance views may request a certain strategy for data reduction. These are intended to improve visualization performance by reducing the amount of data needed to be sent to the client. These strategies will be indicated by additional parameters in the request options. You may choose to handle them or ignore them.

Note: these strategies are currently being tested in core plugins and may change based on developer feedback.

latest request strategy

This request is a "depth based" strategy. When a view is only capable of displaying a single value (or perhaps the last ten values), then it can use the latest request strategy with a size parameter that specifies the number of results it desires. The size parameter is a hint; views must not assume the response will have the exact number of results requested.

example:

{
    start: 1487981997240,
    end: 1487982897240,
    domain: 'utc',
    strategy: 'latest',
    size: 1
}

This strategy says "I want the latest data point in this time range". A provider which recognizes this request should return only one value-- the latest-- in the requested time range. Depending on your back-end implementation, performing these queries in bulk can be a large performance increase. These are generally issued by views that are only capable of displaying a single value and only need to show the latest value.

minmax request strategy

example:

{
    start: 1487981997240,
    end: 1487982897240,
    domain: 'utc',
    strategy: 'minmax',
    size: 720
}

MinMax queries are issued by plots, and may be issued by other types as well. The aim is to reduce the amount of data returned but still faithfully represent the full extent of the data. In order to do this, the view calculates the maximum data resolution it can display (i.e. the number of horizontal pixels in a plot) and sends that as the size. The response should include at least one minimum and one maximum value per point of resolution.

Telemetry Formats

Telemetry format objects define how to interpret and display telemetry data. They have a simple structure, provided here as a TypeScript interface:

interface Formatter {
    key: string; // A string that uniquely identifies this formatter.

    format: (
        value: any, // The raw telemetry value in its native type.
        minValue?: number, // An optional argument specifying the minimum displayed value.
        maxValue?: number, // An optional argument specifying the maximum displayed value.
        count?: number // An optional argument specifying the number of displayed values.
    ) => string; // Returns a human-readable string representation of the provided value.

    parse: (
        value: string | any // A string representation of a telemetry value or an already-parsed value.
    ) => any; // Returns the value in its native type. This function should be idempotent.

    validate: (value: string) => boolean; // Takes a string representation of a telemetry value and returns a boolean indicating whether the provided string can be parsed.
}
Built-in Formats

Open MCT on its own defines a handful of built-in formats:

Number Format (default):

Applied to data with format: 'number'

valueMetadata = {
    format: 'number'
    // ...
};
interface NumberFormatter extends Formatter {
    parse: (x: any) => number;
    format: (x: number) => string;
    validate: (value: any) => boolean;
}
String Format:

Applied to data with format: 'string'

valueMetadata = {
    format: 'string'
    // ...
};
interface StringFormatter extends Formatter {
    parse: (value: any) => string;
    format: (value: string) => string;
    validate: (value: any) => boolean;
}
Enum Format:

Applied to data with format: 'enum'

valueMetadata = {
    format: 'enum',
    enumerations: [
    {
        value: 1,
        string: 'APPLE'
    }, 
    {
        value: 2,
        string: 'PEAR',
    },
    {
        value: 3,
        string: 'ORANGE'
    }]
    // ...
};

Creates a two-way mapping between enum string and value to be used in the parse and format methods. Ex:

  • formatter.parse('APPLE') === 1;
  • formatter.format(1) === 'APPLE';
interface EnumFormatter extends Formatter {
    parse: (value: string) => string;
    format: (value: number) => string;
    validate: (value: any) => boolean;
}
Registering Formats

Formats implement the following interface (provided here as TypeScript for simplicity):

Formats are registered with the Telemetry API using the addFormat function. eg.

openmct.telemetry.addFormat({
    key: 'number-to-string',
    format: function (number) {
        return number + '';
    },
    parse: function (text) {
        return Number(text);
    },
    validate: function (text) {
        return !isNaN(text);
    }
});

Telemetry Data

A single telemetry point is considered a Datum, and is represented by a standard javascript object. Realtime subscriptions (obtained via subscribe) will invoke the supplied callback once for each telemetry datum recieved. Telemetry requests (obtained via request) will return a promise for an array of telemetry datums.

Telemetry Datums

A telemetry datum is a simple javascript object, e.g.:

{
    "timestamp": 1491267051538,
    "value": 77,
    "id": "prop.fuel"
}

The key-value pairs of this object are described by the telemetry metadata of a domain object, as discussed in the Telemetry Metadata section.

Limit Evaluators draft

Limit evaluators allow a telemetry integrator to define which limits exist for a telemetry endpoint and how limits should be applied to telemetry from a given domain object.

A limit evaluator can implement the evalute method which is used to define how limits should be applied to telemetry and the getLimits method which is used to specify what the limit values are for different limit levels.

Limit levels can be mapped to one of 5 colors for visualization: purple, red, orange, yellow and cyan.

For an example of a limit evaluator, take a look at examples/generator/SinewaveLimitProvider.js.

Telemetry Consumer APIs draft

The APIs for requesting telemetry from Open MCT -- e.g. for use in custom views -- are currently in draft state and are being revised. If you'd like to experiment with them before they are finalized, please contact the team via the contact-us link on our website.

Time API

Open MCT provides API for managing the temporal state of the application. Central to this is the concept of "time bounds". Views in Open MCT will typically show telemetry data for some prescribed date range, and the Time API provides a way to centrally manage these bounds.

The Time API exposes a number of methods for querying and setting the temporal state of the application, and emits events to inform listeners when the state changes.

Because the data displayed tends to be time domain data, Open MCT must always have at least one time system installed and activated. When you download Open MCT, it will be pre-configured to use the UTC time system, which is installed and activated, along with other default plugins, in index.html. Installing and activating a time system is simple, and is covered in the next section.

Time Systems and Bounds

Defining and Registering Time Systems

The time bounds of an Open MCT application are defined as numbers, and a Time System gives meaning and context to these numbers so that they can be correctly interpreted. Time Systems are JavaScript objects that provide some information about the current time reference frame. An example of defining and registering a new time system is given below:

openmct.time.addTimeSystem({
    key: 'utc',
    name: 'UTC Time',
    cssClass = 'icon-clock',
    timeFormat = 'utc',
    durationFormat = 'duration',
    isUTCBased = true
});

The example above defines a new utc based time system. In fact, this time system is configured and activated by default from index.html in the default installation of Open MCT if you download the source from GitHub. Some details of each of the required properties is provided below.

  • key: A string that uniquely identifies this time system.
  • name: A string providing a brief human readable label. If the Time Conductor plugin is enabled, this name will identify the time system in a dropdown menu.
  • cssClass: A class name string that will be applied to the time system when it appears in the UI. This will be used to represent the time system with an icon. There are a number of built-in icon classes available in Open MCT, or a custom class can be used here.
  • timeFormat: A string corresponding to the key of a registered telemetry time format. The format will be used for displaying discrete timestamps from telemetry streams when this time system is activated. If the UTCTimeSystem is enabled, then the utc format can be used if this is a utc-based time system
  • durationFormat: A string corresponding to the key of a registered telemetry time format. The format will be used for displaying time ranges, for example 00:15:00 might be used to represent a time period of fifteen minutes. These are used by the Time Conductor plugin to specify relative time offsets. If the UTCTimeSystem is enabled, then the duration format can be used if this is a utc-based time system
  • isUTCBased: A boolean that defines whether this time system represents numbers in UTC terrestrial time.

Getting and Setting the Active Time System

Once registered, a time system can be activated by calling setTimeSystem with the timeSystem key or an instance of the time system. You can also specify valid bounds for the timeSystem.

openmct.time.setTimeSystem('utc', bounds);

The current time system can be retrieved as well by calling getTimeSystem.

openmct.time.getTimeSystem();

A time system can be immediately activated after registration:

openmct.time.addTimeSystem(utcTimeSystem);
openmct.time.setTimeSystem(utcTimeSystem, bounds);

Setting the active time system will trigger a 'timeSystemChanged' event. If you supplied bounds, a 'boundsChanged' event will be triggered afterwards with your newly supplied bounds.

⚠️ Deprecated

  • The method timeSystem() is deprecated. Please use getTimeSystem() and setTimeSystem() as a replacement.

Time Bounds

The TimeAPI provides a getter and setter for querying and setting time bounds. Time bounds are simply an object with a start and an end end attribute.

  • start: A number representing a moment in time in the active Time System. This will be used as the beginning of the time period displayed by time-responsive telemetry views.
  • end: A number representing a moment in time in the active Time System. This will be used as the end of the time period displayed by time-responsive telemetry views.

New bounds can be set system wide by calling setBounds with bounds.

const ONE_HOUR = 60 * 60 * 1000;
let now = Date.now();
openmct.time.setBounds({start: now - ONE_HOUR, now);

Calling getBounds will return the current application-wide time bounds.

openmct.time.getBounds();

To respond to bounds change events, listen for the 'boundsChanged' event.

⚠️ Deprecated

  • The method bounds() is deprecated and will be removed in a future release. Please use getBounds() and setBounds() as a replacement.

Clocks

The Time API requires a clock source which will cause the bounds to be updated automatically whenever the clock source "ticks". A clock is simply an object that supports registration of listeners and periodically invokes its listeners with a number. Open MCT supports registration of new clock sources that tick on almost anything. A tick occurs when the clock invokes callback functions registered by its listeners with a new time value.

An example of a clock source is the LocalClock which emits the current time in UTC every 100ms. Clocks can tick on anything. For example, a clock could be defined to provide the timestamp of any new data received via a telemetry subscription. This would have the effect of advancing the bounds of views automatically whenever data is received. A clock could also be defined to tick on some remote timing source.

The values provided by clocks are simple numbers, which are interpreted in the context of the active Time System.

Defining and registering clocks

A clock is an object that defines certain required metadata and functions:

  • key: A string uniquely identifying this clock. This can be used later to reference the clock in places such as the Time Conductor configuration
  • cssClass: A string identifying a CSS class to apply to this clock when it's displayed in the UI. This will be used to represent the time system with an icon. There are a number of built-in icon classes available in Open MCT, or a custom class can be used here.
  • name: A string providing a human-readable identifier for the clock source. This will be displayed in the clock selector menu in the Time Conductor UI component, if active.
  • description: An optional string providing a longer description of the clock. The description will be visible in the clock selection menu in the Time Conductor plugin.
  • on: A function supporting registration of a new callback that will be invoked when the clock next ticks. It will be invoked with two arguments:
    • eventName: A string specifying the event to listen on. For now, clocks support one event - tick.
    • callback: A function that will be invoked when this clock ticks. The function must be invoked with one parameter - a number representing a valid time in the current time system.
  • off: A function that allows deregistration of a tick listener. It accepts the same arguments as on.
  • currentValue: A function that returns a number representing a point in time in the active time system. It should be the last value provided by a tick, or some default value if no ticking has yet occurred.

A new clock can be registered using the addClock function exposed by the Time API:

var someClock = {
    key: 'someClock',
    cssClass: 'icon-clock',
    name: 'Some clock',
    description: "Presumably does something useful",
    on: function (event, callback) {
        // Some function that registers listeners, and updates them on a tick
    },
    off: function (event, callback) {
        // Some function that unregisters listeners.
    },
    currentValue: function () {
        // A function that returns the last ticked value for the clock
    }
}

openmct.time.addClock(someClock);

An example clock implementation is provided in the form of the LocalClock

Getting and setting active clock

Once registered a clock can be activated by calling the setClock function on the Time API passing in the key or instance of a registered clock. Only one clock may be active at once, so activating a clock will deactivate any currently active clock and start the new clock. clockOffsets must be specified when changing a clock.

Setting the clock triggers a 'clockChanged' event, followed by a 'clockOffsetsChanged' event, and then a 'boundsChanged' event as the offsets are applied to the clock's currentValue().

openmct.time.setClock(someClock, clockOffsets);

Upon being activated, the time API will listen for tick events on the clock by calling clock.on.

The currently active clock can be retrieved by calling getClock.

openmct.time.getClock();

⚠️ Deprecated

  • The method clock() is deprecated and will be removed in a future release. Please use getClock() and setClock() as a replacement.

⚠️ [DEPRECATED] Stopping an active clock

As of July 2023, this method will be deprecated. Open MCT will always have a ticking clock.

The stopClock method can be used to stop an active clock, and to clear it. It will stop the clock from ticking, and set the active clock to undefined.

openmct.time.stopClock();

⚠️ Deprecated

  • The method stopClock() is deprecated and will be removed in a future release.

Clock Offsets

When in Real-time mode, the time bounds of the application will be updated automatically each time the clock "ticks". The bounds are calculated based on the current value provided by the active clock (via its tick event, or its currentValue() method).

Unlike bounds, which represent absolute time values, clock offsets represent relative time spans. Offsets are defined as an object with two properties:

  • start: A number that must be < 0 and which is used to calculate the start bounds on each clock tick. The start offset will be calculated relative to the value provided by a clock's tick callback, or its currentValue() function.
  • end: A number that must be >= 0 and which is used to calculate the end bounds on each clock tick.

The setClockOffsets function can be used to get or set clock offsets. For example, to show the last fifteen minutes in a ms-based time system:

var FIFTEEN_MINUTES = 15 * 60 * 1000;

openmct.time.setClockOffsets({
    start: -FIFTEEN_MINUTES,
    end: 0
})

The getClockOffsets method will return the currently set clock offsets.

openmct.time.getClockOffsets()

Note: Setting the clock offsets will trigger an immediate bounds change, as new bounds will be calculated based on the currentValue() of the active clock. Clock offsets are only relevant when in Real-time mode.

⚠️ Deprecated

  • The method clockOffsets() is deprecated and will be removed in a future release. Please use getClockOffsets() and setClockOffsets() as a replacement.

Time Modes

There are two time modes in Open MCT, "Fixed" and "Real-time". In Real-time mode the time bounds of the application will be updated automatically each time the clock "ticks". The bounds are calculated based on the current value provided by the active clock. In Fixed mode, the time bounds are set for a specified time range. When Open MCT is first initialized, it will be in Real-time mode.

The setMode method can be used to set the current time mode. It accepts a mode argument, 'realtime' or 'fixed' and it also accepts an optional offsets/bounds argument dependent on the current mode.

openmct.time.setMode('fixed');
openmct.time.setMode('fixed', bounds); // with optional bounds

or

openmct.time.setMode('realtime');
openmct.time.setMode('realtime', offsets); // with optional offsets

The getMode method will return the current time mode, either 'realtime' or 'fixed'.

openmct.time.getMode();

Time Mode Helper Methods

There are two methods available to determine the current time mode in Open MCT programmatically, isRealTime and isFixed. Each one will return a boolean value based on the current mode.

if (openmct.time.isRealTime()) {
  // do real-time stuff
}
if (openmct.time.isFixed()) {
  // do fixed-time stuff
}

Time Events

The Time API is a standard event emitter; you can register callbacks for events using the on method and remove callbacks for events with the off method.

For example:

openmct.time.on('boundsChanged', function callback (newBounds, tick) {
    // Do something with new bounds
});

List of Time Events

The events emitted by the Time API are:

  • boundsChanged: emitted whenever the bounds change. The callback will be invoked with two arguments:
    • bounds: A bounds bounds object representing a new time period bound by the specified start and send times.
    • tick: A boolean indicating whether or not this bounds change is due to a "tick" from a clock source. This information can be useful when determining a strategy for fetching telemetry data in response to a bounds change event. For example, if the bounds change was automatic, and is due to a tick then it's unlikely that you would need to perform a historical data query. It should be sufficient to just show any new telemetry received via subscription since the last tick, and optionally to discard any older data that now falls outside of the currently set bounds. If tick is false,then the bounds change was not due to an automatic tick, and a query for historical data may be necessary, depending on your data caching strategy, and how significantly the start bound has changed.
  • timeSystemChanged: emitted whenever the active time system changes. The callback will be invoked with a single argument:
  • clockChanged: emitted whenever the clock changes. The callback will be invoked with a single argument:
    • clock: The newly active clock, or undefined if an active clock has been deactivated.
  • clockOffsetsChanged: emitted whenever the active clock offsets change. The callback will be invoked with a single argument:
  • modeChanged: emitted whenever the time mode changed. The callback will be invoked with one argument:
    • mode: A string representation of the current time mode, either 'realtime' or 'fixed'.

⚠️ Deprecated Events (These will be removed in a future release):

  • boundsboundsChanged
  • timeSystemtimeSystemChanged
  • clockclockChanged
  • clockOffsetsclockOffsetsChanged

The Time Conductor

The Time Conductor provides a user interface for managing time bounds in Open MCT. It allows a user to select from configured time systems and clocks, and to set bounds and clock offsets.

If activated, the time conductor must be provided with configuration options, detailed below.

Time Conductor Configuration

The time conductor is configured by specifying the options that will be available to the user from the menus in the time conductor. These will determine the clocks available from the conductor, the time systems available for each clock, and some default bounds and clock offsets for each combination of clock and time system. By default, the conductor always supports a fixed mode where no clock is active. To specify configuration for fixed mode, simply leave out a clock attribute in the configuration entry object.

Configuration is provided as an array of menu options. Each entry of the array is an object with some properties specifying configuration. The configuration options specified are slightly different depending on whether or not it is for an active clock mode.

Configuration for Fixed Time Mode (no active clock)

  • timeSystem: A string, the key for the time system that this configuration relates to.
  • bounds: A Time Bounds object. These bounds will be applied when the user selects the time system specified in the previous timeSystem property.
  • zoomOutLimit: An optional number specifying the longest period of time that can be represented by the conductor when zooming. If a zoomOutLimit is provided, then a zoomInLimit must also be provided. If provided, the zoom slider will automatically become available in the Time Conductor UI.
  • zoomInLimit: An optional number specifying the shortest period of time that can be represented by the conductor when zooming. If a zoomInLimit is provided, then a zoomOutLimit must also be provided. If provided, the zoom slider will automatically become available in the Time Conductor UI.

Configuration for Active Clock

  • clock: A string, the key of the clock that this configuration applies to.
  • timeSystem: A string, the key for the time system that this configuration relates to. Separate configuration must be provided for each time system that you wish to be available to users when they select the specified clock.
  • clockOffsets: A clockOffsets object that will be automatically applied when the combination of clock and time system specified in this configuration is selected from the UI.

Example conductor configuration

An example time conductor configuration is provided below. It sets up some default options for the UTCTimeSystem and LocalTimeSystem, in both fixed mode, and for the LocalClock source. In this configutation, the local clock supports both the UTCTimeSystem and LocalTimeSystem. Configuration for fixed bounds mode is specified by omitting a clock key.

const ONE_YEAR = 365 * 24 * 60 * 60 * 1000;
const ONE_MINUTE = 60 * 1000;

openmct.install(openmct.plugins.Conductor({
    menuOptions: [
        // 'Fixed' bounds mode configuation for the UTCTimeSystem
        {
            timeSystem: 'utc',
            bounds: {start: Date.now() - 30 * ONE_MINUTE, end: Date.now()},
            zoomOutLimit: ONE_YEAR,
            zoomInLimit: ONE_MINUTE
        },
        // Configuration for the LocalClock in the UTC time system
        {
            clock: 'local',
            timeSystem: 'utc',
            clockOffsets: {start: - 30 * ONE_MINUTE, end: 0},
            zoomOutLimit: ONE_YEAR,
            zoomInLimit: ONE_MINUTE
        },
        //Configuration for the LocaLClock in the Local time system
        {
            clock: 'local',
            timeSystem: 'local',
            clockOffsets: {start: - 15 * ONE_MINUTE, end: 0}
        }
    ]
}));

Indicators

Indicators are small widgets that reside at the bottom of the screen and are visible from every screen in Open MCT. They can be used to convey system state using an icon and text. Typically an indicator will display an icon (customizable with a CSS class) that will reveal additional information when the mouse cursor is hovered over it.

The URL Status Indicator

A common use case for indicators is to convey the state of some external system such as a persistence backend or HTTP server. So long as this system is accessible via HTTP request, Open MCT provides a general purpose indicator to show whether the server is available and returning a 2xx status code. The URL Status Indicator is made available as a default plugin. See the documentation for details on how to install and configure the URL Status Indicator.

Creating a Simple Indicator

A simple indicator with an icon and some text can be created and added with minimal code. An indicator of this type exposes functions for customizing the text, icon, and style of the indicator.

eg.

var myIndicator = openmct.indicators.simpleIndicator();
myIndicator.text("Hello World!");
openmct.indicators.add(myIndicator);

This will create a new indicator and add it to the bottom of the screen in Open MCT. By default, the indicator will appear as an information icon. Hovering over the icon will reveal the text set via the call to .text(). The Indicator object returned by the API call exposes a number of functions for customizing the content and appearance of the indicator:

  • .text([text]): Gets or sets the text shown when the user hovers over the indicator. Accepts an optional string argument that, if provided, will be used to set the text. Hovering over the indicator will expand it to its full size, revealing this text alongside the icon. Returns the currently set text as a string.
  • .description([description]): Gets or sets the indicator's description. Accepts an optional string argument that, if provided, will be used to set the text. The description allows for more detail to be provided in a tooltip when the user hovers over the indicator. Returns the currently set text as a string.
  • .iconClass([className]): Gets or sets the CSS class used to define the icon. Accepts an optional string parameter to be used to set the class applied to the indicator. Any of the built-in glyphs may be used here, or a custom CSS class can be provided. Returns the currently defined CSS class as a string.
  • .statusClass([className]): Gets or sets the CSS class used to determine status. Accepts an optional string parameter to be used to set a status class applied to the indicator. May be used to apply different colors to indicate status.

Custom Indicators

A completely custom indicator can be added by simply providing a DOM element to place alongside other indicators.

    var domNode = document.createElement('div');
    domNode.innerText = new Date().toString();
    setInterval(function () {
        domNode.innerText = new Date().toString();
    }, 1000);

    openmct.indicators.add({
        element: domNode
    });

Priority API

Open MCT provides some built-in priority values that can be used in the application for view providers, indicators, root object order, and more.

Priority Types

Currently, the Open MCT Priority API provides (type: numeric value):

  • HIGH: 1000
  • Default: 0
  • LOW: -1000

View provider Example:

  class ViewProvider {
    ...
    priority() {
        return openmct.priority.HIGH;
    }
}

User API

Open MCT provides a User API which can be used to define providers for user information. The API can be used to manage user information and roles.

Example

Open MCT provides an example user and user provider which can be used as a starting point for creating a custom user provider.

Visibility-Based Rendering in View Providers

To enhance performance and resource efficiency in OpenMCT, a visibility-based rendering feature has been added. This feature is designed to defer the execution of rendering logic for views that are not currently visible. It ensures that views are only updated when they are in the viewport, similar to how modern browsers handle rendering of inactive tabs but optimized for the OpenMCT tabbed display. It also works when views are scrolled outside the viewport (e.g., in a Display Layout).

Overview

The show function is responsible for the rendering of a view. An Intersection Observer is used internally to determine whether the view is visible. This observer drives the visibility-based rendering feature, accessed via the renderWhenVisible function provided in the viewOptions parameter.

Implementing Visibility-Based Rendering

The renderWhenVisible function is passed to the show function as part of the viewOptions object. This function can be used for all rendering logic that would otherwise be executed within a requestAnimationFrame call. When called, renderWhenVisible will either execute the provided function immediately (via requestAnimationFrame) if the view is currently visible, or defer its execution until the view becomes visible.

Additionally, renderWhenVisible returns a boolean value indicating whether the provided function was executed immediately (true) or deferred (false).

Monitoring of visibility begins after the first call to renderWhenVisible is made.

Here’s the signature for the show function:

show(element, isEditing, viewOptions)

  • element (HTMLElement) - The DOM element where the view should be rendered.
  • isEditing (boolean) - Indicates whether the view is in editing mode.
  • viewOptions (Object) - An object with configuration options for the view, including:
    • renderWhenVisible (Function) - This function wraps the requestAnimationFrame and only triggers the provided render logic when the view is visible in the viewport.

Example

An OpenMCT view provider might implement the show function as follows:

// Define your view provider
const myViewProvider = {
  // ... other properties and methods ...
  show: function (element, isEditing, viewOptions) {
    // Callback for rendering view content
    const renderCallback = () => {
      // Your view rendering logic goes here
    };
    
    // Use the renderWhenVisible function to ensure rendering only happens when view is visible
    const wasRenderedImmediately = viewOptions.renderWhenVisible(renderCallback);

    // Optionally handle the immediate rendering return value
    if (wasRenderedImmediately) {
      console.debug('🪞 Rendering triggered immediately as the view is visible.');
    } else {
      console.debug('🛑 Rendering has been deferred until the view becomes visible.');
    }
  }
};

Note that renderWhenVisible defers rendering while the view is not visible and caters to the latest execution call. This provides responsiveness for dynamic content while ensuring performance optimizations.

Ensure your view logic is prepared to handle potentially multiple deferrals if using this API, as only the last call to renderWhenVisible will be queued for execution upon the view becoming visible.