A custom data provider connects to and fetches data from remote data sources. These data sources may be hosted static files, such as a CSV, remote APIs, or SQL or NoSQL databases. You can have one or more custom data providers in a custom data app. The custom data providers can adhere to two broadly defined types:
-
Full-fetch providers are designed to pull an entire dataset from a remote data source and perform post-processing functions after all of the data have been retrieved. This type of provider may be used where:
- datasets are small and the entire dataset can be fetched
- the remote API does not support filtering or queries
- the remote API is slow to respond
All queries and filtering performed in your ArcGIS client will only be performed on the dataset that is currently loaded into memory.
-
Pass-through providers allow for the remote API to perform some post-processing. By providing the analogous GeoServices API query parameters to the remote API, the remote API can return a subset of the data. This type of provider may be used where:
- datasets are large and the entire dataset cannot be fetched
- the remote API does support filtering or queries
- the remote API is quick to respond
Queries and filters in your ArcGIS client are translated in your custom data provider code into a query that your remote data source can understand. The request is sent to this remote data source, and the response is used to update what is currently in memory in your ArcGIS client. See the use of query parameters at the bottom of this page and in the example here: MongoDB sample code. See the topic on Pass-through providers for more information.
To create a custom data provider, open a command prompt on your
development machine in your custom data app directory, and run the cdf createprovider <name
command. The command will create a
project template for a custom data provider in the providers folder
inside the custom data app.
Understanding the Model Class
The model.js file contains a Model class for fetching data from your
data sources. The Model field in the index.js file must point to
this class. For each provider, you must implement the get
function
that includes your custom logic for fetching data and formatting it as GeoJSON.
The get
function accepts two arguments: req
and callback
. You must
use the req
argument to pass a standard Express.js request object, but the
callback
function is optional to return data parsed as GeoJSON. The syntax that
uses async
should not be combined with the syntax that uses callback
. Using
these two together is antipattern for JavaScript and can lead to issues with
error handling that may lead to process termination. The following code shows
examples of using the Model class in a model.js file:
// model.js
function Model() {}
Model.prototype.getData = (req, callback) => { // if using 'callback', do not use 'async'
// implement provider-specific logic here
// ...
const geojson = {
type: "FeatureCollection",
features: []
};
geojson.metadata = { idField: "id", maxRecordCount: 5000 }
callback(null, geojson);
};
// export model
module.exports = Model;
// model.js
class Model {
// implement some code here, such as a constructor with a database connection
async getData (req) { // if using 'async', do not use 'callback' as a second argument
// implement provider-specific logic here
// ...
const geojson = {
type: "FeatureCollection",
features: []
};
return {
...geojson,
metadata: {
idField: "id", // use the id field as the ObjectID
maxRecordCount: 5000
}
};
}
}
// export model
module.exports = Model;
To callback with an error, pass the error message as the first argument
to the callback
function as follows:
callback(err)
To callback with properly-formatted GeoJSON, call the
callback
function as follows:
callback(null, geojson)
All of the features in the GeoJSON features array must have the same geometry-type. This requirements differs from the GeoJSON specification which allows for mixed geometries. However, feature layers require uniform geometries, so it is important that this rule is enforced.
Custom Data Feeds requires that the GeoJSON be decorated with a property called metadata
.
It should have an object as its value. This object contains information important for
translating the GeoJSON to its feature service equivalent. Key GeoJSON metadata properties can be
referenced here: CDF API Reference. An example
of the GeoJSON metadata field is shown below.
geojson.metadata = {
idField: "id",
inputCrs: 3857
};
If your data does not contain a field that can be used as an OBJECTID
, then do not define the id
property in the GeoJSON metadata. In such cases, Custom Data Feeds will generate an OBJECTID
field and assign a unique value to each feature. This is not the recommended approach, as it can lead to rare OBJECTID
collisions and cannot be used with the pass-through pattern implementation. It can also cause issues when queries or client features (e.g., pop-ups) reference those generated IDs. The best practice is to include a stable, unique, non-null identifier in your source data and set it as the id
.
If your data is in WGS84 (WKID 4326), there is no need to specify the input
because no reprojection will need to be done by custom data feeds processing and returning geoJSON. However, if your data is stored in a different CRS, you must include the WKID in the metadata to ensure correct reprojection.
Understanding Provider Configuration
Define your configurable parameters, such as API keys and database connection strings, in the default.json file found in the config folder. Once defined, you can load the config module and access the parameters in any JavaScript file in the src directory typing this line:
const config = require('../config/default.json');
The top-level keys in the default.json file must be unique for all providers both in your development and production environments. If two provider configurations have the same top-level keys in their default.json files, additional providers with that key will not be able to be registered. In this case, the registration API will return an error.
Below is an example of a configuration file for a MongoDB provider, where mongodb
is the
top-level key.
{
"mongodb": {
"connectString": "<your connection string>",
"db": "<your db name>",
"collection": "<your collection name>"
}
}
Use Route Parameters to Fetch Data (Deprecated)
host
and id
are deprecated in ArcGIS Enterprise SDK 11.5. It is highly recommended that you use service parameters instead. See the sub-topic below.
You can use route parameters to build the URL to the remote data source.
Each provider provides two route parameters: host
and id
. The code
snippet below shows how to use these parameters to build a URL.
Model.prototype.getData(request, callback) {
const {params: {host, id}} = request
request(`https://${host}/resource${id}`, (err, res, body) => {
if (err) return callback(err)
// format response body as GeoJSON
callback(null, body)
})
}
To access the host
and id
parameters, set the properties.hosts
field to true
and theproperties.disable
field to false
in
the cdconfig.json file located in the root of your custom data provider
directory.
Creating Your Own Service Parameters
Developers have the ability to define as many service parameters as is required in their custom data provider. The developer chooses a key
, label
, and description
for each parameter. The key
is used in the provider
code, whereas the label
and description
are user-friendly text that are viewable to publishers and administrators. These service parameters are defined in cdconfig.json as shown below.
{
"name": "my-provider",
"arcgisVersion": "11.5.0",
"parentServiceType": "FeatureServer",
"customdataRuntimeVersion": "1",
"type": "provider",
"properties": {
"hosts": false,
"disableIdParam": true,
"serviceParameters": [
{
"label": "Doc ID",
"key": "doc_id",
"description": "Google Sheets document ID."
},
{
"label": "Geometry Column",
"key": "geom_column",
"description": "Name of the column in the file that contains the feature's geometry."
},
{
"label": "Data Coordinate System",
"key": "coordinate_system",
"description": "The WKID representing the coordinate system the remote data is stored in."
}
]
}
}
The properties.service
array is populated with objects containing label
, key
, and description
for each service parameter.
All parameters entered in this array are considered required parameters, and publishers will need to supply values for them at service
creation. properties.hosts
and properties.disable
are still present to maintain backward compatbility, but their default
values are set to false
and true
, respectively, rendering them inactive by default.
It is highly recommended that your service parameters are given useful label
and description
values because these values will be shown in ArcGIS
Enterprise clients such as Enterprise Manager and Portal for ArcGIS. Descriptive values for these attributes will make the service creation process
easier for publishers.
If using the ArcGIS Enterprise administrator APIs for creating custom data feature services, you will need to ensure that the service.json file includes these parameters in the json
object as shown below.
...rest of service.json
"jsonProperties": {"customDataProviderInfo": {
"forwardUserIdentity": false,
"dataProviderName": "my-provider",
"dataProviderHost": "",
"dataProviderId": "",
"serviceParameters": {
"doc_id": "parameter value",
"geom_value": "parameter value",
"coordinate_system": "parameter value"
}
}},
... rest of service.json
The values for these parameters are accessible in the provider's model.js methods just as the former host
and id
were on the req.params
object. Following the example above, the Google Sheets document id could be accessed in the get
method by req.params.doc
.
Development and Testing
When developing and testing the provider locally with the Custom Data Feeds CLI tool, you will need to use a client that can send the appropriate headers to simulate what happens in a production environment. Include a header named x-esri-cdf-service-params
whose value is an object with key
and value pairs. For example: {"doc
.
The authorize()
Method and the _user
Object
The authorize()
method is an optional method that can be employed in the model.js file. If used, authorize()
will be executed before get
. Inside this method, throwing an error will end the request, and neither the get
or the edit
methods will be executed.
Here is an example of how this method could be used to only allow requests to proceed for a specific set of allowed users.
const cdConfig = require('../cdconfig.json');
class Model {
#allowedUsers = [
'johnDoe1',
'johnDoe2',
'johnDoe3'
]
constructor({logger}) {
this.logger = logger;
}
async authorize(req) {
const requestUsername = req._user?.username;
if(this.#allowedUsers.includes(requestUsername)) {
this.logger.info(`In Provider "${cdConfig.name}", "authorize" method: pass`)
return;
}
this.logger.info(`In Provider "${cdConfig.name}", "authorize" method: reject`)
const error = new Error('Unauthorized');
error.code = 403;
throw error;
}
...
The _user
object is only available on the request object if the property forward
is set to true
in the service JSON. This property must be set by an administrator and is not set by default. Set the property in the json
object as seen below.
...rest of service.json
"jsonProperties": {"customDataProviderInfo": {
"forwardUserIdentity": true,
"dataProviderName": "my-provider",
"dataProviderHost": "",
"dataProviderId": "",
"serviceParameters": {
"doc_id": "parameter value",
"geom_value": "parameter value",
"coordinate_system": "parameter value"
}
}},
... rest of service.json
After setting forward
to true
, the _user
object will be availabe on the request object in the get
and edit
methods as well. User information that is available depends on whether the intial request was authorized (request made with a valid token) or unauthorized (request made without a valid ArcGIS Enterprise Server or Portal token). See the table below for expected properties available on the _user
object.
Request | ArcGIS Server | Parameter | Type |
---|---|---|---|
Authorized Request | Federated | username | string |
privileges | object | ||
role | string | ||
role | string | ||
id | string | ||
org | string | ||
Unfederated | username | string | |
server | object | ||
Unauthorized Request | Federated | none | empty object |
Unfederated | server * | object* |
Development and Testing
When developing and testing the provider locally with the Custom Data Feeds CLI tool, you will need to use a client that can send the appropriate headers to simulate what happens in a production environment. Include a header named x-esri-request-user
whose value is an object with key
and value pairs. For example: {"username"
.