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 and the
callback function to return data parsed as GeoJSON. The following code shows
examples of using the Model class in a template model.js file:
// model.js
function Model() {}
Model.prototype.getData = (req, callback) => {
// implement provider-specific logic here
// ...
const geojson = {
type: "FeatureCollection",
features: []
};
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, callback) {
// implement provider-specific logic here
// ...
const geojson = {
type: "FeatureCollection",
features: []
};
callback(null, geojson);
}
}
// 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 GeoJSON specification which allows for mixed geometries. However, feature services 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"
};
If your data does not contain a field that can be used as the OBJECTID
, you can leave
id
undefined. In such cases Custom Data Feeds will create an OBJECTID
for each
feature by default. However, this is less than ideal as it impacts performance and in
very rare cases may lead to OBJECTID
collisions. The best practice is to have a field
on the raw data that meets the requirements of id
.
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, provider configurations will be overwritten.
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
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.