Fetchr

Universal data access layer for web applications.

Typically on the server, you call your API or database directly to fetch some data. However, on the client, you cannot always call your services in the same way (i.e, cross domain policies). Instead, XHR/fetch requests need to be made to the server which get forwarded to your service.

Having to write code differently for both environments is duplicative and error prone. Fetchr provides an abstraction layer over your data service calls so that you can fetch data using the same API on the server and client side.

Install

npm install fetchr --save

Important: when on browser, Fetchr relies fully on Fetch API. If you need to support old browsers, you will need to install a polyfill as well (eg. https://github.com/github/fetch).

Setup

Follow the steps below to setup Fetchr properly. This assumes you are using the Express framework.

1. Configure Server

On the server side, add the Fetchr middleware into your express app at a custom API endpoint.

Fetchr middleware expects that you're using the body-parser middleware (or an alternative middleware that populates req.body) before you use Fetchr middleware.

import express from 'express';
import Fetcher from 'fetchr';
import bodyParser from 'body-parser';
const app = express();

// you need to use body-parser middleware before fetcher middleware
app.use(bodyParser.json());

app.use('/myCustomAPIEndpoint', Fetcher.middleware());

2. Configure Client

On the client side, it is necessary for the xhrPath option to match the path where the middleware was mounted in the previous step

xhrPath is an optional config property that allows you to customize the endpoint to your services, defaults to /api.

import Fetcher from 'fetchr';
const fetcher = new Fetcher({
    xhrPath: '/myCustomAPIEndpoint',
});

3. Register data services

You will need to register any data services that you wish to use in your application. The interface for your service will be an object that must define a resource property and at least one CRUD operation. The resource property will be used when you call one of the CRUD operations.

// app.js
import Fetcher from 'fetchr';
import myDataService from './dataService';
Fetcher.registerService(myDataService);

// dataService.js
export default {
    // resource is required
    resource: 'data_service',
    // at least one of the CRUD methods is required
    read: async function ({ req, resource, params, config }) {
        return { data: 'foo' };
    },
    // other methods
    // create: async function({ req, resource, params, body, config }) {},
    // update: async function({ req, resource, params, body, config }) {},
    // delete: async function({ req, resource, params, config }) {}
};

4. Instantiating the Fetchr Class

Data services might need access to each individual request, for example, to get the current logged in user's session. For this reason, Fetcher will have to be instantiated once per request.

On the serverside, this requires fetcher to be instantiated per request, in express middleware. On the clientside, this only needs to happen on page load.

// app.js - server
import express from 'express';
import Fetcher from 'fetchr';
import myDataService from './dataService';
const app = express();

// register the service
Fetcher.registerService(myDataService);

// register the middleware
app.use('/myCustomAPIEndpoint', Fetcher.middleware());

app.use(function (req, res, next) {
    // instantiated fetcher with access to req object
    const fetcher = new Fetcher({
        xhrPath: '/myCustomAPIEndpoint', // xhrPath will be ignored on the serverside fetcher instantiation
        req: req,
    });

    // perform read call to get data
    fetcher
        .read('data_service')
        .params({ id: 42 })
        .then(({ data, meta }) => {
            // handle data returned from data fetcher in this callback
        })
        .catch((err) => {
            // handle error
        });
});

// app.js - client
import Fetcher from 'fetchr';
const fetcher = new Fetcher({
    xhrPath: '/myCustomAPIEndpoint', // xhrPath is REQUIRED on the clientside fetcher instantiation
});
fetcher
    .read('data_api_fetcher')
    .params({ id: 42 })
    .then(({ data, meta }) => {
        // handle data returned from data fetcher in this callback
    })
    .catch((err) => {
        // handle errors
    });

// for create you can use the body() method to pass data
fetcher
    .create('data_api_create')
    .body({ some: 'data' })
    .then(({ data, meta }) => {
        // handle data returned from data fetcher in this callback
    })
    .catch((err) => {
        // handle errors
    });

Usage Examples

See the simple example.

Service Metadata

Service calls on the client transparently become fetch requests. It is a good idea to set cache headers on common fetch calls. You can do so by providing a third parameter in your service's callback. If you want to look at what headers were set by the service you just called, simply inspect the third parameter in the callback.

Note: If you're using promises, the metadata will be available on the meta property of the resolved value.

// dataService.js
export default {
    resource: 'data_service',
    read: async function ({ req, resource, params, config }) {
        return {
            data: 'response', // business logic
            meta: {
                headers: {
                    'cache-control': 'public, max-age=3600',
                },
                statusCode: 200, // You can even provide a custom statusCode for the fetch response
            },
        };
    },
};

fetcher
    .read('data_service')
    .params({id: ###})
    .then(({ data, meta }) {
        // data will be 'response'
        // meta will have the header and statusCode from above
    });

There is a convenience method called fetcher.getServiceMeta on the fetchr instance. This method will return the metadata for all the calls that have happened so far in an array format. In the server, this will include all service calls for the current request. In the client, this will include all service calls for the current session.

Updating Configuration

Usually you instantiate fetcher with some default options for the entire browser session, but there might be cases where you want to update these options later in the same session.

You can do that with the updateOptions method:

// Start
const fetcher = new Fetcher({
    xhrPath: '/myCustomAPIEndpoint',
    xhrTimeout: 2000,
});

// Later, you may want to update the xhrTimeout
fetcher.updateOptions({
    xhrTimeout: 4000,
});

Error Handling

When an error occurs in your Fetchr CRUD method, you should throw an error object. The error object should contain a statusCode (default 500) and output property that contains a JSON serializable object which will be sent to the client.

export default {
    resource: 'FooService',
    read: async function create(req, resource, params, configs) {
        const err = new Error('it failed');
        err.statusCode = 404;
        err.output = { message: 'Not found', more: 'meta data' };
        err.meta = { foo: 'bar' };
        throw err;
    },
};

And in your service call:

fetcher
    .read('someData')
    .params({ id: '42' })
    .catch((err) => {
        // err instanceof FetchrError -> true
        // err.message -> "Not found"
        // err.meta -> { foo: 'bar' }
        // err.name = 'FetchrError'
        // err.output -> { message: "Not found", more: "meta data" }
        // err.rawRequest -> { headers: {}, method: 'GET', url: '/api/someData' }
        // err.reason -> BAD_HTTP_STATUS | BAD_JSON | TIMEOUT | ABORT | UNKNOWN
        // err.statusCode -> 404
        // err.timeout -> 3000
        // err.url -> '/api/someData'
    });

Abort support

An object with an abort method is returned when creating fetchr requests on client. This is useful if you want to abort a request before it is completed.

const req = fetcher
    .read('someData')
    .params({ id: 42 })
    .catch((err) => {
        // err.reason will be ABORT
    });

req.abort();

Timeouts

xhrTimeout is an optional config property that allows you to set timeout (in ms) for all clientside requests, defaults to 3000. On the clientside, xhrPath and xhrTimeout will be used for all requests. On the serverside, xhrPath and xhrTimeout are not needed and are ignored.

import Fetcher from 'fetchr';
const fetcher = new Fetcher({
    xhrPath: '/myCustomAPIEndpoint',
    xhrTimeout: 4000,
});

If you want to set a timeout per request you can call clientConfig with a timeout property:

fetcher
    .read('someData')
    .params({ id: 42 })
    .clientConfig({ timeout: 5000 }) // wait 5 seconds for this request before timing out
    .catch((err) => {
        // err.reason will be TIMEOUT
    });

Params Processing

For some applications, there may be a situation where you need to process the service params passed in the request before they are sent to the actual service. Typically, you would process them in the service itself. However, if you need to perform processing across many services (i.e. sanitization for security), then you can use the paramsProcessor option.

paramsProcessor is a function that is passed into the Fetcher.middleware method. It is passed three arguments, the request object, the serviceInfo object, and the service params object. The paramsProcessor function can then modify the service params if needed.

Here is an example:

/**
    Using the app.js from above, you can modify the Fetcher.middleware
    method to pass in the paramsProcessor function.
 */
app.use(
    '/myCustomAPIEndpoint',
    Fetcher.middleware({
        paramsProcessor: function (req, serviceInfo, params) {
            console.log(serviceInfo.resource, serviceInfo.operation);
            return Object.assign({ foo: 'fillDefaultValueForFoo' }, params);
        },
    }),
);

Response Formatting

For some applications, there may be a situation where you need to modify the response before it is passed to the client. Typically, you would apply your modifications in the service itself. However, if you need to modify the responses across many services (i.e. add debug information), then you can use the responseFormatter option.

responseFormatter is a function that is passed into the Fetcher.middleware method. It is passed three arguments, the request object, response object and the service response object (i.e. the data returned from your service). The responseFormatter function can then modify the service response to add additional information.

Take a look at the example below:

/**
    Using the app.js from above, you can modify the Fetcher.middleware
    method to pass in the responseFormatter function.
 */
app.use(
    '/myCustomAPIEndpoint',
    Fetcher.middleware({
        responseFormatter: function (req, res, data) {
            data.debug = 'some debug information';
            return data;
        },
    }),
);

Now when an request is performed, your response will contain the debug property added above.

CORS Support

Fetchr provides CORS support by allowing you to pass the full origin host into corsPath option.

For example:

import Fetcher from 'fetchr';
const fetcher = new Fetcher({
    corsPath: 'http://www.foo.com',
    xhrPath: '/fooProxy',
});
fetcher.read('service').params({ foo: 1 }).clientConfig({ cors: true });

Additionally, you can also customize how the GET URL is constructed by passing in the constructGetUri property when you execute your read call:

import qs from 'qs';
function customConstructGetUri(uri, resource, params, config) {
    // this refers to the Fetcher object itself that this function is invoked with.
    if (config.cors) {
        return uri + '/' + resource + '?' + qs.stringify(this.context);
    }
    // Return `falsy` value will result in `fetcher` using its internal path construction instead.
}

import Fetcher from 'fetchr';
const fetcher = new Fetcher({
    corsPath: 'http://www.foo.com',
    xhrPath: '/fooProxy',
});
fetcher.read('service').params({ foo: 1 }).clientConfig({
    cors: true,
    constructGetUri: customConstructGetUri,
});

CSRF Protection

You can protect your Fetchr middleware paths from CSRF attacks by adding a middleware in front of it:

app.use('/myCustomAPIEndpoint', csrf(), Fetcher.middleware());

You could use https://github.com/expressjs/csurf for this as an example.

Next you need to make sure that the CSRF token is being sent with our requests so that they can be validated. To do this, pass the token in as a key in the options.context object on the client:

const fetcher = new Fetcher({
    xhrPath: '/myCustomAPIEndpoint', // xhrPath is REQUIRED on the clientside fetcher instantiation
    context: {
        // These context values are persisted with client calls as query params
        _csrf: 'Ax89D94j',
    },
});

This _csrf will be sent in all client requests as a query parameter so that it can be validated on the server.

Service Call Config

When calling a Fetcher service you can pass an optional config object.

When this call is made from the client, the config object is used to set some request options and can be used to override default options:

//app.js - client
const config = {
    timeout: 6000, // Timeout (in ms) for each request
    unsafeAllowRetry: false, // for POST requests, whether to allow retrying this post
};

fetcher.read('service').params({ id: 1 }).clientConfig(config);

For requests from the server, the config object is simply passed into the service being called.

Retry

You can set Fetchr to automatically retry failed requests by specifying a retry configuration in the global or in the request configuration:

// Globally
const fetchr = new Fetchr({
    retry: { maxRetries: 2 },
});

// Per request
fetchr.read('service').clientConfig({
    retry: { maxRetries: 1 },
});

With the above configuration, Fetchr will retry twice all requests that fail but only once when calling read('service').

You can further customize how the retry mechanism works. These are all settings and their default values:

const fetchr = new Fetchr({
  retry: {
    maxRetries: 2, // amount of retries after the first failed request
    interval: 200, // maximum interval between each request in ms (see note below)
    statusCodes: [0, 408], // response status code that triggers a retry (see note below)
  },
  unsafeAllowRetry: false, // allow unsafe operations to be retried (see note below)
}

interval

The interval between each request respects the following formula, based on the exponential backoff and full jitter strategy published in this AWS architecture blog post:

Math.random() * Math.pow(2, attempt) * interval;

attempt is the number of the current retry attempt starting from 0. By default interval corresponds to 200ms.

statusCodes

For historical reasons, fetchr only retries 408 responses and no responses at all (for example, a network error, indicated by a status code 0). However, you might find useful to also retry on other codes as well (502, 503, 504 can be good candidates for an automatic retries).

unsafeAllowRetry

By default, Fetchr only retries read requests. This is done for safety reasons: reading twice an entry from a database is not as bad as creating an entry twice. But if your application or resource doesn't need this kind of protection, you can allow retries by setting unsafeAllowRetry to true and fetchr will retry all operations.

Context Variables

By Default, fetchr appends all context values to the request url as query params. contextPicker allows you to have greater control over which context variables get sent as query params depending on the request method (GET or POST). This is useful when you want to limit the number of variables in a GET url in order not to accidentally cache bust.

contextPicker follows the same format as the predicate parameter in lodash/pickBy with two arguments: (value, key).

const fetcher = new Fetcher({
    context: {
        // These context values are persisted with client calls as query params
        _csrf: 'Ax89D94j',
        device: 'desktop',
    },
    contextPicker: {
        GET: function (value, key) {
            // for example, if you don't enable CSRF protection for GET, you are able to ignore it with the url
            if (key === '_csrf') {
                return false;
            }
            return true;
        },
        // for other method e.g., POST, if you don't define the picker, it will pick the entire context object
    },
});

const fetcher = new Fetcher({
    context: {
        // These context values are persisted with client calls as query params
        _csrf: 'Ax89D94j',
        device: 'desktop',
    },
    contextPicker: {
        GET: ['device'], // predicate can be an array of strings
    },
});

Custom Request Headers

When calling a Fetcher service you can add custom request headers.

A request contains custom headers when you add headers option to 'clientConfig'.

const config = {
    headers: {
        'X-VERSION': '1.0.0',
    },
};

fetcher.read('service').params({ id: 1 }).clientConfig(config);

All requests contain custom headers when you add headers option to constructor arguments of 'Fetcher'.

import Fetcher from 'fetchr';
const fetcher = new Fetcher({
    headers: {
        'X-VERSION': '1.0.0',
    },
});

Stats Monitoring & Analysis

To collect fetcher service's success/failure/latency stats, you can configure statsCollector for Fetchr. The statsCollector function will be invoked with one argumment: stats. The stats object will contain the following fields:

• resource: The name of the resource for the request
• operation: The name of the operation, create|read|update|delete
• params: The params object for the resource
• statusCode: The status code of the response
• err: The error object of failed request; null if request was successful
• time: The time spent for this request, in milliseconds

Fetcher Instance

import Fetcher from 'fetchr';
const fetcher = new Fetcher({
    xhrPath: '/myCustomAPIEndpoint',
    statsCollector: function (stats) {
        // just console logging as a naive example.  there is a lot more you can do here,
        // like aggregating stats or filtering out stats you don't want to monitor
        console.log(
            'Request for resource',
            stats.resource,
            'with',
            stats.operation,
            'returned statusCode:',
            stats.statusCode,
            ' within',
            stats.time,
            'ms',
        );
    },
});

Server Middleware

app.use(
    '/myCustomAPIEndpoint',
    Fetcher.middleware({
        statsCollector: function (stats) {
            // just console logging as a naive example.  there is a lot more you can do here,
            // like aggregating stats or filtering out stats you don't want to monitor
            console.log(
                'Request for resource',
                stats.resource,
                'with',
                stats.operation,
                'returned statusCode:',
                stats.statusCode,
                ' within',
                stats.time,
                'ms',
            );
        },
    }),
);

API

• Fetchr

License

This software is free to use under the Yahoo! Inc. BSD license. See the LICENSE file for license text and copyright information.