enterprise-fetch
v1.0.0
Published
A fetch wrapper to leverage enterprise features such as retry policy and request timeout
Downloads
199
Maintainers
Readme
Enterprise Fetch
A fetch wrapper with some nice enterprise features that are missing from the default fetch implementations, such as request timeout, request proxy, and a fully configurable retry policy. Designed to work in both browser and NodeJS server environments.
ES6 module and CJS available for developers and pre-built minified bundles that can be used straight in the browser
Simple usage
The default export provides a fetch wrapper with a default retry policy set
Use it as a drop-in replacement for your existing fetch implementation
Examples
import enterpriseFetch from 'enterprise-fetch';
// enterpriseFetch(url: RequestInfo, init?: FetchInit): Promise<Response>
enterpriseFetch('https://httpbin.org/json')
.then((res) => res.json())
.then((json) => console.log(json));
enterpriseFetch('https://httpbin.org/post', {
method: 'post',
body: JSON.stringify({ a: 1 }),
headers: { 'Content-Type': 'application/json' },
})
.then((res) => res.json())
.then((json) => console.log(json));
Async / await
I prefer writing JavaScript in this way and would be the recommended approach where possible
const res = await enterpriseFetch('https://httpbin.org/post', {
method: 'post',
body: JSON.stringify({ a: 1 }),
headers: { 'Content-Type': 'application/json' },
});
const json = await res.json();
console.log(json);
Default retry policy
const fetchDefaults = {
// The timeout to apply before sending Abort signal
// to all requests that do not supply a timeout option
timeout: 60 * 1000,
// Retry policy for all fetch requests unless overridden
retry: {
retries: 3,
minTimeout: 400,
factor: 2,
},
// Do retry function to examine failures and apply custom
// retry logic return true to retry the fetch call
doRetry: async (attempt, res, { url, options }) => {
// Get the retry policy from options or fetchDefaults
const { retry = fetchDefaults.retry } = options || fetchDefaults;
if (attempt <= retry.retries) {
// Retry request on any network error,
// or 4xx or 5xx status codes. No retry on 404
if (!res.status || (res.status >= 400 && res.status !== 404)) {
const counter = `${attempt}/${retry.retries}`;
if ('message' in res)
console.warn(
`[fetch] Attempt ${counter} ${
res.name || ('type' in res && res.type)
}: ${res.message} ${url || ''}`
);
else
console.warn(
`[fetch] Attempt ${counter} ${res.status}: ${res.statusText} ${
url || ''
}`
);
return true;
}
}
return false;
},
};
Advanced usage
We can garnish any request init
object with some additional options as well as the full set of RequestInit
options supported by the standard fetch implementation
Fetch with timeout
Uses AbortController
signal instance to terminate the request after a specified timeout duration (in ms).
enterpriseFetch('https://httpbin.org/json', {
timeout: 5000,
})
.then((res) => res.json())
.then((json) => console.log(json));
Fetch with retries
Retry the request three times, with an initial wait timeout of 1000ms with an exponential backoff factor of 500ms.
e.g. wait 1000ms after first failed request, wait 1500ms between the second and 3000ms between the third request
This is simply the options object from https://www.npmjs.com/package/promise-retry
enterpriseFetch('https://httpbin.org/json', {
retry: {
retries: 3,
minTimeout: 1000,
factor: 500,
},
timeout: 5000,
})
.then((res) => res.json())
.then((json) => console.log(json));
Fetch with retry policy and custom retry logic
Custom retry logic is required to exploit domain specific rules when deciding if another retry is required.
For example, checking the headers for existance of a specific error key, or re-setting use-once headers such as an OAuth 1.0 bearer token for the next request
enterpriseFetch('https://httpbin.org/json', {
retry: {
retries: 3,
minTimeout: 1000,
factor: 500,
},
timeout: 5000,
doRetry: async (attempt, response, { options }) => {
const { retry } = options;
if (attempt <= retry.retries) {
// Retry request on any network error,
// or any 4xx or 5xx status codes, except 404
if (!res.status || (res.status >= 400 && res.status !== 404)) {
const counter = `${attempt}/${retry.retries}`;
if ('message' in res)
console.warn(
`[fetch] Attempt ${counter} ${
res.name || ('type' in res && res.type)
}: ${res.message} ${url || ''}`
);
else
console.warn(
`[fetch] Attempt ${counter} ${res.status}: ${res.statusText} ${
url || ''
}`
);
return true;
}
}
return false;
},
})
.then((res) => res.json())
.then((json) => console.log(json));
Application level retry policy
Create a fetch retry policy that can be used throughout your application.
Create a configuration file in your app to define fetch defaults that can be applied to all requests using this fetch wrapper.
You can then create an instance of fetchWithDefaults(fetchDefaults)
, supplying your fetchDefaults
. Then with this you can either redefine global fetch, or export the created instance of fetchWithDefaults(fetchDefaults)
to import wherever it is needed in your app.
import { fetchWithDefaults } from 'enterprise-fetch';
import { FetchInit } from 'enterprise-fetch/dist/models';
const fetchDefaults = {
// The timeout to apply to requests that do not supply a timeout option
timeout: 40 * 1000,
// Retry policy for all fetch requests
retry: {
retries: 3,
minTimeout: 400,
factor: 2,
},
// Do retry function to examine failures and apply custom retry logic
// return true to retry the fetch call
doRetry: async (attempt, res, { url, options }) => {
// Get the retry policy from options or fetchDefaults
const { retry = fetchDefaults.retry } = options || fetchDefaults;
if (attempt <= retry.retries) {
const counter = `${attempt}/${retry.retries}`;
Logger.warning(
`[fetch] Attempt ${counter} ${res.status}: ${res.statusText} ${
url || ''
}`
);
if ('clone' in res) {
const text = await res.clone().text();
const body = isJson(text) ? JSON.parse(text) : text;
if (body.error) {
// if (body.error && body.error.code === 'DUP_RCRD') {
// return false;
// }
}
}
// Retry request on any network error, or 4xx or 5xx status codes
if (
(res.status >= 400 && res.status !== 404) ||
('type' in res && res.type === 'aborted') ||
('name' in res && res.name === 'AbortError') ||
('code' in res && res.code === 'ETIMEDOUT')
) {
return true;
}
}
return false;
},
} as FetchInit;
// Redeclare fetch and apply our defaults
const fetch = fetchWithDefaults(fetchDefaults);
export default fetch;
Proxying requests (NodeJS)
NodeJS only, we can leverage the https-proxy-agent
to send our requests to a proxy server via the agent
option in RequestInit
, this is very useful for debugging and tracing fetch requests which is not natively possible with NodeJS.
To activate set process env variable http_proxy
to a valid http proxy server. e.g. process.env.http_proxy = http://127.0.0.1:8888
Browser bundles
Browser bundles are available for you to use if you are not using a module bundler for your application.
Simple example
<html>
<head>
<script
type="text/javascript"
src="https://unpkg.com/enterprise-fetch"
></script>
<script>
var element = function (id) {
return document.getElementById(id);
};
var goFetch = function () {
var uri = element('uri').value;
window['enterprise-fetch'].default(uri).then(function (response) {
alert(response.status);
});
};
</script>
</head>
<body>
<input type="text" id="uri" value="https://httpbin.org/status/500" />
<input type="button" id="fetch" value="Fetch" onclick="goFetch()" />
</body>
</html>
Legacy browsers
A cjs
bundle exists in addition to the esm
bundle loaded by default if you require legacy browser support. The cjs
bundle is 2x the size of the esm
bundle due to the syntax transpilation and polyfills.
Polyfills will be required for browsers that do not have native support for
- AbortController
- fetch
You could use a polyfill service to polyfill missing features by adding a snippet like the below before loading this script
<script
type="text/javascript"
src="https://polyfill.io/v3/polyfill.min.js?features=AbortController%2Cfetch"
></script>
Development scripts
npm start
builds the library and runs through all tests in command linenpm run build
build the library from source and create an instrumented copy of the build for coverage reportingnpm test
starts a http server and runs through all tests in command linenpm run cy:open
starts a http server and opens cypress testing gui for you to execute test specs interactively in a browser windownpm run report:coverage
checks code coverage and starts a http server rendering the test coverage report for review in a browsernpm run serve
starts a http server with a simple browser testing page available at the/test
route