Skip to content

Latest commit

 

History

History
310 lines (243 loc) · 12.7 KB

README.md

File metadata and controls

310 lines (243 loc) · 12.7 KB
Raw

fetch

Built latest

A small and pluggable lib to fetch a resource and cache the result.

Usage

By default fetch will treat all response codes except 200, 301 and 404 as errors. 404 will yield null and 200 the body.

By default, responses will be parsed as JSON. To fetch plain text - specify a contentType (see behavior options )

Caching

Fetch will parse the cache-control header. If fetch encounters private, no-cache, max-age=0 or must-revalidate it wont cache. Otherwise it will respect the max-age header.

Callback usage:

var fetchBuilder = require("exp-fetch");
var behavior = {};
var fetch = fetchBuilder(behavior).fetch;
fetch("http://example.com/resource.json", function (err, content) {
    // Do something with the result
});

Promise usage:

var fetchBuilder = require("exp-fetch");
var behavior = {};
var fetch = fetchBuilder(behavior).fetch;
fetch("http://example.com/resource.json").then(function (content) {
    // Do something with the result
    const json = content;
});

Custom request headers

var fetchBuilder = require("exp-fetch");
var behavior = {};
var fetch = fetchBuilder(behavior).fetch;
var options = {
  url: "http://example.com/resource.json",
  headers: {
    "User-agent": "exp-fetch"
  }
}
fetch(options, function (err, content) {
    // Do something with the result
});

Specifying contentType

To fetch XML parsed to a JSON response

var fetchBuilder = require("exp-fetch");
var behavior = { contentType: "xml" };
var fetch = fetchBuilder(behavior).fetch;
fetch("http://example.com/resource.xml", function (err, content) {
  //Do something with the result (it will be JSON-formatted)
  const json = content;
});

To fetch plain text or html

var fetchBuilder = require("exp-fetch");
var behavior = { contentType: "text" };
var fetch = fetchBuilder(behavior).fetch;
fetch("http://example.com", function (err, content) {
  // Do something with the result (which will be a string)
  const textString = content;
});

Using one build with many verbs

const behavior = {};
const request = require("exp-fetch")(behavior);

// using the .get/.post/.etc methods will override httpMethod from behavior with
// the specified verb
const getRes = await request.get("foo.com");
const postRes = await request.post("foo.com/epic-endpoint", {foo: "bar"});
const patchRes = await request.patch("foo.com/epic-endpoint", {foo: "bar"});
const headRes = await request.head("foo.com/epic-endpoint");
const optionsRes = await request.options("foo.com/epic-endpoint");
const putRes = await request.put("foo.com/epic-endpoint", {foo: "bar"});
const deleteRes = await request.del("foo.com/epic-endpoint");

Issuing POST requests:

var fetchBuilder = require("exp-fetch");
var behavior = { httpMethod: "POST"};
var poster = fetchBuilder(behavior).fetch;
var body = {
    "query": "some string"
};

poster("http://example.com/query", body, function (err, content) {
    // Do something with the result
    // The result will be cached by `url` + ` ` + `sha of body`
});

Allowed behavior options for fetchBuilder

Note: these options are used for the fetchBuilder, not the fetch-function

  • agent: (default: null), keepAlive Agent instance.
  • cache: (default: an instance of AsyncCache) (https://github.com/ExpressenAB/exp-asynccache). To disable caching set {cache: null}
  • cacheKeyFn: (default: caches on the url + sha1 of the body) An optional formatting function for finding the cache-key. One might, for example, want to cache on an url with the get params stripped or use some headers as part of the cache key.
  • cacheNotFound: (default: false). If set it will cache 404s, if given a number it will cache the 404 for that time. If the maxAgeFn is given, it will get this time as the first parameter.
  • cacheValueFn: (default: caches the response body) An optional function for change what will be returned and cached from fetch.
  • clone: (default: true), should fetch clone objects before handing them from the cache.
  • contentType: (default: json), expected content type. When set to xml the response will be parsed from XML to JSON. Other standard content-type values such as text or text/html are also accepted.
  • getCorrelationId: (default: null), for each request call this function to pass as the correlation id header specified below. Does not pass correlation id if function is not defined or if it returns null.
  • correlationIdHeader: (default: correlation-id), header to use when passing correlation id.
  • deepFreeze: (default:false). When this option is set to true it will freeze the response recursively so that it or any objects it contains can't be modified. ("use strict" is needed)
  • errorOnRemoteError: (default: true). If set it will treat a remote > 200 statusCode as an error.
  • followRedirect: (default: true), should fetch follow redirects (and cache the redirect chain)
  • freeze: (default:true). When this option is set to false it will not freeze the response so it can be modified. ("use strict" is needed)
  • httpMethod: (default: "GET"), the HTTP-method that should be used to make requests.
  • logger: A logger object implementing error, warning, info, debug for example https://github.com/tj/log.js
  • maxAgeFn: (default: respects the cache-control header)
  • onError: If given a function, it will be called each time fetch encounters a non 200 nor 404 response
  • onNotFound: If given a function, it will be called each time fetch encounters a 404
  • onRequestInit: If given a function, it will be called before the actual request is made, see Hooks for signature
  • onSuccess: If given a function, it will be called each time fetch encounters a 200
  • requestTimeFn: (default log with level debug) If given a function, it will be called when the request returned and processed from remote end.
  • retry: see got for details, defaults to 0
  • timeout: see got for details, defaults to 20000ms
  • hooks: see got for details, defaults to empty object

The difference between freeze and deepFreeze is that deepFreeze walks the object graph and freezes any child objects in the retrieved data. freeze only freezes the root object but still allows modifications to nested objects. deepFreeze will be slower since it is recursive.

CacheKeyFn

var fetchBuilder = require("exp-fetch");
var keyFinder = function (url) {
    return url.replace(/\//g, "");
}
var fetch = fetchBuilder({cacheKeyFn: keyFinder}).fetch;
Promise.all([
   fetch("http://example.com/foo/bar")
   fetch("http://example.com/foobar")
]).then(function (result) {
   result[0] === result[1];
});

CacheValueFn

var fetchBuilder = require("exp-fetch");
var valueFn  = function (body, headers, statusCode) {
    return {
        body: body,
        headers: headers,
        statusCode: statusCode
    };
}
var fetch = fetchBuilder({cacheValueFn: valueFn}).fetch;
fetch("http://example.com/resource.json", function (err, value) {
  // value will be something like:
  // { statusCode: 200, headers: { "content-type": "application/json" }, body: { "resource": "body" } }
})

maxAgeFn

var fetchBuilder = require("exp-fetch");
function cacheNothing(maxAge, key, res, content) {
    return -1;
}
var fetch = fetchBuilder({maxAgeFn: cacheNothing}).fetch;

Hooks

They are: onError, onNotFound and onSuccess. Signature:

function onError(url, cacheKey, res, content) {
    //
}

And onRequestInit with signature:

function onRequestInit(requestOptions, cacheKey) {
    //
}

The function will be called once before the actual request is made, i.e. not found in cache. Subsequent redirect requests does not call the function. The requestOptions argument is a copy of the request options and will not alter the request.

Useful when mocking requests, e.g:

var url = require("url");
var nock = require("nock");

function onRequestInit(requestOptions, cacheKey) {
    var callUrl = url.parse(requestOptions.url);
    var path = callUrl.path;
    var host = callUrl.protocol + "//" + callUrl.host;

    nock(host).get(path).reply(200, {mock: true});
}

var fetch = fetchBuilder({onRequestInit: onRequestInit}).fetch;

And requestTimeFn with signature:

function requestTimeFn(requestOptions, took) {
    console.log("REQUEST", requestOption.method, ":", requestOption.url, "took", took,  "ms");
}

Init cache function

The fetch lib provides a convenient initLRUCache-method which sets up a cache purging it's expired content.

var initLRUCache = require("exp-fetch").initLRUCache;
var cache = new AsyncCache(initLRUCache({ age: 60, size: 2000});

Allowed params:

  • size or max: the max allowed size, the unit is set by the length method. Default is value.length. Default: 2000000
  • length: the length function, default is v && v.length || 1
  • age or maxAge: the maximum number of seconds a key will be kept in the cache. Default 60

Stats

Get statistics for number calls and cache hit ratio:

var fetchBuilder = require("exp-fetch");
var behavior = {};
var stats = fetchBuilder(behavior).stats;
console.log("Hit ratio", stats().cacheHitRatio);

Timeout

Example: When setting up the exp-fetch timeout configuration, you can adjust the values of "socket" and "request" to ensure that the request doesn't exceed a predetermined response time. For example, if you know that the server response time is 3 seconds or 3000 milliseconds, you may want to choose values just slightly higher than 3 seconds.

In this case, the configuration code would look like the following:

timeout: {
    socket: 3500, // slightly higher than 3 seconds
    request: 4000, // slightly higher than 3 seconds 
}
The values of 3500 and 4000 milliseconds are chosen because they are just higher than the expected server response time of 3000 milliseconds. By setting the values in this way, the request will timeout if the server takes longer than expected to respond.

Please note that the situation described below is uncommon. However, if you have complete control over the responding server and can modify the timeout, you may be interested in the following information.

In such a case, you can refer to the example file located at examples/timeout.js. To run this file, simply copy it to the root directory and execute it with Node.

It's worth noting that having control over the responding server and its timeout can provide more flexibility and customization options for your requests.

Please note that running the node timeout.js command in the root directory should result in a successful fetch. However, if you decrease the timeout value of the "socket" option to 3000 milliseconds or 3 seconds, you will encounter an ESOCKETTIMEDOUT error.

This error indicates that the "socket" option must be set to a value higher than the server delay, and the "request" option must be set to a value higher than the timeout value of the "socket" option for the timeout options to function properly.

In most cases you would not have control over the responding server. In that case you need to add some retry logic see the examples/retry.js

  retry: {
    limit: 3,
    methods: [ "POST" ],
    statusCodes: [ 408, 500, 502, 503, 504 ],
    maxRetryAfter: 4000,
  },
property values description
limit 3 The maximum amount of times to retry the request.
methods [ "POST" ], The HTTP methods that should be retried.
statusCodes [ 408, 500, 502, 503, 504 ] The HTTP status codes that should be retried.
maxRetryAfter 4000 The maximum amount of time in milliseconds that the request should be retried after.

These values and property are examples and you can tweak and find other implementations based on your use case.

NOTE: You can copy the examples/retry.js to root and run it with node node retry.js In retry.js script the server delay is simulated to be delayed and different timeouts are passed to the server response, which should be a more relastic scenario.

NOTE: Basically if you have a timout configuration that starts to throw ESOCKETTIMEDOUT error you can try to add some retry logic. The timout option can be left in place and will work if server timout does not increase. If server timeout would increase then the retry options would kick in and rescue the fetch.