|
| 1 | +## JavaScript Fetch Retry (On failure) |
| 2 | + |
| 3 | +> There are many cases, where network failure happens randomly. This affects the consistency of the results, especially in API calls. So, below is an implementation of `fetch-retry` function which retries `fetch` upon failure upto `n` times. |
| 4 | +
|
| 5 | +> Fetch (or axios library) works great for most AJAX requests in JavaScript. However, network fails happen randomly. To catch this issue, we will implement a function `fetch_retry(url, options, n)`, which does `fetch(url, options)` but retries it upto n times upon failure. This increases the chances of success and displaying consistent results. |
| 6 | +
|
| 7 | +### Approach 1: Loop over n times and call fetch |
| 8 | +```js |
| 9 | +function fetch_retry(url, options, n) { |
| 10 | + for (let i = 0; i < n; i++) { |
| 11 | + fetch(url, options); |
| 12 | + if (success) return result; |
| 13 | + } |
| 14 | +} |
| 15 | +``` |
| 16 | +> Above approach will NOT work. Since FETCH is an asynchronous function, the program will not wait for the results before continuing. `n` fetches will be called at the same time, regardless of whether the previous calls succeed. |
| 17 | +
|
| 18 | +> `fetch` returns a Promise. So if when call succeeds, the function inside the `.then` will be called. And if the call fails, error will be caught in the `catch` block. |
| 19 | +
|
| 20 | +### fetch_retry outline |
| 21 | +```js |
| 22 | +function fetch_retry(url, options, n) { |
| 23 | + fetch(url, options) |
| 24 | + .then(function(result) { |
| 25 | + // on success |
| 26 | + }).catch(function(error) { |
| 27 | + // on failure |
| 28 | + }); |
| 29 | +} |
| 30 | +``` |
| 31 | + |
| 32 | +> We will use a promise to make fetch_retry work synchronously. So `fetch_retry` will return a Promise, that resolves if any attempt out of n attempts succeed, and rejects if all `n` attempts succeed. |
| 33 | +```js |
| 34 | +function fetch_retry(url, options, n) { |
| 35 | + return new Promise(function(resolve, reject) { |
| 36 | + fetch(url, options) |
| 37 | + .then(function(result) { |
| 38 | + // on success |
| 39 | + }).catch(function(error) { |
| 40 | + // on failure |
| 41 | + }) |
| 42 | + }); |
| 43 | +} |
| 44 | +``` |
| 45 | + |
| 46 | +> If fetch succeeds, we can resolve the Promise returned by calling the resolve function. |
| 47 | +```js |
| 48 | +function fetch_retry(url, options, n) { |
| 49 | + return new Promise(function(resolve, reject) { |
| 50 | + fetch(url, options) |
| 51 | + .then(function(result) { |
| 52 | + // on success |
| 53 | + resolve(result); // Call resolve to resolve the returning promise |
| 54 | + }).catch(function(error) { |
| 55 | + // on failure |
| 56 | + }) |
| 57 | + }); |
| 58 | +} |
| 59 | +``` |
| 60 | + |
| 61 | +> If fetch fails, we call the fetch_retry function recursively, and reduce the the value of `n` |
| 62 | +```js |
| 63 | +function fetch_retry(url, options, n) { |
| 64 | + return new Promise(function(resolve, reject) { |
| 65 | + fetch(url, options) |
| 66 | + .then(function(result) { |
| 67 | + // on success |
| 68 | + resolve(result); |
| 69 | + }).catch(function(error) { |
| 70 | + // on failure |
| 71 | + fetch_retry(url, options, n - 1) // Call fetch_retry recursively |
| 72 | + .then(/* one of the remaining n - 1 calls to fetch will succeed */) |
| 73 | + .catch(/* remaining n - 1 fetch failed */) |
| 74 | + }) |
| 75 | + }); |
| 76 | +} |
| 77 | +``` |
| 78 | + |
| 79 | +> `fetch_retry` will return a Promise and resolve if any of the `n-1` attempts succeed and rejects if all the `n-1` attempts fail. |
| 80 | +
|
| 81 | +> If `fetch_retry` in the on failure case, succeeds, Promise will resolve, and reject if all the calls fail. |
| 82 | +```js |
| 83 | +function fetch_retry(url, options, n) { |
| 84 | + return new Promise(function(resolve, reject) { |
| 85 | + fetch(url, options) |
| 86 | + .then(function(result) { |
| 87 | + // on success |
| 88 | + resolve(result); |
| 89 | + }).catch(function(error) { |
| 90 | + // on failure |
| 91 | + fetch_retry(url, options, n - 1) |
| 92 | + .then(resolve) // Resolve |
| 93 | + .catch(reject); // Reject |
| 94 | + }) |
| 95 | + }); |
| 96 | +} |
| 97 | +``` |
| 98 | + |
| 99 | +> Since we have recursion involved, we need a base case to exit the function when all the fetch calls have failed `n` times. So if `n === 1` and fetch call fails, we reject with error from fetch, without calling fetch_retry any further |
| 100 | +```js |
| 101 | +function fetch_retry(url, options, n) { |
| 102 | + return new Promise(function(resolve, reject) { |
| 103 | + fetch(url, options) |
| 104 | + .then(function(result) { |
| 105 | + // on success |
| 106 | + resolve(result); |
| 107 | + }).catch(function(error) { |
| 108 | + if (n === 1) return reject(error); |
| 109 | + fetch_retry(url, options, n - 1) |
| 110 | + .then(resolve) |
| 111 | + .catch(reject); |
| 112 | + }) |
| 113 | + }); |
| 114 | +} |
| 115 | +``` |
| 116 | + |
| 117 | +> Since fetch itself returns a Promise, the `return Promise` part is redundant. Also we do not need to explicitly call resolve and reject on failure after `fetch_retry` is called in the catch section. So by cleaning up, we have below: |
| 118 | +```js |
| 119 | +function fetch_retry(url, options, n) { |
| 120 | + return fetch(url, options) |
| 121 | + .catch(function(error) { |
| 122 | + if (n === 1) throw error; |
| 123 | + return fetch_retry(url, options, n - 1); |
| 124 | + }); |
| 125 | +} |
| 126 | +``` |
| 127 | + |
| 128 | +> Converting above to use arrow functions |
| 129 | +```js |
| 130 | +const fetch_retry = (url, options, n) => fetch(url, options).catch(error => { |
| 131 | + if (n === 1) throw error; |
| 132 | + return fetch_retry(url, options, n - 1); |
| 133 | +}); |
| 134 | +``` |
| 135 | + |
| 136 | +> Using ES7 async await |
| 137 | +```js |
| 138 | +const fetch_retry = async (url, options, n) => { |
| 139 | + try { |
| 140 | + return await fetch(url, options); |
| 141 | + } catch(error) { |
| 142 | + if (n === 1) throw error; |
| 143 | + return await fetch_retry(url, options, n - 1); |
| 144 | + } |
| 145 | +}; |
| 146 | +``` |
| 147 | + |
| 148 | +> async await without recursion |
| 149 | +```js |
| 150 | +const fetch_retry = async(url, options, n) => { |
| 151 | + let error; |
| 152 | + |
| 153 | + for (let i = 0; i < n; i++) { |
| 154 | + try { |
| 155 | + return await fetch(url, options); |
| 156 | + } catch(err) { |
| 157 | + error = err; |
| 158 | + } |
| 159 | + } |
| 160 | + |
| 161 | + throw error; |
| 162 | +} |
| 163 | +``` |
| 164 | + |
| 165 | +> Google Phone Interview question: Call an API, which is unstable in returning a response, and fails occassionally. So take the average of 3 successfull responses and return that as a result. However, if the entire process is not complete within 10 seconds, show an error message to the user. |
| 166 | +```js |
| 167 | +// Need to revisit |
| 168 | +const fetch_retry = async(url, options, n) => { |
| 169 | + let error; |
| 170 | + let result = []; |
| 171 | + let i = 0; |
| 172 | + |
| 173 | + while (true) { |
| 174 | + try { |
| 175 | + if (i === n) { |
| 176 | + return result.reduce((acc, currentValue) => acc + currentValue, 0); |
| 177 | + } |
| 178 | + response = await fetch(url, options); |
| 179 | + result.push(response); |
| 180 | + i++; |
| 181 | + } catch(err) { |
| 182 | + error = err; |
| 183 | + } |
| 184 | + } |
| 185 | + |
| 186 | + throw error; |
| 187 | +} |
| 188 | +let response = fetch_retry(url, options, 3); |
| 189 | + |
| 190 | +setTimeout(function() { |
| 191 | + if (!response) { |
| 192 | + console.log("There was an error with the API"); |
| 193 | + } else { |
| 194 | + console.log("Result is: ", response); |
| 195 | + } |
| 196 | +}, 10000) |
| 197 | +``` |
0 commit comments