setupCache()
The setupCache
function receives the axios instance and a set of optional properties described below. This modifies the axios instance in place and returns it.
const axios = setupCache(axiosInstance, OPTIONS);
TIP
The setupCache
function receives global options and all request specifics ones too. This way, you can customize the defaults for all requests.
TIP
If you want to use the same cache interceptor for all your axios instances, you can call setupCache
with the default axios instance.
import Axios from 'axios';
setupCache(Axios, OPTIONS);
storage
optional- Type:
AxiosStorage
- Default:
buildMemoryStorage()
A storage interface is the entity responsible for saving, retrieving and serializing data received from network and requested when a axios call is made.
See the Storages page for more information.
generateKey
optional- Type:
KeyGenerator<unknown, unknown>
- Default:
defaultKeyGenerator
The generateKey
property defines the function responsible for generating unique keys for each request cache.
By default, it employs a strategy that prioritizes the id
if available, falling back to a string generated using various request properties. The default implementation generates a 32-bit hash key using the method
, baseURL
, params
, data
, and url
of the request.
WARNING
In any persistent cache scenario where hitting over 77K unique keys is a possibility, you should use a more robust hashing algorithm.
waiting
optional- Type:
Map<string, Deferred<void>>
- Default:
new Map
A simple object that will hold a promise for each pending request. Used to handle concurrent requests.
You shouldn’t change this property, but it is exposed in case you need to use it as some sort of listener or know when a request is waiting for others to finish.
headerInterpreter
optional- Type:
HeaderInterpreter
- Default:
defaultHeaderInterpreter
The function used to interpret all headers from a request and determine a time to live (ttl
) number.
WARNING
Many REST backends returns some variation of Cache-Control: no-cache
or Cache-Control: no-store
headers, which tell us to ignore caching at all. You shall disable headerInterpreter
for those requests.
If the debug mode prints Cache header interpreted as 'dont cache'
this is probably the reason.
The possible returns are:
'dont cache'
: the request will not be cached.'not enough headers'
: the request will find other ways to determine the TTL value.number
: used as the TTL value.{ cache: number, stale: number }
: used as the TTL value and stale TTL value
Example of a custom headerInterpreter
import {
setupCache,
type HeaderInterpreter
} from 'axios-cache-interceptor';
const myHeaderInterpreter: HeaderInterpreter = (headers) => {
if (headers['x-my-custom-header']) {
const seconds = Number(headers['x-my-custom-header']);
if (seconds < 1) {
return 'dont cache';
}
return seconds;
}
return 'not enough headers';
};
requestInterceptor
optional- Type:
AxiosInterceptor<CacheRequestConfig<unknown, unknown>>
- Default:
defaultRequestInterceptor()
The function that will be used to intercept the request before it is sent to the axios adapter.
It is the main function of this library, as it is the bridge between the axios request and the cache.
It wasn’t meant to be changed, but if you need to, you can do it by passing a new function to this property.
See its code for more information here.
responseInterceptor
optional- Type:
AxiosInterceptor<CacheAxiosResponse<unknown, unknown>>
- Default:
defaultResponseInterceptor()
The function that will be used to intercept the request after it is returned by the axios adapter.
It is the second most important function of this library, as it is the bridge between the axios response and the cache.
It wasn’t meant to be changed, but if you need to, you can do it by passing a new function to this property.
See its code for more information here.
debug
dev only optional- Type:
(msg: { id?: string; msg?: string; data?: unknown }) => void
orundefined
- Default:
undefined
WARNING
This option only works when targeting a Development build.
The debug option will print debug information in the console. It is good if you need to trace any undesired behavior or issue. You can enable it by setting debug
to a function that receives an string and returns nothing.
Read the Debugging page for the complete guide.
Example of a custom debug function
// Will print debug info in the console.
setupCache(axiosInstance, { debug: console.log });
// Own logging platform.
setupCache(axiosInstance, {
debug: ({ id, msg, data }) =>
myLoggerExample.emit({ id, msg, data })
});
// Disables debug. (default)
setupCache(axiosInstance, { debug: undefined });