Api middleware with redux-observable - redux

I'm refactoring my react/redux app to use redux-observable instead of redux-thunk. Using thunk, I have an api middleware set up to listen for any actions with a CALL_API key and do some manipulation of the data, prepare headers, prepare full url, perform an api call using axios, and also do some additional action dispatches related to an api call.
Importantly, the api middleware dispatches a REQUEST_START action which gives the request an id and sets its status to pending in the network part of my state. When the promise from axios resolves or rejects, the middleware dispatches a REQUEST_END action, updating the state so that the current request is set to resolved or rejected. Then the response is returned to the calling action creator that initially dispatched the CALL_API action.
I have not been able to figure out how to do this with redux-observable. The part about the api middleware described above that I want to replicate is the REQUEST_START and REQUEST_END action dispatches. It's very convenient to have a centralized place where all api call related stuff is handled. I know I can effectively dispatch the REQUEST_START and REQUEST_END actions in each of my epics that does an api call, but I don't want to have to repeat the same code in many places.
I managed to partially solve this by creating an apiCallEpic which listens for actions with type CALL_API and does the above setup for api calls. However, an issue (or rather, something I don't like) is that the epic that initiates the api call (e.g. getCurrentUserEpic) essentially gives up control to apiCallEpic.
So, for example, when the api call succeeds and has a response, I may want to format that response data in some way before dispatching an action to be handled by my reducer. That is, getCurrentUserEpic should do some formatting of data returned from api call before sending to reducer. I was able to achieve something close to this by passing a payloadHandler callback function defined in getCurrentUserEpic that the apiCallEpic can call if/when it gets a successful response. However, I don't like this callback architecture and it seems like there's got to be a better way.
Here is some code that demonstrates my use of api middleware using thunk.
import axios from 'axios';
// actionCreators.js
// action types
const CALL_API = "CALL_API";
const FETCH_CURRENT_USER = "FETCH_CURRENT_USER";
const RECEIVE_CURRENT_USER = "RECEIVE_CURRENT_USER";
// action creators for request start and end
export const reqStart = (params = {}) => (dispatch) => {
const reduxAction = {
type: REQ_START,
status: 'pending',
statusCode: null,
requestId: params.requestId,
}
dispatch(reduxAction);
}
export const reqEnd = (params = {}) => (dispatch) => {
const {
requestId,
response = null,
error = null,
} = params;
let reduxAction = {}
if (response) {
reduxAction = {
type: REQ_END,
status: 'success',
statusCode: response.status,
requestId,
}
}
else if (error) {
if (error.response) {
reduxAction = {
type: REQ_END,
status: 'failed',
statusCode: error.response.status,
requestId,
}
}
else {
reduxAction = {
type: REQ_END,
status: 'failed',
statusCode: 500,
requestId,
}
}
}
dispatch(reduxAction);
}
// some api call to fetch data
export const fetchCurrentUser = (params = {}) => (dispatch) => {
const config = {
url: '/current_user',
method: 'get',
}
const apiCall = {
[CALL_API]: {
config,
requestId: FETCH_CURRENT_USER,
}
}
return dispatch(apiCall)
.then(response => {
dispatch({
type: RECEIVE_CURRENT_USER,
payload: {response},
})
return Promise.resolve({response});
})
.catch(error => {
return Promise.reject({error});
})
}
// apiMiddleware.js
// api endpoint
const API_ENTRY = "https://my-api.com";
// utility functions for request preparation
export const makeFullUrl = (params) => {
// ...prepend endpoint url with API_ENTRY constant
return fullUrl
}
export const makeHeaders = (params) => {
// ...add auth token to headers, etc.
return headers;
}
export default store => next => action => {
const call = action[CALL_API];
if (call === undefined) {
return next(action);
}
const requestId = call.requestId;
store.dispatch(reqStart({requestId}));
const config = {
...call.config,
url: makeFullUrl(call.config),
headers: makeHeaders(call.config);
}
return axios(config)
.then(response => {
store.dispatch(reqEnd({
response,
requestId,
}))
return Promise.resolve(response);
})
.catch(error => {
store.dispatch(reqEnd({
error,
requestId,
}))
return Promise.reject(error);
})
}
// reducers.js
// Not included, but you can imagine reducers handle the
// above defined action types and update the state
// accordingly. Most usefully, components can always
// subscribe to specific api calls and check the request
// status. Showing loading indicators is one
// use case.
Here's the code I've implemented to accomplish a similar thing with redux-observable.
export const fetchCurrentUserEpic = (action$, state$) => {
const requestType = FETCH_CURRENT_USER;
const successType = RECEIVE_CURRENT_USER;
const requestConfig = {
url: "/current_user",
method: "get",
}
const payload = {requestConfig, requestType, successType};
const payloadNormalizer = ({response}) => {
return {currentUser: response.data.data};
}
return action$.ofType(FETCH_CURRENT_USER).pipe(
switchMap((action) => of({
type: CALL_API,
payload: {...payload, requestId: action.requestId, shouldFail: action.shouldFail, payloadNormalizer},
})),
)
}
export const apiEpic = (action$, state$) => {
return action$.ofType(CALL_API).pipe(
mergeMap((action) => (
concat(
of({type: REQ_START, payload: {requestId: action.payload.requestId, requestType: action.payload.requestType}}),
from(callApi(action.payload.requestConfig, action.payload.shouldFail)).pipe(
map(response => {
return {
type: action.payload.successType,
payload: action.payload.payloadNormalizer({response})
}
}),
map(() => {
return {
type: REQ_END,
payload: {status: 'success', requestId: action.payload.requestId, requestType: action.payload.requestType},
}
})
)
)
).pipe(
catchError(error => {
console.log('error', error);
return of({type: REQ_END, payload: {status: 'failed', requestId: action.payload.requestId, requestType: action.payload.requestType}, error});
})
)
)
)
}
Any comments or suggestions are appreciated!

I've found redux-fetch-epic-builder (A lib for building "fetch actions" and generic epics handled by redux-observable) to be similar to what you are trying to achieve here (beware it uses rxjs 5, this guide to rescue). It uses fetch, not axios, but it's easy to replace that. Plus it has transformers for successful/failed actions.
The library is a bit old, but the base idea to overcome boilerplate code is still valid: Generic epic-builder to fetch data with calls to API(s).
I am a novice in React / Redux / RxJS, but the only problem I see with the redux-fetch-epic-builder is the way to configure the client (in axios terms). That is, I am not fully satisfied with (due to it being not FSA or RSAA):
//action creators
const getComments = (id, page = 1) => ({
type: GET_COMMENTS,
host: 'http://myblog.com',
path: `/posts/${id}/comments`,
query: {
page,
},
})
// ...
const epics = [
buildEpic(GET_COMMENTS),
]
but this may still be an elegant way. And the license allow to develop the library further. I have not converted the example from the library documentation to your user-related example, but with react-observable there is certainly no need to introduce a separate "api middleware". (Also, I like /SUBACTION better than _SUBACTION, but it's trivial to change.)

Related

Handling errors with redux-toolkit

The information about the error in my case sits deeply in the response, and I'm trying to move my project to redux-toolkit. This is how it used to be:
catch(e) {
let warning
switch (e.response.data.error.message) {
...
}
}
The problem is that redux-toolkit doesn't put that data in the rejected action creator and I have no access to the error message, it puts his message instead of the initial one:
While the original response looks like this:
So how can I retrieve that data?
Per the docs, RTK's createAsyncThunk has default handling for errors - it dispatches a serialized version of the Error instance as action.error.
If you need to customize what goes into the rejected action, it's up to you to catch the initial error yourself, and use rejectWithValue() to decide what goes into the action:
const updateUser = createAsyncThunk(
'users/update',
async (userData, { rejectWithValue }) => {
const { id, ...fields } = userData
try {
const response = await userAPI.updateById(id, fields)
return response.data.user
} catch (err) {
if (!err.response) {
throw err
}
return rejectWithValue(err.response.data)
}
}
)
We use thunkAPI, the second argument in the payloadCreator; containing all of the parameters that are normally passed to a Redux thunk function, as well as additional options: For our example async(obj, {dispatch, getState, rejectWithValue, fulfillWithValue}) is our payloadCreator with the required arguments;
This is an example using fetch api
import { createSlice, createAsyncThunk } from "#reduxjs/toolkit";
export const getExampleThunk = createAsyncThunk(
'auth/getExampleThunk',
async(obj, {dispatch, getState, rejectWithValue, fulfillWithValue}) => {
try{
const response = await fetch('https://reqrefs.in/api/users/yu');
if (!response.ok) {
return rejectWithValue(response.status)
}
const data = await response.json();
return fulfillWithValue(data)
}catch(error){
throw rejectWithValue(error.message)
}
}
)
Simple example in slice:
const exampleSlice = createSlice({
name: 'example',
initialState: {
httpErr: false,
},
reducers: {
//set your reducers
},
extraReducers: {
[getExampleThunk.pending]: (state, action) => {
//some action here
},
[getExampleThunk.fulfilled]: (state, action) => {
state.httpErr = action.payload;
},
[getExampleThunk.rejected]: (state, action) => {
state.httpErr = action.payload;
}
}
})
Handling Error
Take note:
rejectWithValue - utility (additional option from thunkAPI) that you can return/throw in your action creator to return a rejected response with a defined payload and meta. It will pass whatever value you give it and return it in the payload of the rejected action.
For those that use apisauce (wrapper that uses axios with standardized errors + request/response transforms)
Since apisauce always resolves Promises, you can check !response.ok and handle it with rejectWithValue. (Notice the ! since we want to check if the request is not ok)
export const login = createAsyncThunk(
"auth/login",
async (credentials, { rejectWithValue }) => {
const response = await authAPI.signin(credentials);
if (!response.ok) {
return rejectWithValue(response.data.message);
}
return response.data;
}
);

Accessing the server response when an error occurs in my saga request

I'm using the library redux-saga-requests for network communication and high level state management. Following the guidance from the documentation, I've set up a global error handler to display a notification in case the server request times out or the user has connection troubles.
However, when I get a server error (500 for example), I can't access the response in my action. The server response is just not there.
Is there a way to access this object while still staying within the confines and rules of redux-saga?
Here's the initial store setup - note that onErrorSaga doesn't give me the network response, so I might as well go with the default behaviour - sending a suffixed _ERROR action that can be caught by my reducer:
function* onErrorSaga(error, action) {
console.log('error?', error)
console.log('action?', action)
// None of these contain the server response,
// just the error message and stack thrown by the client.
yield { error }
}
function* rootSaga(axiosInstance) {
yield createRequestInstance({
driver: {
default: createAxiosDriver(axiosInstance),
},
onError: onErrorSaga,
})
yield watchRequests()
}
const configureStore = initialState => {
const sagaMiddleware = createSagaMiddleware()
const middlewares = [thunkMiddleware, requestsPromiseMiddleware({ auto: true }), sagaMiddleware, trackingMiddleware]
const store = createStore(rootReducer, initialState, composeWithDevTools(applyMiddleware(...middlewares)))
sagaMiddleware.run(rootSaga, axiosInstance)
return store
}
export const configureStoreAsync = () => {
return new Promise((resolve, reject) => {
const store = configureStore()
store
.dispatch(fetchAppInitialization())
.then(result => resolve(store))
.catch(e => reject(store))
})
}
Here's how I'm currently handling errors in my state (which is then displayed by a notification component):
import { CLOSE_NOTIFICATION } from '../actions/actions.notification'
export default (state = [], action) => {
if (action.type.endsWith('_ERROR')) {
return {
type: 'warning',
title: action.error.message,
message: `${action.error.stack.substring(0, 200)}...`,
open: true,
timestamp: new Date(),
}
}
switch (action.type) {
case CLOSE_NOTIFICATION:
return {
...state,
open: false,
}
default:
return state
}
}
Turns out it was way easier than I thought. The action object does in fact contain the server response.
The path to the response is:
action.error.response.data

Proper way to handle errors in redux using redux-promise

I'm trying "redux-promise".
When there's no error in the flow, my code works properly. But, let's say that the API is down or I have a typo in the URL. In those cases, I expect to handle the error in the proper way.
This is the API: https://jsonplaceholder.typicode.com/users
(in the snippet I'm adding random text at the end to produce the 404)
Action creator
export async function fetchUsers() {
const request = await axios
.get('https://jsonplaceholder.typicode.com/userssdfdsfdsf')
.catch(error => console.log('ERROR', error))
return {
type: FETCHING_USERS,
payload: request
};
}
Reducer
export default (state = [], action) => {
switch (action.type) {
case FETCHING_USERS:
return [...state, ...action.payload.data]
default:
return state
}
}
I can see the error logged in the console
ERROR Error: Request failed with status code 404
But, once the action is dispatched its payload is undefined
action {type: "FETCHING_USERS", payload: undefined}
I don't know where is the best place to handle this: action creator, reducer, etc. I shouldn't check if payload is something in the reducer and, if not, return state or do nothing. I want to understand which would be the best approach to handle this.
You may look at source of redux-promise, as it very simple.
Redux-promise expects either promise or action with payload set to some promise. I think you're going to use leter case.
Code may look like (just example, not tested):
export function fetchUsers() {
const request = axios.get('https://jsonplaceholder.typicode.com/userssdfdsfdsf');
return {
type: FETCHING_USERS,
payload: request
};
}
In this case redux-promise will await for resolution of promise returned by axios.get and dispatch your action but payload replaced with promise result. In case of error, redux-promise will catch it and dispatch action with error = true (you may want to handle action.error === true case in reducer)
In the reducer you should check for the existence of the error field in action:
export default function(state = null, action) {
if (action.error) {
//handle
}
switch(action.type) {
This code is for the action.
export const fetchUsers = () => async dispatch => {
try {
const res = await axios.get('https://jsonplaceholder.typicode.com/userssdfdsfdsf');
dispatch({
type: FETCHING_USERS,
payload: res.data
});
} catch (err) {
dispatch({
type: FETCHING_ERROR
});
}
};
In reducer do this.
Define the initial state and use this code.
export default function(state = initialState, action) {
const { type, payload } = action;
switch (type) {
case FETCHING_USERS:
return {
...state,
loading: false,
user: payload,
};
case FETCH_ERROR:
return {
...state,
token: null,
loading: false,
};

Redux: Reusable actions

I have an action, that uses a redux thunk, that looks like so:
export function fetchData(query) {
return dispatch => {
return fetch(`http://myapi?query=${query}` ,{mode: 'cors'})
.then(response => response.json())
.then(json => { dispatch(someOtherAction(json)) })
}
}
}
and then my someOtherAction actually updates state:
export function someOtherAction(data) {
return {
action: types.SOME_ACTION,
data
}
}
But i want it to be possible for the fetchData action creator to be reusable so that different parts of my app can fetch data from myapi and then have different parts of the state based on that.
I'm wondering what is the best way to reuse this action? Is it acceptable to pass a second parameter in to my fetchData action creator that stipulates which action is called on a successful fetch:
export function fetchData(query, nextAction) {
return dispatch => {
return fetch(`http://myapi?query=${query}` ,{mode: 'cors'})
.then(response => response.json())
.then(json => { dispatch(nextAction(json)) })
}
}
}
Or is there an accepted way of doing this sort of thing?
I use a middleware for that. I have defined the fetch call in there, then in my actions I send the URL to fetch and the actions to dispatch when completed. This would be a typical fetch action:
const POSTS_LOAD = 'myapp/POST_L';
const POST_SUCCESS = 'myapp/POST_S';
const POST_FAIL = 'myapp/POST_F';
export function fetchLatestPosts(page) {
return {
actions: [POSTS_LOAD, POST_SUCCESS, POST_FAIL],
promise: {
url: '/some/path/to/posts',
params: { ... },
headers: { ... },
},
};
}
When calling that action, the POST_LOAD action will be dispatch automatically by the middleware just before the fetch request it's executed. If everything goes well the POST_SUCCESS action will be dispatched with the json response, if something goes wrong the POST_FAIL action will be dispatched by the middleware.
All the magic it's in the middleware! And it's something similar to this:
export default function fetchMiddleware() {
return ({ dispatch, getState }) => {
return next => action => {
if (typeof action === 'function') {
return action(dispatch, getState);
}
const { promise, actions, ...rest } = action;
if (!promise) {
return next(action);
}
const [REQUEST, SUCCESS, FAILURE] = actions;
next({ ...rest, type: REQUEST }); // <-- dispatch the LOAD action
const actionPromise = fetch(promise.url, promise); // <-- Make sure to add the domain
actionPromise
.then(response => response.json())
.then(json => next({ ...rest, json, type: SUCCESS })) // <-- Dispatch the success action
.catch(error => next({ ...rest, error, type: FAILURE })); // <-- Dispatch the failure action
return actionPromise;
};
};
}
This way I have all my requests on a single place and I can define the actions to run after the request it's completed.
------------EDIT----------------
In order to get the data on the reducer, you need to use the action name you defined on the original action creator. The following example shows how to handle the POST_SUCCESS action from the middleware to get the posts data from the json response.
export function reducer(state = {}, action) {
switch(action.type) {
case POST_SUCCESS: // <-- Action name
return {
...state,
posts: action.json.posts, // <-- Getting the data from the action
}
default:
return state;
}
}
I hope this helps!

Why does redux-mock-store don't show an action dispatched in catch promises?

I'm very bad when it comes to thinking of a title question, sorry for that.
My Problem:
I'm unit testing my async redux actions like it's suggested in the docs. I mock the API calls with nock and check for the dispatched actions with redux-mock-store. It works great so far, but I have one test that fails even though it clearly does work. The dispatched action neither does show up in the array returned by store.getActions() nor is the state changed in store.getState(). I'm sure that it does happen because I can see it when I test manually and observe it with Redux Dev Tools.
The only thing that is different in this action dispatch is that it is called in a promise in a catch of another promise. (I know that sounds confusing, just look at the code!)
What my code looks like:
The action:
export const login = (email, password) => {
return dispatch => {
dispatch(requestSession());
return httpPost(sessionUrl, {
session: {
email,
password
}
})
.then(data => {
dispatch(setUser(data.user));
dispatch(push('/admin'));
})
.catch(error => {
error.response.json()
.then(data => {
dispatch(setError(data.error))
})
});
};
}
This httpPost method is just a wrapper around fetch that throws if the status code is not in the 200-299 range and already parses the json to an object if it doesn't fail. I can add it here if it seems relevant, but I don't want to make this longer then it already is.
The action that doesn't show up is dispatch(setError(data.error)).
The test:
it('should create a SET_SESSION_ERROR action', () => {
nock(/example\.com/)
.post(sessionPath, {
session: {
email: fakeUser.email,
password: ''
}
})
.reply(422, {
error: "Invalid email or password"
})
const store = mockStore({
session: {
isFetching: false,
user: null,
error: null
}
});
return store.dispatch(actions.login(
fakeUser.email,
""))
.then(() => {
expect(store.getActions()).toInclude({
type: 'SET_SESSION_ERROR',
error: 'Invalid email or password'
})
})
});
Thanks for even reading.
Edit:
The setErroraction:
const setError = (error) => ({
type: 'SET_SESSION_ERROR',
error,
});
The httpPostmethod:
export const httpPost = (url, data) => (
fetch(url, {
method: 'POST',
headers: createHeaders(),
body: JSON.stringify(data),
})
.then(checkStatus)
.then(response => response.json())
);
const checkStatus = (response) => {
if (response.status >= 200 && response.status < 300) {
return response;
}
const error = new Error(response.statusText);
error.response = response;
throw error;
};
Because of you are using nested async function in catch method - you need to return the promise:
.catch(error => {
return error.response.json()
.then(data => {
dispatch(setError(data.error))
})
});
Otherwise, dispatch will be called after your assertion.
See primitive examples:
https://jsfiddle.net/d5fynntw/ - Without returning
https://jsfiddle.net/9b1z73xs/ - With returning

Resources