Previously I worked with Express a bit and now learning Koa.
In Express, when I did a simple get method. The following code works
.get('/applications', (req, res, next) => {
const getTable = `SELECT * FROM applicationtable`
db.query(getTable, values, (err, result) => {
if (err) {
console.log(err)
}else {
res.json(result)
}
})
})
However when I tried to do the same thing in Koa with
.get('/applications', (ctx, next) => {
const getTable = `SELECT * FROM applicationtable`
db.query(getTable, values, (err, result) => {
if (err) {
console.log(err)
}else {
ctx.body = result
}
})
})
It returns a 404 error. I have to explicitly wrap the above one in Koa with a promise and put async await then only it will work, as shown below:
.get('/applications', async(ctx, next) => {
const getTable = `SELECT * FROM applicationtable`
const item = await new Promise(function(resolve, reject) {
db.query(getTable, (err, result) => {
if (err) {
reject(err)
}else {
resolve(result)
}
})
})
ctx.body = item;
})
My question is , why in Express the code does not have to be explicitly wrapped in async await? In what way Koa is behaving differently than Express with the examples here?
Koa is all designed around promises. The request handler itself is async which means it returns a promise and various things happen in the framework when that promise resolves. As such, you can't freely mix plain asynchronous callbacks with promises.
In the case of your second code block, your request handler returns (which resolves the async promise) BEFORE your asynchronous callback gets called and thus before you set ctx.body. So, Koa goes to process the request (thinking you are done), but ctx.body hasn't been set yet so it figures this must be a 404.
Wrapping your async operation in a promise and using await with it, then chains it into the main request handler promise so that promise doesn't resolve until your asynchronous operation is done (what you want).
Bottom line, use promises that are linked into the request handler (either with await or by returning a promise) for all your asynchronous operations in the request handler.
Here's a nice little example in the Koa developer's guide: https://github.com/koajs/koa/blob/master/docs/guide.md#async-operations
Express, on the other hand, does not do anything when your request handler returns. It doesn't do anything until you either call next() to continue looking for matches with other route handlers or call res.send() or some similar API call that sends a response so you're free to do that in a plain asynchronous callback like you show.
This is part of the core design difference between Express and Koa.
Related
I'm new to node and banging my head against a wall on what should be a simple node+express+googlecal+pug issue
node/express route accepts requests and calls controller
controller ensures validation of auth and then...
executes a successful gcal function...console.log has the data i need
trying to directly (in controller function) returns "Cannot set headers after they are sent to the client"....why is a call to Gcal API forcing a response back to client?
Trying to make it more micro via individual calls to each function results in same result
What am I missing here?
getcalendars: async function(oAuth2Client, res) {
const calendar = google.calendar({ version: "v3", auth: oAuth2Client });
cal = await calendar.calendarList.list(
{},
(err, result) => {
//console.log("HEADERS SENT1?: "+res.headersSent);
if (err) {
console.log('The API returned an error: ' + err);
return;
}
console.log(JSON.stringify(result));
message2 = JSON.stringify(result)
res.render('schedules', {message2: message2})
return
});
},
EDIT: Calling function
router.route('/dashboard/schedules')
.get(async function(req, res) {
if (req.session.loggedin) {
//x = gcalController.getcalendars(req, res);
token = await gcalController.gettoken(req, res);
isAuth = await gcalController.calauth(token);
listcalendars = await gcalController.getcalendars(isAuth,res);
} else {
res.redirect("/?error=failedAuthentication")
//res.send('Please login to view this page!');
}
});
Can't set headers already sent happens when you're sending a response more than once. Usually you can terminate the function by returning your res.send() call.
It looks like the express middleware that created the res object is sending a response by the time your res.render() gets pulled out of the microtask queue.
Can you show the full code? It seems that this is probably originating in the scope where getcalendars is called.
Question: Using Express is it possible to return an error from inside an IIFE and advance to my error handling middleware?
Background: The IIFE is used to create an async container to wrap await statements. I can't see a way out of this and I wonder if I'm using the wrong basic, pattern altogether.
Simplified Example:
app.get('/', function(req, res, next) {
(async function() {
try {
let example = await someLogic(x);
} catch(err) {
return next(new Error('oops'));
}
})();
console.log('main endpoint');
});
app.use(function(err, req, res, next) {
console.log('my error', err.message);
});
Using Express is it possible to return an error from inside an IIFE and advance to my error handling middleware?
Yes, that works fine. It will call next(err) just fine. But, your return will return only from the IIFE and the rest of your request handler after the try/catch will still execute (not sure if you want that or not).
FYI, it's probably simpler to declare the request handler as async and then you don't need the IIFE wrapper:
app.get('/', async function(req, res, next) {
try {
let example = await someLogic(x);
console.log('main endpoint');
// send some response here
} catch(err) {
return next(new Error('oops'));
}
});
How does the the error handler get triggered?
In sample code, I find that it is placed at the bottom of all the middleware functions. Is the position important?
You can refer below example for some details.
Here, for the '/' GET endpoint a middleware explicitly throws an error 'problem error'.
At this point, express error handler mechanism is triggered and it looks for an error handler (with err as a param). As a result, subsequent 'Hello' is not sent back to the client as its handler is not an error one.
Subsequent error handlers logErrors, clientErrorHandler and errorHandler are invoked one by one to perform relevant tasks with final one writing back the response.
The reason they are placed at the end is to catch the errors thrown by declared middlewares and handle them elegantly i.e. printing it, logging it, sending mails etc. Think of it in terms of a try catch mechanism in other languages like Java. If declared above other middlewares, they would render useless as the errors won't be handled. You can see the difference in output by swapping the order of 'GET' request with the error handlers.
const express = require('express');
const app = express();
app.get('/', (req, res, next) => next(new Error('problem error')), (req, res) => {
res.status(200).send("Hello");
});
app.use(logErrors);
app.use(clientErrorHandler);
app.use(errorHandler);
function logErrors (err, req, res, next) {
console.error(err.stack)
next(err)
}
function clientErrorHandler (err, req, res, next) {
if (req.xhr) {
res.status(500).send({ error: 'Something failed!' })
} else {
next(err)
}
}
function errorHandler (err, req, res, next) {
if (res.headersSent) {
return next(err)
}
res.status(500)
res.render('error', { error: err })
}
app.listen(3000, () => console.log('Example app listening on port 3000!'))
I actually have a couple of questions about Express middleware chaining and thought I'd group them in this single post since they're closely related.
Q1: In a middleware function, if I send() the Response, pass to the next function, or just return, does any of the remaining code get executed?
router.get('/a', (req, res, next) => {
res.sendStatus(200);
// Does this ever get executed?
});
router.get('/b', (req, res, next) => {
next();
// Does this ever get executed?
});
router.get('/c', (req, res, next) => {
return;
// Does this stop the middleware chain? What happens to the request/response?
});
Q2: Can you pass on to another middleware function from inside of the preceding one? My reason for this is that I have to supply an argument to create a certain middleware, and that argument is dependent on which Express app object is the owner of the Request. So I need to access a property of the Request before it is passed to the middleware that needs to be created using that property. Would this work: ?
router.get('/d', (req, res, next) => {
var middlewareFunction = createDynamicMiddleware();
middlewareFunction(req, res, next); // Will this still work as expected?
});
Q3: How are the Request/Response arguments passed to the next middleware when you call next() without any arguments?
Q1: In a middleware function, if I send() the Response, pass to the next function, or just return, does any of the remaining code get
executed?
router.get('/a', (req, res, next) => {
res.sendStatus(200);
// Does this ever get executed?
// Yes, but don't call next() or res.send() or it will throw an error
});
router.get('/b', (req, res, next) => {
next();
// Does this ever get executed?
// Yes, but don't call next() or res.send() or it will throw an error
});
router.get('/c', (req, res, next) => {
return;
// Does this stop the middleware chain? What happens to the request/response?
// The request get stuck and after a while it will fail due 'timeout'
});
Q2: Can you pass on to another middleware function from inside of the preceding one? My reason for this is that I have to supply an
argument to create a certain middleware, and that argument is
dependent on which Express app object is the owner of the Request. So
I need to access a property of the Request before it is passed to the
middleware that needs to be created using that property.
Your factory of dynamic middleware should return a new function
Functions that returns new functions, are also knows as curried functions, here is a practical example:
// based on 'someId' we will return a specific middleware
req.get('/user/:someId',
createDynamicMiddleware,
(req,res) => {
res.send('ok')
})
function createDynamicMiddleware(req,res,next){
if(req.params.someId === 'me') {
return function(req, res, next){
// do some stuff if someId is me
return next();
}
} else {
return function(req, res, next){
// do some other stuff otherwise
return next();
}
}
}
Q3: How are the Request/Response arguments passed to the next middleware when you call next() without any arguments?
Express handle this for you!
Also, if you add a property to req or res objects, it will be present in the next middlewares
For example:
router.get('/', (req,res,next) => {
req.customProperty = 'hello!';
return next();
}, (req, res) => {
console.log(req.myCustomProperty)
// hello!
})
I'm looking for a clean way to have my express app return 405 Method Not Allowed if a client sends a request that matches a mapped url route but does not match the mapped HTTP method.
My current implementation is to have a default "catch-all" handler that tries to match the url against the register routes, ignoring the HTTP method. If there is a match, then we know to return a 405, otherwise we let express do its default 404 behavior.
I'm hoping there is a better way that doesn't involve running all the route matching twice (once by express, once by my handler).
Here is an approach that I have used successfully with multiple Django applications and now with Node and Express. It is also follows RFC 2616 (HTTP/1.1) that says the following about HTTP 405:
The response MUST include an Allow header containing a list of valid
methods for the requested resource.
So, the key point is to route the requests to the same handler without regard to methods.
app.all('/page/:id', page.page);
app.all('/page/:id/comments', page.comments);
app.all('/page/:id/attachments', page.attachments);
...
The next point is to validate the method in the handler function 'comments'. Note that the handler is responsible for handling all the methods. In Django's world this is the only way to go because the framework forces you to separate the routing of the URLs from the actual action about to be performed against the resource the URL represents.
In the handler you could check the method like this...
exports.comments = function (req, res) {
if (req.route.method === 'get') {
res.send(200, 'Hello universe.');
} else {
res.set('Allow', 'GET');
res.send(405, 'Method Not Allowed');
}
}
...but as you can expect the code will quickly become repetitious and not nice to read especially when you have many handler functions and many different sets of allowed methods.
Therefore I prepared a shortcut function named restful for the job. Define the function wherever you want. I personally would place it in helpers.js under the same directory where the handler functions are implemented.
var restful = function (req, res, handlers) {
//
// This shortcut function responses with HTTP 405
// to the requests having a method that does not
// have corresponding request handler. For example
// if a resource allows only GET and POST requests
// then PUT, DELETE, etc requests will be responsed
// with the 405. HTTP 405 is required to have Allow
// header set to a list of allowed methods so in
// this case the response has "Allow: GET, POST" in
// its headers [1].
//
// Example usage
//
// A handler that allows only GET requests and returns
//
// exports.myrestfulhandler = function (req, res) {
// restful(req, res, {
// get: function (req, res) {
// res.send(200, 'Hello restful world.');
// }
// });
// }
//
// References
//
// [1] RFC-2616, 10.4.6 405 Method Not Allowed
// https://www.rfc-editor.org/rfc/rfc2616#page-66
//
// [2] Express.js request method
// http://expressjs.com/api.html#req.route
//
var method = req.route.method; // [2]
if (!(method in handlers)) {
res.set('Allow', Object.keys(handlers).join(', ').toUpperCase());
res.send(405);
} else {
handlers[method](req, res);
}
}
With restful it is now quite painless to handle 405 responses automatically and having proper Allow header being set. Just give a function for each method you allow and restful does the rest.
So lets modify the previous example:
exports.comments = function (req, res) {
restful(req, res, {
get: function (req, res) {
res.send(200, 'Hello restful universe.');
}
});
}
Why the name restful? In RESTful web it is quite essential for the API to obey the conventions like responsing with HTTP 405 to the request having non-supported method. Many of those conventions could be integrated to restful when needed. Therefore the name is restful and not something like auto405 or http405handler.
Hope this helps. Any thoughts?
Method 1: Use .route() and .all()
// Your route handlers
const handlers = require(`./handlers.js`);
// The 405 handler
const methodNotAllowed = (req, res, next) => res.status(405).send();
router
.route(`/products`)
.get(handlers.getProduct)
.put(handlers.addProduct)
.all(methodNotAllowed);
This works because requests are passed to the handlers in the order they are attached to the route (the request "waterfall"). The .get() and .put() handlers will catch GET and PUT requests, and the rest will fall through to the .all() handler.
Method 2: Middleware
Create middleware which checks for allowed methods, and returns a 405 error if the method is not whitelisted. This approach is nice because it allows you to see and set the allowed methods for each route along with the route itself.
Here's the methods.js middleware:
const methods = (methods = ['GET']) => (req, res, next) => {
if (methods.includes(req.method)) return next();
res.error(405, `The ${req.method} method for the "${req.originalUrl}" route is not supported.`);
};
module.exports = methods;
You would then use the methods middleware in your routes like this:
const handlers = require(`./handlers.js`); // route handlers
const methods = require(`./methods.js`); // methods middleware
// allows only GET or PUT requests
router.all(`/products`, methods([`GET`, `PUT`]), handlers.products);
// defaults to allowing GET requests only
router.all(`/products`, methods(), handlers.products);
Due to ambiguity, there really is no other way. Personally, I would do something like this:
var route = '/page/:id/comments'
app.get(route, getComments)
app.all(route, send405)
function send405(req, res, next) {
var err = new Error()
err.status = 405
next(err)
}
Either way, you have to check the routes twice.
Kinda old question but here is what i did. I just put this after all my routes but before my 400 handler
// Handle 405 errors
app.use(function(req, res, next) {
var flag = false;
for (var i = 0; i < req.route.stack.length; i++) {
if (req.method == req.route.stack[i].method) {
flag = true;
}
}
if (!flag) {
err = new Error('Method Not Allowed')
err.status = 405;
return next(err)
}
next();
});
I have been doing it this way:
Say if you have GET and POST method handlers for /. You can wrap the path with app.route or router.route and assign the handlers accordingly.
app.route("/").get((req, res) => {
/* DO SOMETHING*/
}).post((req, res) => {
/* DO SOMETHING*/
}).all((req, res) => {
res.status(405).send();
});
http://expressjs.com/en/4x/api.html#app.route
http://expressjs.com/en/4x/api.html#router.route
A request will get matched to the route and filtered through the handlers. If a handler is present, it will get handled as usual. Else, it will reach the all handler that will set the status code to 405 and ending the request.
I fixed it like this :
/*paths here*/
router.get('/blah/path1', blah.do_something );
router.post('/blah/path2', blah.do_something_else );
/* if we get here we haven't already gone off down another path */
router.all('/*', (req,res) => { res.status(405),
res.json({'status':405,
'message':req.method + ' not allowed on this route'})
});
/* simples */
I thought this was a pretty interesting problem and so I dove deeeeeep down into the depths of the express app function and found a way to dynamically build a 405 error that includes all of the possible routes (without having to manually update anything when you add a new route).
app.use("", (req, _, next) => {
const err = buildError(app, req);
if (!err) return next();
return next(err);
});
For those interested you can find the npm package here https://www.npmjs.com/package/express-ez-405, and below is a quick example of what it looks like to use it.
const express = require("express");
const { buildError } = require("express-ez-405");
const app = express();
app.use(express.json());
const userRouter = require("./routes/user");
const mainRouter = require("./routes/main");
const nestedRouter = require("./routes/nested");
// // Routes
app.use("/main", mainRouter);
app.use("/user", userRouter);
app.use("/nested/route", nestedRouter);
// Routes
// 405 & 404 error catcher
app.use("", (req, _, next) => {
const err = buildError(app, req);
if (!err) return next();
return next(err);
});
// 405 & 404 error catcher
// Error handling
app.use((err, _, res, __) =>
res.status(err.status).json({ message: err.message })
);
// Error handling
const PORT = 3000;
app.listen(PORT, () => {
console.log(`Server listening on port: ${PORT}...`);
});
Once this in in there you never have to worry about updating it regardless of whether you add, remove, or change routing.