Ok,
So I am using the puppeteer framework and I have an async function that interact with a webpage.
This function clicks and selects and elements of a webpage while it waiting for the traffic of the page to be idle.
This function works most of the time, but sometimes it stalls.
I want to be able to set a timeout so that if the function is taking longer than a certain amount of time, it throws an error and I can run it again.
So far I cannot seem to get this to work because I cannot get the callback function I pass to setTimeOut() to 'interact' with the outer function.
My code looks like this:
const scrap_webtite = async page => {
/* scrap the site */
try{ // catch all
// set timeout
let timed_out_ID = setTimeout(()=> throw "timeOut", 1000);
// run the async
let el = await sometimes_stalls_function(page);
// if function ran finished correcly
clearTimeout(timed_out_ID);
// save el
save_el(el);
}
}catch(e){
console.error("Something went wrong!", e);
// this makes the function run again
// here is where I want to ideally catch the timeout error
return false
}
}
I have also tried wrapping the setTimeOut function in an Promise as per this post and the using the .then().catch() callbacks to try to catch the error to no avail.
Apologies if this is a stupid question, thank for you help.
The problem you're running into is essentially that the error thrown in setTimeout() is not related to your function flow, and thus can't be caught there. You can essentially think of the timer's callback function as a "detached" function: the variables from the parent scope will still be available, but you can't return a value to the parent directly etc.
To work around this problem you have a few options, Promise.race() is one possible solution. The idea is to first make an async version of a timeout:
const rejectAfter = (timeout) => {
return new Promise((resolve, reject) => {
setTimeout(() => reject(), timeout);
});
};
Then extract your business logic out into a separate async function a-la:
const doTheThing = async () => {
// TODO: Implement
};
And finally in your scraping function, use Promise.race() to use the result from whichever of the two finishes first:
const scrape = async (page) => {
try {
const el = await Promise.race([
rejectAfter(1000),
doTheThing()
]);
} catch(error) {
// TODO: Handle error
}
}
try turning everything in the try block into a promise
const scrap_webtite = async page => {
/* scrap the site */
try{ // catch all
return await new Promise(async(r,j)=>{
// set timeout
let timed_out_ID = setTimeout(()=>j("timeOut"),1000);
// run the async
let el = await sometimes_stalls_function(page);
// if function ran finished correcly
clearTimeout(timed_out_ID);
// save el
r(save_el(el));
})
}catch(e){
console.error("Something went wrong!", e);
// this makes the function run again
// here is where I want to ideally catch the timeout error
return false
}
}
I have a Selenium webdriverIO V5 framework. The issue I am facing here is, the below code works fine on Mac OS, but it does not work correctly on the Windows OS. In the Windows OS it gets stuck with an infinite loop issue.
The below code functionality is: Merge yaml files (which contains locators) and return the value of the locator by passing the key:
const glob = require('glob');
const yamlMerge = require('yaml-merge');
const sleep = require('system-sleep');
let xpath;
class Page {
getElements(elementId) {
function objectCollector() {
glob('tests/wdio/locators/*.yml', function (er, files) {
if (er) throw er;
xpath = yamlMerge.mergeFiles(files);
});
do {
sleep(10);
} while (xpath === undefined);
return xpath;
}
objectCollector();
return xpath[elementId];
}
}
module.exports = new Page();
Since you are waiting on the results of a callback, I would recommend returning a new Promise from your getElements function and resolve() the value you receive inside the callback. Then when you call getElements, you will need to resolve that Promise or use the await notation. The function will stop at that point and wait until the Promise resolves, but the event loop will still continue. See some documentation for more information.
I'll write an example below of what your code might look like using a Promise, but when you call getElements, you will need to put the keyword await before it. If you want to avoid that, you could resolve the Promise from objectCollector while you're in getElements and remove the async keyword from its definition, but you really should not get in the way of asynchronous JavaScript. Also, you can probably shorten the code a bit because objectCollector looks like an unnecessary function in this example:
const glob = require('glob')
const yamlMerge = require('yaml-merge')
const sleep = require('system-sleep')
let xpath
class Page {
function async getElements(elementId) {
function objectCollector() {
return new Promise((resolve,reject) => {
glob('tests/wdio/locators/*.yml', function (er, files) {
if (er) reject(er)
resolve(yamlMerge.mergeFiles(files))
})
})
}
let xpath = await objectCollector()
return xpath[elementId]
}
}
module.exports = new Page();
I am attempting to read a stream line by line and for each line do some async processing. The issue I'm having is how to determine when the operations for all lines are complete. I thought I could use the readline "close" event but that seems to be triggered long before the async operations I've started on the "line" event complete.
Here's the event handlers in question:
logReader.on("line", inputLine => {
self._processLogLine(inputLine, () => {
if (self.streamFinished) {
completionCallback(err);
}
});
});
logReader.on("close", () => {
self.streamFinished = true;
});
The completionCallback method should be called when all line processing on the stream is done.
I've thought about adding counters for line operations started/completed and calling the completionCallback when they match but that seems awkward.
Here is one way you could do it basically keeping track of each lines success. I don't have enough of your code to test it...
const lineProcessing = [];
const doCompletion = () => {
if (self.streamFinished && !lineProcessing.includes(false)) {
// Not sure where 'err' comes from
completionCallback(err);
}
}
logReader.on("line", inputLine => {
lineProcessing.push(false);
let processingIndex = lineProcessing.length - 1;
self._processLogLine(inputLine, () => {
lineProcessing[processingIndex] = true;
doCompletion();
});
});
logReader.on("close", () => {
self.streamFinished = true;
});
It's a little difficult to tell for sure what needs to happen without more context. But here's a quick and somewhat dirty fix that should be adaptable to your situation. It shows how you can accomplish this with promises:
let promises = []
logReader.on("line", inputLine => {
promises.push(new Promise(resolve => {
self._processLogLine(inputLine, resolve)
})
})
logReader.on("close", () => {
Promise.all(promises)
.then(completionCallback)
})
Essentially, just create an array of promises. When you know that all promises have been added to the array, call Promise.all() on it.
If you use this approach, you'd probably want to add error checking (rejecting and catching).
Look into generators. They give you a way to do a "for" loop...and when you're done, you just return from the generator function.
So you can have a generator to read the file, and a function that uses the generator, then when the loop ends, call your callback.
Can anyone recommend a pattern for instantly retrieving data from a function that returns a Promise?
My (simplified) example is an AJAX preloader:
loadPage("index.html").then(displayPage);
If this is downloading a large page, I want to be able to check what's happening and perhaps cancel the process with an XHR abort() at a later stage.
My loadPage function used to (before Promises) return an id that let me do this later:
var loadPageId = loadPage("index.html",displayPage);
...
doSomething(loadPageId);
cancelLoadPage(loadPageId);
In my new Promise based version, I'd imagine that cancelLoadPage() would reject() the original loadPage() Promise.
I've considered a few options all of which I don't like. Is there a generally accepted method to achieve this?
Okay, let's address your bounty note first.
[Hopefully I'll be able to grant the points to someone who says more than "Don't use promises"... ]
Sorry, but the answer here is: "Don't use promises". ES6 Promises have three possible states (to you as a user): Pending, Resolved and Rejected (names may be slightly off).
There is no way for you to see "inside" of a promise to see what has been done and what hasn't - at least not with native ES6 promises. There was some limited work (in other frameworks) done on promise notifications, but those did not make it into the ES6 specification, so it would be unwise of you to use this even if you found an implementation for it.
A promise is meant to represent an asynchronous operation at some point in the future; standalone, it isn't fit for this purpose. What you want is probably more akin to an event publisher - and even that is asynchronous, not synchronous.
There is no safe way for you to synchronously get some value out of an asynchronous call, especially not in JavaScript. One of the main reasons for this is that a good API will, if it can be asynchronous, will always be asynchronous.
Consider the following example:
const promiseValue = Promise.resolve(5)
promiseValue.then((value) => console.log(value))
console.log('test')
Now, let's assume that this promise (because we know the value ahead of time) is resolved synchronously. What do you expect to see? You'd expect to see:
> 5
> test
However, what actually happens is this:
> test
> 5
This is because even though Promise.resolve() is a synchronous call that resolves an already-resolved Promise, then() will always be asynchronous; this is one of the guarantees of the specification and it is a very good guarantee because it makes code a lot easier to reason about - just imagine what would happen if you tried to mix synchronous and asynchronous promises.
This applies to all asynchronous calls, by the way: any action in JavaScript that could potentially be asynchronous will be asynchronous. As a result, there is no way for you do any kind of synchronous introspection in any API that JavaScript provides.
That's not to say you couldn't make some kind of wrapper around a request object, like this:
function makeRequest(url) {
const requestObject = new XMLHttpRequest()
const result = {
}
result.done = new Promise((resolve, reject) => {
requestObject.onreadystatechange = function() {
..
}
})
requestObject.open(url)
requestObject.send()
return requestObject
}
But this gets very messy, very quickly, and you still need to use some kind of asynchronous callback for this to work. This all falls down when you try and use Fetch. Also note that Promise cancellation is not currently a part of the spec. See here for more info on that particular bit.
TL:DR: synchronous introspection is not possible on any asynchronous operation in JavaScript and a Promise is not the way to go if you were to even attempt it. There is no way for you to synchronously display information about a request that is on-going, for example. In other languages, attempting to do this would require either blocking or a race condition.
Well. If using angular you can make use of the timeout parameter used by the $http service if you need to cancel and ongoing HTTP request.
Example in typescript:
interface ReturnObject {
cancelPromise: ng.IPromise;
httpPromise: ng.IHttpPromise;
}
#Service("moduleName", "aService")
class AService() {
constructor(private $http: ng.IHttpService
private $q: ng.IQService) { ; }
doSomethingAsynch(): ReturnObject {
var cancelPromise = this.$q.defer();
var httpPromise = this.$http.get("/blah", { timeout: cancelPromise.promise });
return { cancelPromise: cancelPromise, httpPromise: httpPromise };
}
}
#Controller("moduleName", "aController")
class AController {
constructor(aService: AService) {
var o = aService.doSomethingAsynch();
var timeout = setTimeout(() => {
o.cancelPromise.resolve();
}, 30 * 1000);
o.httpPromise.then((response) => {
clearTimeout(timeout);
// do code
}, (errorResponse) => {
// do code
});
}
}
Since this approach already returns an object with two promises the stretch to include any synchronous operation return data in that object is not far.
If you can describe what type of data you would want to return synchronously from such a method it would help to identify a pattern. Why can it not be another method that is called prior to or during your asynchronous operation?
You can kinda do this, but AFAIK it will require hacky workarounds. Note that exporting the resolve and reject methods is generally considered a promise anti-pattern (i.e. sign you shouldn't be using promises). See the bottom for something using setTimeout that may give you what you want without workarounds.
let xhrRequest = (path, data, method, success, fail) => {
const xhr = new XMLHttpRequest();
// could alternately be structured as polymorphic fns, YMMV
switch (method) {
case 'GET':
xhr.open('GET', path);
xhr.onload = () => {
if (xhr.status < 400 && xhr.status >= 200) {
success(xhr.responseText);
return null;
} else {
fail(new Error(`Server responded with a status of ${xhr.status}`));
return null;
}
};
xhr.onerror = () => {
fail(networkError);
return null;
}
xhr.send();
return null;
}
return xhr;
case 'POST':
// etc.
return xhr;
// and so on...
};
// can work with any function that can take success and fail callbacks
class CancellablePromise {
constructor (fn, ...params) {
this.promise = new Promise((res, rej) => {
this.resolve = res;
this.reject = rej;
fn(...params, this.resolve, this.reject);
return null;
});
}
};
let p = new CancellablePromise(xhrRequest, 'index.html', null, 'GET');
p.promise.then(loadPage).catch(handleError);
// times out after 2 seconds
setTimeout(() => { p.reject(new Error('timeout')) }, 2000);
// for an alternative version that simply tells the user when things
// are taking longer than expected, NOTE this can be done with vanilla
// promises:
let timeoutHandle = setTimeout(() => {
// don't use alert for real, but you get the idea
alert('Sorry its taking so long to load the page.');
}, 2000);
p.promise.then(() => clearTimeout(timeoutHandle));
Promises are beautiful. I don't think there is any reason that you can not handle this with promises. There are three ways that i can think of.
The simplest way to handle this is within the executer. If you would like to cancel the promise (like for instance because of timeout) you just define a timeout flag in the executer and turn it on with a setTimeout(_ => timeout = true, 5000) instruction and resolve or reject only if timeout is false. ie (!timeout && resolve(res) or !timeout && reject(err)) This way your promise indefinitely remains unresolved in case of a timeout and your onfulfillment and onreject functions at the then stage never gets called.
The second is very similar to the first but instead of keeping a flag you just invoke reject at the timeout with proper error description. And handle the rest at the then or catch stage.
However if you would like to carry the id of your asych operation to the sync world then you can also do it as follows;
In this case you have to promisify the async function yourself. Lets take an example. We have an async function to return the double of a number. This is the function
function doubleAsync(data,cb){
setTimeout(_ => cb(false, data*2),1000);
}
We would like to use promises. So normally we need a promisifier function which will take our async function and return another function which when run, takes our data and returns a promise. Right..? So here is the promisifier function;
function promisify(fun){
return (data) => new Promise((resolve,reject) => fun(data, (err,res) => err ? reject(err) : resolve(res)));
}
Lets se how they work together;
function promisify(fun){
return (data) => new Promise((resolve,reject) => fun(data, (err,res) => err ? reject(err) : resolve(res)));
}
function doubleAsync(data,cb){
setTimeout(_ => cb(false, data*2),1000);
}
var doubleWithPromise = promisify(doubleAsync);
doubleWithPromise(100).then(v => console.log("The asynchronously obtained result is: " + v));
So now you see our doubleWithPromise(data) function returns a promise and we chain a then stage to it and access the returned value.
But what you need is not only a promise but also the id of your asynch function. This is very simple. Your promisified function should return an object with two properties; a promise and an id. Lets see...
This time our async function will return a result randomly in 0-5 secs. We will obtain it's result.id synchronously along with the result.promise and use this id to cancel the promise if it fails to resolve within 2.5 secs. Any figure on console log Resolves in 2501 msecs or above will result nothing to happen and the promise is practically canceled.
function promisify(fun){
return function(data){
var result = {id:null, promise:null}; // template return object
result.promise = new Promise((resolve,reject) => result.id = fun(data, (err,res) => err ? reject(err) : resolve(res)));
return result;
};
}
function doubleAsync(data,cb){
var dur = ~~(Math.random()*5000); // return the double of the data within 0-5 seconds.
console.log("Resolve in " + dur + " msecs");
return setTimeout(_ => cb(false, data*2),dur);
}
var doubleWithPromise = promisify(doubleAsync),
promiseDataSet = doubleWithPromise(100);
setTimeout(_ => clearTimeout(promiseDataSet.id),2500); // give 2.5 seconds to the promise to resolve or cancel it.
promiseDataSet.promise
.then(v => console.log("The asynchronously obtained result is: " + v));
You can use fetch(), Response.body.getReader(), where when .read() is called returns a ReadableStream having a cancel method, which returns a Promise upon cancelling read of the stream.
// 58977 bytes of text, 59175 total bytes
var url = "https://gist.githubusercontent.com/anonymous/"
+ "2250b78a2ddc80a4de817bbf414b1704/raw/"
+ "4dc10dacc26045f5c48f6d74440213584202f2d2/lorem.txt";
var n = 10000;
var clicked = false;
var button = document.querySelector("button");
button.addEventListener("click", () => {clicked = true});
fetch(url)
.then(response => response.body.getReader())
.then(reader => {
var len = 0;
reader.read().then(function processData(result) {
if (result.done) {
// do stuff when `reader` is `closed`
return reader.closed.then(function() {
return "stream complete"
});
};
if (!clicked) {
len += result.value.byteLength;
}
// cancel stream if `button` clicked or
// to bytes processed is greater than 10000
if (clicked || len > n) {
return reader.cancel().then(function() {
return "read aborted at " + len + " bytes"
})
}
console.log("len:", len, "result value:", result.value);
return reader.read().then(processData)
})
.then(function(msg) {
alert(msg)
})
.catch(function(err) {
console.log("err", err)
})
});
<button>click to abort stream</button>
The method I am currently using is as follows:
var optionalReturnsObject = {};
functionThatReturnsPromise(dataToSend, optionalReturnsObject ).then(doStuffOnAsyncComplete);
console.log("Some instant data has been returned here:", optionalReturnsObject );
For me, the advantage of this is that another member of my team can use this in a simple way:
functionThatReturnsPromise(data).then(...);
And not need to worry about the returns object. An advanced user can see from the definitions what is going on.
Is there a possibility to create atomic database transactions with bookshelf? I'm having a problem with duplicates in the database. The problematic code is as below:
bookshelf.transaction(function (t) {
var modelLocation = new Models.Location({'name':event.venue});
modelLocation.fetch({transacting:t})
.then(function (fetchedLocation) {
if (!fetchedLocation) {
modelLocation.save(null,{transacting:t}).then(function (savedModel) {
t.commit(savedModel)
}).catch(function (err) {
t.rollback(err)
});
}
else{
t.commit(fetchedLocation)
}
})
})
I call the method containing this code almost simultaniously and asynchronously 20 times. From these 20, there are 5 duplicate datasets. This results in around 2-3 duplicates in the database. The current workaround is to wrap the whole thing in a setTimeout with a random timout between 0 and 10 seconds which almost never gives me duplicates. But this is obviously not a production ready solution.
OK so in the end, I decided to go with the async.js library and it's queue.
The queue guarantees that maximum n async tasks are executed concurrently. In this case 1.
I made a module which exports a queue instance. This way I can use it across multiple modules. It simply waits for the promise to fulfill.
var async = require('async');
module.exports = async.queue(function (task, callback) {
task().then(function () {
callback();
});
},1);
Then in the module, where I need an "atomic" transaction I have the following code:
var queue = require('./transactionQueue');
...
...
queue.push(function(){
return bookshelf.transaction(function (t) {
var modelLocation = new Models.Location({'name':event.venue});
return modelLocation
.fetch({transacting:t})
.then(function (fetchedLocation) {
if (!fetchedLocation) {
return modelLocation
.save(null,{transacting:t});
}
});
});
});
It's important to wrap the transaction into a function so it won't get executed right away.
Since Bookshelf transactions are promises you do not need to explicitly call commit() or rollback(). Just let the fulfilled promise to commit itself, or you can force a rollback by throwing an exception.
In your code there was apparently a small bug that could be causing the trouble: an argument missing from the fetch()'s then() -- this argument is the result from the fetch() invocation, an instance if the object was found or null if not.
bookshelf.transaction(function (t) {
var modelLocation = new Models.Location({'name':event.venue});
return modelLocation
.fetch()
.then(function (fetchedLocation) {
if (!fetchedLocation) {
modelLocation
.save(null,{transacting:t});
}
})l
});
I am not able to test that now, but I hope it helps.