Interacting with element throw Stale Element Reference - webdriver

I have this site: http://embed.plnkr.co/Bs5iDqtXSSnvye2ORI6k/preview
Code:
var app = angular.module('plunker', []);
var a = new Array(1000);
for (var i = 0; i< 1000; i++) {
a[i] = 'Name' + i;
}
app.controller('MainCtrl', function($scope, $interval) {
$scope.names = a;
$scope.start = function () {
$interval(function () {
$scope.names.pop();
}, 50);
}
});
And the following spec:
'use strict';
describe('Name list', function () {
it('should get the text of the last name', function () {
browser.driver.get('http://embed.plnkr.co/Bs5iDqtXSSnvye2ORI6k/preview');
browser.switchTo().frame(browser.driver.findElement(protractor.By.tagName('iframe')));
element(by.buttonText('start')).click();
expect(element.all(by.tagName('span)).last().getText()).toBe('Name999');
});
});
And this config:
'use strict';
// An example configuration file.
exports.config = {
baseUrl: 'http://localhost:3000',
seleniumAddress: 'http://localhost:4444/wd/hub',
specs: ['stale.spec.js']
};
And when I run Protractor I get the following error:
StaleElementReferenceError: stale element reference: element is not
attached to the page document (Session info: chrome=43.0.2357.81)
(Driver info: chromedriver=2.15.322455
(ae8db840dac8d0c453355d3d922c91adfb61df8f),platform=Mac OS X 10.10.3
x86_64) (WARNING: The server did not provide any stacktrace
information) Command duration or timeout: 9 milliseconds For
documentation on this error, please visit:
http://seleniumhq.org/exceptions/stale_element_reference.html Build
info: version: '2.45.0', revision: '5017cb8', time: '2015-02-26
23:59:50' System info: host: 'ITs-MacBook-Pro.local', ip:
'129.192.20.150', os.name: 'Mac OS X', os.arch: 'x86_64', os.version:
'10.10.3', java.version: '1.8.0_31' Driver info:
org.openqa.selenium.chrome.ChromeDriver Capabilities
[{applicationCacheEnabled=false, rotatable=false,
mobileEmulationEnabled=false,
chrome={userDataDir=/var/folders/rr/63848xd90yscgwpkfn8srbyh0000gq/T/.org.chromium.Chromium.rarNyX},
takesHeapSnapshot=true, databaseEnabled=false, handlesAlerts=true,
version=43.0.2357.81, platform=MAC, browserConnectionEnabled=false,
nativeEvents=true, acceptSslCerts=true, locationContextEnabled=true,
webStorageEnabled=true, browserName=chrome, takesScreenshot=true,
javascriptEnabled=true, cssSelectorsEnabled=true}] Session ID:
235ec977a69d98c7f5b75a329e8111b2
This means that the element I try to interact with (getting the text of the element), isn't attached to the DOM anymore. This example is really my spec simplyfied. What really happens in my real spec is I try to get the text of the last element of a list of elements (generated by ng-repeat). What also happens is that the model updates, by removing the last element of the array representing the list of element. This example above is just something to reproduce the error (every time).
If I comment out this line: element(by.buttonText('start')).click(); the spec is successful.

I struggled a lot with this and tried to figure out why this would happen. I first thought that the element finder which points to the last element of the list was created long before the interaction was done, so there was no surprise to me that the element could be detached from the DOM in that period of time between the creation of the element finder and the interaction.
What I've later found out is that the element is found just before the interaction is done, every time you interact with something. So pointing to the last element should actually point to the last element of the time interacting with the element.
Using browser.pause() I was able to see what WebDriver really does, and this is two tasks where in between the error is thrown:
(pending) Task::414<WebDriver.findElements(By.tagName("span"))>
| | | | | | Task: WebDriver.findElements(By.tagName("span"))
| | | | | | at Array.forEach (native)
Here in between, the DOM is updated according to the model, and the last element of the list is detached.
(pending) Task::1170<WebElement.getText()>
| | | | | | Task: WebElement.getText()
| | | | | | at Array.forEach (native)
The DOM is updated in this small hole of execution. Currently the model updates every 50 ms, and this is sure to throw a Stale Element Reference error. But if I increase the interval to, say 1000 ms, then the chance to get the error is much less. So it depends on how fast your computer runs if you get this error.
The fix is up to you, but with this information it should be a bit clearer what to do, I hope.

The browser is running asynchronously from your protractor tests. This example really highlights that nicely (its a problem for many protractor tests, but not usually so obvious). This is compounded by what looks like a single line:
expect(element.all(by.tagName('span')).last().getText()).toBe('Name999');
actually requires several round-trips to the browser (there are a lot of promises being returned and resolved: element.all, last, getText). For most web pages that are "passive" once they've stabilized, this isn't a problem, but for any web page that dynamically changes without user inputs, testing with protractor can be painful.
To make a lookup and test "atomic" in the browser, thus side-steping this issue, you can define a function to run in the browser (my CSS DOM-fu is weak, so hopefully someone can tweak this to make it less terrible):
expect(browser.executeScript(function() {
var spans = document.getElementsByTagName('span');
var lastSpan = spans[spans.length - 1]; // XXX handle empty spans
return lastSpan.innerText;
}).toBe('Name999');
Beware that the "function" is serialized and executed on the browser, so no variables from an enclosing scope will work. Also, this approach loses any browser-compatibility magic hiding in protractor or webdriver (e.g., I wouldn't be surprised if getText() wasn't simply a innerText accessor).
Also, note that you still have a race condition. Between the 'click' to get things started and this code to actually inspect the DOM, it may or may not have been mutated (and "Name999" might be gone).

Related

Cypress - How can I run test files in order

When I press the "run all specs" button or use the run command that runs all files in Cypress it runs all test files alphabetically, so I don't want that.
I want to sort all of them with my own rules.
Let's say I have 3 steps in a chat app test.
Can connect the chat app
Can connect the chat
Can the user send a message
I want to test every step without being tied to each other.
What I mean, Testing one of their own function.
What I do is as follows
chat_app_connect.spec.js
describe('Server Connecting Test', () => {
it('Visit Server page', () => {
cy.visit('https://chat.page..');
});
it('Check welcome messages', () => {
cy.contains('Live Support');
cy.contains('Hello, Stranger');
});
it('Check URL and status of circle', () => {
// URL
cy.url()
.should('include', '/hello');
// Status Circle
cy.get('circle')
.should('have.class', 'positive');
});
});
chat_connect.spec.js
import './chat_app_connect.spec.js';
describe('Chat Connecting Test', () => {
it('Type customer name', () => {
cy.get('input')
.clear()
.type('E2E Test');
});
it('Click to the submit button', () => {
cy.get('.submit-button')
.click();
});
it('Check URL and status of circle', () => {
// URL
cy.url()
.should('equal', 'https://client.dev.octopus.chat/');
// Status Circle
cy.get('circle', { timeout: 5000 })
.should('have.class', 'positive');
});
});
chatting.spec.js
import './chat_connect.spec.js';
describe('Chatting Tests', () => {
it('Type a test message then press Enter and check the message if it sent', () => {
// Type
cy.get('#chat-message')
.clear()
.type('Hey I\'m a test message{enter}');
// Check the message
cy.get('.message-list')
.should('contain', 'Hey I\'m a test message');
});
});
as you see every test is tied to each other, and that is mean when I tried to test just catting functionality its call every test and the whole tests will be tested.
I don't know if it is the right way or not.
what should I do in this case or can it be an acceptable way
I have a particular case where I launch multiple instances of an app, rather than using fixtures or test data, I simply integrate user feedback as Cypress tests from login on forwards.
In any case, I used the specPattern config in cypress.json to set the spec file run order:
{
"baseUrl": "http://localhost:5000",
"specPattern": [
"login/*.js",
"leads/new-lead.spec.js",
"leads/leads-list.spec.js",
"leads/lead-detail.spec.js",
"leads/lead-modify.spec.js",
//...
]
}
No file numbering needed :D
The easiest solution is most likely to add a prefix to all your test files, such as:
01-chat_app_connect.spec.js
02-chat_connect.spec.js
etc.
Cypress is going to take those files in alphabetical order, which you can "trick" into your wanted behavior by using a number as a prefix.
Jean Lescure's answer was a lifesaver. We needed to run tests based on priority without having a bunch of duplicated tests or symlinks. The following worked for us in our default cypress config file:
"integrationFolder":"cypress/integration",
"testFiles": [
"high_priority_specs/**/*.js",
"medium_priority_specs/**/*.js",
"low_priority_specs/**/*.js"
]
To change the level of priority we used 3 configs files that were loaded using the cypress --configFile argument. To run the higher priority tests (smoke tests only) we used the following:
"integrationFolder":"cypress/integration",
"testFiles": [
"high_priority_specs/**/*.js"
]
Cypress does not intentionally let you do this, and for good reasons:
It's generally indicative of poor test design. Tests should not depend on the state of one another. Any test should be able to be run successfully in isolation from the rest of the test suite.
You'll never be able to take advantage of cypress' built in ability to run tests in parallel since you can't guarantee one spec will be ran after another
Here is a relevant discussion about this that gets into more detail: https://github.com/cypress-io/cypress/issues/390
However, if you decide to do this anyway, you can do it by prefixing the name of the specs with a number:
01-some-spec.js
02-alphabetically-first-spec.js
03-some-other-spec.js
In addition to #Brendan answer, if you have a nested folder structure, this approach will work as well.
01-folder-name
|
- 01-some-spec.js

Using Flow union types for Redux actions

Following the style of this Facebook app sample using Redux and Flow together, I made an action type in this manner:
type Action =
| { type: 'ADD_FILES', files: Array<{ id: number, file: File }> }
| { type: 'HANDLE_IMAGE_PUBLISHED', id: number, name: string }
| { type: 'SET_IMAGE_UPLOAD_PROGRESS', id: number, progress: number }
;
But I've found that when I try to process my actions with a reducer, Flow complains if I try to access the name or progress properties, saying "Property not found in object type".
That is, in my reducer, if I check that action.type === 'HANDLE_IMAGE_PUBLISHED' and then access action.name, Flow complains. And the same thing goes for for accessing action.progress property when action.type === 'SET_IMAGE_UPLOAD_PROGRESS'. Both these property accesses should be legit under their respective circumstances, as far as I can tell, but Flow complains.
Yet for some reason it's OK for me to access action.id anywhere, even though one of the types in my union doesn't specify an id property. I'm very confused.
Here is a live demo in the Flow REPL. What am I doing wrong?
This is simply a case of a type refinement invalidation:
https://flow.org/en/docs/lang/refinements/#toc-refinement-invalidations
Because you are using the value in a callback, Flow pessimistically assumes that you could have re-assigned action before the callback runs (it does not know that the map callback is called immediately). It also does not do the analysis to see that there is no place, in fact, that you re-assign it.
All that's needed is to pull the action out as a const:
export default (state: Array<ImageRecordModel> = [], action_: Action): Array<ImageRecordModel> => {
const action = action_;
(tryflow link)
You may also want to consider enabling const params in your .flowconfig. This does basically what you expect: treats all params as const:
[options]
experimental.const_params=true

Meteor : "console" variable undefined inside require call

I'm facing a strange problem on Meteor and I can't resolve it :
I'm developping a WebRTC app using Meteor, PeerJS and AdapterJS (which give an WebRTC plugin for unsupported browser like Safari or IE). These two libs are downloaded using NPM : meteor npm install peerjs/adapterjs
So in my view's controller I have :
view.js
//import Peer from 'peerjs'; => same error with "import"
//import AdapterJS from 'adapterjs';
Template.view.onRendered(function(){
AdapterJS = require("adapterjs");
Peer = require("peerjs");
//var peerkey="..."
var peer = new Peer({
key: peerkey, // get a free key at http://peerjs.com/peerserver
debug: 3,
config: {'iceServers': [
{ url: 'stun:stun.l.google.com:19302' },
{ url: 'stun:stun1.l.google.com:19302' },
]}
});
But when I run my controller, I get an exception because "console" is undefined inside peerjs/util.js function when calling the peerjs constructor :
Uncaught TypeError: Cannot read property 'log' of undefined
Strangly, when I only require "peerjs", there is no exeption...
I tried to change the order of require functions but it won't work.
Other variable like "alert", "window.console" work and are defined inside the module but "console" not.. :/
Any suggestion can help ^^
Thanks in advance.
EDIT : If I add a breakpoint on the first line of node_module/peerjs/lib/util.js, I see that the "console" variable is "undefined" inside util.js but .... it is defined inside the caller function (fileEvaluate) !
EDIT2 : I tried something else to check if the code inside adapterjs redefine or change something : I put 'require("adapterjs")' inside a timeout function with a long delay (10 seconds) and .... console is still undefined inside peer module ! But when I comment require("adapterjs"), no error, console is defined ! I think that Meteor do something special before running the controller script depending on require functions...
EDIT3 : Here is a git repo to test the project : gitlab.com
If you display the dev console, you will see exceptions.
I found a solution, although I don't fully understand it myself. It's something to do with the way that Meteor imports modules, and Peerjs runs afoul of that.
Basically I copied node_modules/peerjs/dist/peer.js into the client directory, so that Meteor will load it "as is".
A small change to main.js as follows:
import './main.html';
// I placed peer.js from the node_modules/peerjs/dist/peer.js into the client folder and it works fine
// import {Peer} from 'peerjs';
import {AdapterJS as Adapter} from 'adapterjs';
Template.hello.onCreated(function helloOnCreated() {
// counter starts at 0
window.peer = new Peer({
and it works fine :)
I see in line 860+ of adapter.js that console is being defined (part of the shim) from https://github.com/Temasys/AdapterJS/blob/master/source/adapter.js
// IE 9 is not offering an implementation of console.log until you open a console
if (typeof console !== 'object' || typeof console.log !== 'function') {
/* jshint -W020 */
console = {} || console;
// Implemented based on console specs from MDN
// You may override these functions
console.log = function (arg) {};
console.info = function (arg) {};
console.error = function (arg) {};
This code defines the console if it doesn't find one to it's liking? Does this mean you are using IE9 or some other incompatible browser?
Try pasting this into your console and see what it tells you:
if (typeof console !== 'object' || typeof console.log !== 'function') alert("Console not present, needs to be shimmed?"); else console.log("console is ok");
Presumably the reason you are using adapter.js is for compatibility purposes - this will help you trouble shoot it. Please let me know what you find, as I will be following you down this path :)

How to delete a large node in firebase

I have a Firebase child node with about 15,000,000 child objects with a total size of about 8 GB of data.
exampele data structure:
firebase.com/childNode/$pushKey
each $pushKey contains a small flat dictionary:
{a: 1.0, b: 2.0, c: 3.0}
I would like to delete this data as efficiently and easy as possible. How?
What i Tried:
My first try was a put request:
PUT firebase.com/childNode.json?auth=FIRE_SECRET
data-raw: null
response: {
"error": "Data requested exceeds the maximum size that can be accessed with a single request. Contact support#firebase.com for help."
}
So that didn't work, let's do a limit request:
PUT firebase.com/childNode.json?auth=FIRE_SECRET&orderBy="$key"&limitToFirst=100
data-raw: null
response: {
"error": "Querying related parameters not supported on this request type"
}
No luck so far :( What about writing a script that will get the first X number of keys and then create a patch request with each value set to null?
GET firebase.com/childNode.json?auth=FIRE_SECRET&shallow=true&orderBy="$key"&limitToLast=100
{
"error" : "Mixing 'shallow' and querying parameters is not supported"
}
It's really not going to be easy this one? I could remove the shallow requirement and get the keys, and finish the script. I was just hoping there would be a easier/more efficient way???
Another thing i tried were to create a node script that listen for childAdded and then directly tries to remove those children?
ref.authWithCustomToken(AUTH_TOKEN, function(error, authData) {
if (error) {console.log("Login Failed!", error)}
if (!error) {console.log("Login Succeeded!", authData)}
ref.child("childNode").on("child_added", function(snap) {
console.log(`found: ${snap.key()}`)
ref.child("childNode").child(snap.key()).remove( function(err) {
if (!err) {console.log(`deleted: ${snap.key()}`)}
})
})
})
This script actually hangs right now, but earlier I did receive somethings like a max stack limit warning from firebase. I know this is not a firebase problem, but I don't see any particular easy way to solve that problem.
Downloading a shallow tree, will download only the keys. So instead of asking the server to order and limit, you can download all keys.
Then you can order and limit it client-side, and send delete requests to Firebase in batches.
You can use this script for inspiration: https://gist.github.com/wilhuff/b78e7391396e09f6c614
Use firebase cli tool for this: firebase database:remove --project .
In Browser Console this is fastest way
database.ref('data').limitToFirst(10000).once('value', snap => {
var updates = {};
snap.forEach(snap => {
updates[snap.key] = null;
});
database.ref('data').update(updates);
});

Marionette js itemview not defined: then on browser refresh it is defined and all works well - race condition?

Yeah it's just the initial browser load or two after a cache clear. Subsequent refreshes clear the problem up.
I'm thinking the item views just aren't fully constructed in time to be used in the collection views on the first load. But then they are on a refresh? Don't know.
There must be something about the code sequence or loading or the load time itself. Not sure. I'm loading via require.js.
Have two collections - users and messages. Each renders in its own collection view. Each works, just not the first time or two the browser loads.
The first time you load after clearing browser cache the console reports, for instance:
"Uncaught ReferenceError: MessageItemView is not defined"
A simple browser refresh clears it up. Same goes for the user collection. It's collection view says it doesn't know anything about its item view. But a simple browser refresh and all is well.
My views (item and collection) are in separate files. Is that the problem? For instance, here is my message collection view in its own file:
messagelistview.js
var MessageListView = Marionette.CollectionView.extend({
itemView: MessageItemView,
el: $("#messages")
});
And the message item view is in a separate file:
messageview.js
var MessageItemView = Marionette.ItemView.extend({
tagName: "div",
template: Handlebars.compile(
'<div>{{fromUserName}}:</div>' +
'<div>{{message}}</div>' +
)
});
Then in my main module file, which references each of those files, the collection view is constructed and displayed:
main.js
//Define a model
MessageModel = Backbone.Model.extend();
//Make an instance of MessageItemView - code in separate file, messagelistview.js
MessageView = new MessageItemView();
//Define a message collection
var MessageCollection = Backbone.Collection.extend({
model: MessageModel
});
//Make an instance of MessageCollection
var collMessages = new MessageCollection();
//Make an instance of a MessageListView - code in separate file, messagelistview.js
var messageListView = new MessageListView({
collection: collMessages
});
App.messageListRegion.show(messageListView);
Do I just have things sequenced wrong? I'm thinking it's some kind of race condition only because over 3G to an iPad the item views are always undefined. They never seem to get constructed in time. PC on a hard wired connection does see success after a browser refresh or two. It's either the load times or the difference in browsers maybe? Chrome IE and Firefox on a PC all seem to exhibit the success on refresh behavior. Safari on iPad fails always.
PER COMMENT BELOW, HERE IS MY REQIRE BLOCK:
in file application.js
require.config({
paths: {
jquery: '../../jquery-1.10.1.min',
'jqueryui': '../../jquery-ui-1.10.3.min',
'jqueryuilayout': '../../jquery.layout.min-1.30.79',
underscore: '../../underscore',
backbone: '../../backbone',
marionette: '../../backbone.marionette',
handlebars: '../../handlebars',
"signalr": "../../jquery.signalR-1.1.3",
"signalr.hubs": "/xyvidpro/signalr/hubs?",
"debug": '../../debug',
"themeswitchertool": '../../themeswitchertool'
},
shim: {
'jqueryui': {
deps: ['jquery']
},
'jqueryuilayout': {
deps: ['jquery', 'jqueryui']
},
underscore: {
exports: '_'
},
backbone: {
deps: ["underscore", "jquery"],
exports: "Backbone"
},
marionette: {
deps: ["backbone"],
exports: "Marionette"
},
"signalr": {
deps: ["jquery"],
exports: "SignalR"
},
"signalr.hubs": {
deps: ["signalr"],
exports: "SignalRHubs"
},
"debug": {
deps: ["jquery"]
},
"themeswitchertool": {
deps: ["jquery"]
}
}
});
require(["marionette", "jqueryui", "jqueryuilayout", "handlebars", "signalr.hubs", "debug", "themeswitchertool"], function (Marionette) {
window.App = new Marionette.Application();
//...more code
})
Finally, inside the module that uses creates the collection views in question, the list of external file dependencies is as follows:
var dependencies = [
"modules/chat/views/userview",
"modules/chat/views/userlistview",
"modules/chat/views/messageview",
"modules/chat/views/messagelistview"
];
Clearly the itemViews are listed before collectionViews. This seems correct to me. Not sure what accounts for the collectionViews needing itemViews before they are defined. And why is all ok after a browser refresh?
The sequence in which you load files is most likely wrong: you need to load the item view before the collection view.
Try putting all of your code in the same file in the proper order, and see if it works.
The free preview to my book on Marionette can also guide you to displaying a collection view.
Edit based on calirification:
The dependencies listed for the module are NOT loaded linearly. That is precisely what RequireJS was designed to avoid. Instead the way to get the files loaded properly (i.e. in the correct order), is by defining a "chain" of dependencies that RequireJS will compute and load.
What you need to do is define (e.g.) your userlistview to depend on userview. In this way, they will get loaded in the proper order by RequireJS. You can see an example of a RequireJS app here (from by book on RequireJS and Marionette). Take a look at how each module definition decalre which modules it depends on (and that RequireJS therefore needs to load before). Once again, listing the modules sequentially within a dependecy array does NOT make them get loaded in that sequence, you really need to use the dependency chain mechanism.

Resources