I have a so called chat view that basically consists of a repeat.for that loops over all the messages and renders the message views using <compose>.
The problem is that it becomes quite slow once the message count exceeds 50 and the user navigates between chats (which triggers the repeat.for update, as I replace the dataset in the VM).
I feel like I am doing something wrong in regards of handling a view like this. Can I get some input regarding other alternatives?
I have tried UI virtualization, but unfortunately the current plugin does not offer support for features that I require (variable height items, bottom-up alignment).
I have also done quite a bit of optimizations regarding bindings, most are one-time and updates to the dataset are debounced. But this did not improve things, as the main bottleneck is the initial load (binding the views the first time).
Thanks!
Example of current approach:
<li repeat.for="message of chat.messages">
<compose view-model.bind="getMessageViewFromMessage(message) & oneTime"
model.bind="message & oneTime"
containerless>
</compose>
</li>
I think you need to consider not using <compose> at all for this. Is there a reason you need <compose>? When you think about it, the <compose element has to re-run through the same view instantiation/binding logic Aurelia does for everything else, every time a message is displayed.
I would personally create a HTML partial with some bindable properties and inside of the loop, reference it. So you might have chat-message.html and then display it like this:
<li repeat.for="message of messages">
<chat-message message.bind="message"></chat-message>
</li>
Where possible and in most cases it should be, avoid dynamic composition for potentially large sets of repeated items.
You should check out the aurelia-ui-virtualization library. Once you load it, you can replace repeat.for with virtual-repeat.for in places like this and you'll get a virtualized repeater that will help improve perf with this type of situation.
Related
Example of the HTML of a dropdown element:
<div aria-owns="ember-basic-dropdown-content-ember1234" tabindex="0" data-ebd-id="ember1234-trigger" role="button" id="ember1235" class="ember-power-select-trigger ember-basic-dropdown-trigger ember-view"> <!---->
<span class="ember-power-select-status-icon"></span>
</div>
The xpath and CSS selector also contain the same ember id.
xpath : //*[#id="ember1235"]
css selector : #ember1235
The ember id would change from id="ember1235" to say, id="ember1265" when there is a change in the UI.
I am using id to locate the element. But every time it changes I need to modify the code. Is there any other attribute I could use for Ember JS UI elements?
There is quite a lot to discuss in your question but hopefully we will have a good answer for you #PriyaK
The first thing to mention is that Ember IDs may not be the best method to select an element in the DOM. As you have already mentioned, they can change from time to time and also it doesn't really give you a great semantic thing to select in your selenium test so it might seem a bit out of context when looking back.
One thing that you could try is to either pass a class to the ember-power-select component (the one that provides the HTML that you used in your example) and use that to select the element, something like:
<PowerSelect
#class="my-fancy-class"
as |name|
>
{{name}}
</PowerSelect>
Then you should be able to select the selected value by using the CSS selector .my-fancy-class span (because the component outputs the selected value in a span)
We just tried this in an example app but it didn't actually work 🤔 Never fear, you can also do something like this and it should work with the same selector as before:
<div class="my-fancy-class">
<PowerSelect as |name|>
{{name}}
</PowerSelect>
</div>
This is fine, but there are also a few issues using classes for selectors in tests. One example of a problem that might crop up is that your tests might all suddenly stop working if you did a style refactor and changed or removed some of the classes on your elements. One technique that has become popular in the Ember community is to use data-test- attributes on your DOM nodes like this:
<div data-test-my-fancy-select>
<PowerSelect
#class="my-fancy-class"
as |name|
>
{{name}}
</PowerSelect>
</div>
which can then be accessed by the following selector: [data-test-my-fancy-select] span. This is great for a few reasons! Firstly it separates the implementation of your application and tests from your styling and avoids the issue I described above. The second benefit of this method is that using what #Gokul suggested in the comments, the ember-test-selectors package, you can make use of these data-test- selectors in your development and test environments but they will be automatically removed from your production build. This is great to keep your DOM clean in production but also, depending on the size of your application, could save you a reasonable amount of size in your templates on aggregate.
I know you say that you are using selenium for your testing but it's also worth mentioning that if you're using the built-in Ember testing system you will be able to make use of some testing helpers that addons may provide you. ember-power-select is one of those addons that provides specific testing helpers and you can read more about it in their documentation: https://ember-power-select.com/docs/test-helpers
I hope this answers any questions you had!
This question was answered as part of "May I Ask a Question" Season 3 Episode 1. If you would like to see us discuss this answer in full you can check out the video here: https://www.youtube.com/watch?v=1DAJXUucnQU
Here is my problem:
I typically have a paginated datagrid with a lot of rows. I want to be able to edit the data for each row of course but I have 2 constraints on this:
I need the edition form to replace the content of the page (I don't want a popup, modal dialog or side panel)
I don't want to lose the state of the datagrid: maybe I navigated down 5 pages in the datagrid and I don't want to be reloaded on the first page. And actually, I'd rather not reload the data I already had (the edited data will be updated automatically by my persistence layer anyway).
Ideally, I would have liked to have some kind of subrouter but I'm not sure how it would fit the first requirement. Otherwise, I could have a component that would be hidden by default and positions itself on top of the datagrid view when necessary but that feels quite hacky and forces me to have everything in the same template. And I will have to handle a stack of these components if I have several different 'full-screen panel'...
Any idea on a correct way to implement this?
Thanks!
I tried different solutions to no avail unfortunately. I had a long discussion with #Kukks on gitter and we agreed that using subrouters and viewports might be a bit overkill for my use case.
I reverted to my original idea of using absolutely positioned components to hide the previous one in a kind of "deck layout". This is not ideal as I would have liked completely separated views and using components forces me to declare them in the main view but it works well and is very easy to implement...
So: not as clean as I would have liked but much easier to implement and less convoluted.
Consider using Router View Ports
http://aurelia.io/hub.html#/doc/article/aurelia/router/latest/router-configuration/9
I am an Automation Engineer, I am currently trying to create test cases for a webpage that are sustainable ( that I can run at a later time and have still pass)
Here is my problem:
I am trying to select multiple web buttons that have the same exact class name. Now I can 'select' these buttons but these are only temporary x paths that are subject to change.
I need UNIQUE ID's (or some way of distinguishing them) for the same web elements. The only difference in the x paths are:
HTML format code that I can find each button, however if one button is moved my tests will fail.
HTML code that is the class name + nth of the button. But again my tests will fail if a button is taken out of the webpage.
//*[#id="tenant-details-accordion"]/div[1]/div[2]/div/div[2]/div[1]/div/a/div
//*[#id="tenant-details-accordion"]/div[1]/div[2]/div/div[2]/div[2]/div/a/div
//*[#id="tenant-details-accordion"]/div[1]/div[2]/div/div[2]/div[3]/div/a/div
^^The above code is how I currently find each button with Selenium
If I copy each classes x path this is what I get
<div class="src-js-components-DateControl-___DateControl__dateControl___2nYAL"><a tabindex="0"><div class="src-js-components-DateControl-___DateControl__icon___2z6Ak null"></div><!-- react-text: 392 -->Set +<!-- /react-text --></a></div>
<div class="src-js-components-DateControl-___DateControl__dateControl___2nYAL"><a tabindex="0"><div class="src-js-components-DateControl-___DateControl__icon___2z6Ak null"></div><!-- react-text: 386 -->Set +<!-- /react-text --></a></div>
<div class="src-js-components-DateControl-___DateControl__dateControl___2nYAL"><a tabindex="0"><div class="src-js-components-DateControl-___DateControl__icon___2z6Ak null"></div><!-- react-text: 398 -->Set +<!-- /react-text --></a></div>
I have talked to the Development team about this issue however they tell me that assigning Unique ID's to these web elements is a big no-no. (something to do with globalization of the project when we go to production) Even if they did assign Unique Id's they tell me that they would have to strip them before the project can be sent to production. Which ultimately would render my tests useless in the end...
I now come to Stack Overflow for help, everything I look up online cannot give me an answer, however this does seem to be a valid question between QA and Development departments.
Is there a way to assign Id's to a web element so that a tester using selenium can identify them and yet it will not affect a Developers ability to use that element through an entire project?
Edit:
With the framework that my company uses, I am not actually writing any code in selenium. I save the Xpath to an object and then later in a manual test call that object with a pre-existing method. I can ONLY select Xpaths to be saved in an object.
I'll talk to Dev later to have them explain the big 'no-no' so that I may be able to communicate that with all of you..Thank you for your time and input
For example:
BirthDateSet= someXpath
Click() using {BirthDateSet}
Here are some pictures to help give you a visual
You can simplify your XPathto make it less sensitive to possible changes in DOM:
//div[#class="src-js-components-DateControl-___DateControl__icon___2z6Ak null"][N] # Where N is button index
If this part 2z6Ak of class name is dinamically generated, try:
//div[starts-with(#class, "src-js-components-DateControl-___DateControl__icon___")][N] # Where N is button index
Another way of reaching to the required button using xpath is following:
//*[#id='tenant-details-accordion']/descendant::div[contains(#class, 'DateControl__icon')][1]
Above xpath should be able to select the first occurring button. Similarly, for the second occurring button, the xpath will become:
//*[#id='tenant-details-accordion']/descendant::div[contains(#class, 'DateControl__icon')][2]
Hence, you can continue to replace the predicate([2]) based on occurrence level of the required button on page.
you can ask devs to add attribute to those buttons
so instead of:
<button class="common_class_name">
<button class="common_class_name">
<button class="common_class_name">
for each button you will see
<button class="common_class_name" test-id="uniqueAttributeForButton1">
<button class="common_class_name" test-id="uniqueAttributeForButton2">
<button class="common_class_name" test-id="uniqueAttributeForButton3">
etc
And in your tests you can find those buttons using css selector
e.g. C# code:
var buttonLocator = "[test-id=\"uniqueAttributeForButton1\"]";
WebElement button = driver.FindElement(By.CssSelector(buttonLocator));
if devs doesnt want to add unique IDs or attributes (it's weird why they don't want to do that) you could try something like this:
//return div class 'col-sm-4' which contain text 'Activation Date'
var parentDiv = driver.FindElement(By.XPath("//div[contains(#class, 'col-sm-4') and contains(., 'Activation Date')]"));
//should return div with icon you want e.g. Click
var childElement = parentDiv.FindElement(By.CssSelector(".DateControl__icon"));
childElement.Click();
You can go with the nth approach to find your unique element or with another, that won't solve your real problem which is that your colleague developers aren't cooperating with you. The automated tests are supposed to be (among other things) a service for the developers to know they haven't broken anything as fast as possible, but if they aren't giving any effort, then WHY should you invest in it at all..?
They probably won't even look at the results. Especially when your locators are in the form of div[2]/div[3]... which would break very often and you'll lose their trust in the automation.
I'm no front-end expert but I've seen enterprise products in production with G10N and L10N and other long words, that have id's in their HTML, so if I were you I'd dig deeper to understand why? that big no-no, which sounds very odd to me...
So you asked the wrong question, it should be something like ~ How to add unique id's to the HTML using React in a global web application? and the ones who should be asking it are the developers.
As a side note, as many people tend to think, automation is not mainly about moving manual tests to automated it also finds "bugs" in the communication inside the company. To quote a great blog post on this subject:
Counter-intuitively, the bigger obstacles when developing automation
are either low priority bugs or even things that are not bugs
whatsoever. These obstacles can be symptoms of a bad design, bad
development practices and even symptoms of inefficient communication
patterns in the organization (that are usually caused by the
organizational structure! Conway’s law is a classic example for this).
Good luck!
EDIT:
Thanks for the code.. So if you look at your code, you can see that although what you're after have similar attributes, there are elements in the page that would allow you to identify what you're after. So let's say you want the SET link after Birth Date, then use:
//div[contains(text(),'Birth Date')]/div/a/div
So to reuse this xpath, just replace the text String.format("//div[contains(text(),'%s')]/div/a/div", labelName);
OLD: if you want to temporarily assign the elements an attribute, I believe you can do that through javascript element.setAttribute() which you can run through the executeScript() function.
Don't the buttons have text? Are you saying that the buttons swap places with one another? or don't you want to use the index of the buttons?
I understand you want to select the "Set +" buttons, correct?
If so you can try to find all buttons under the #id="tenant-details-accordion" element with a specific text like that:
//*[#id="tenant-details-accordion"]//*[text()='Set +']
I have two scenarios I need help with, and I thought posting them together would prove more
valuable for myself, and other viewers.
Setup:
Worklight 6.1
dojo 1.9
Application:
MainView.html (Contains Body, and a transition Div, and NorthSouthView.js script reference)
View1.html (Contains a single Div that displays and unordered list
View2.html (Contains a single Div that Displays <p> data, and also plays audio)
View3.html (Contains a single Div that Displays instructional information)
application-descriptor <mainFile> MainView.html </mainFile>
All of the views are stored together in the application. There are no external http queries made by the application. All view transitions are local to the application.
Scenario #1:
At application start the MainView.html is invoked by worklight. Anticipated format::
<body>
<div>
<h1 id="SSheader" data-dojo-type="dojox.mobile.Heading" data-dojo-props='fixed:"top"'>Loan Snapshot</h1>
</div>
<div id="TransitionViewDiv">
/* Would like to load content of View1.html, View2.html, or View3.html here */
</div>
<script>SetView.js</script>
</body>
Description + Questions:
When the application starts, SetView.js will be loaded, and the purpose of this script is to look at localStorage and determine which view should be displayed. (View1, View2, or View3). The goal is to keep SSheader fixed at the top of the screen at all times. Only TransitionViewDiv would update.
Questions:
1) What coding approach can be used in SetView.js to load one of the 3 possible views into the TransitionViewDiv?. I did findin dojo 1.9 specs an example using dojox/mobile/ViewController but I was not able to get the sample code to work as documented by dojo.
2) Is the approach of having the TransitionViewDiv necessary, or could View1, 2 or 3 be loaded without TransitionViewDiv? Keep in mind that each view View1, 2, and 3 are defined as individual Div's.
Appreciate any advice to accomplish the above approach, or welcome any suggestion on the best practices to accomplish the transition.
Scenario #2:
As a follow-on to the scenario 1 above. Once View1, 2 or 3 is successfully loaded the views will have buttons defined that will want to cause the transition to another one of the remaining views. So, if inside SetView.js the decision is to slide in View2 to be displayed, View2
will have buttons that will want to load for example View3.html.
Description + Questions:
1) Would the best approach to load View3.html from View2.html be to use the moveTo on the button click, or should the button use the callback to invoke javascript to cause the transition similar to what was used to load the initial view?
Appreciate any advice on the best practices to managing multiple view stored in independent files. In the end the application will have upwards of 15+ ViewXX.html files each containing a Div. Based on this, having all of the views in one html file and forcing the hide, and show is not feasible.
Appreciate your time and help
To load an HTML fragment (View1.html, View2.html or View3.html), you can use the dojox/mobile/ContentPane. This widget allows you to provide a href property that can be used to specify the location of the view.
You can also alter it later on by setting the href property again, for example:
registry.byId("myContentPane").set("href", "View2.html");
You should keep the div#TransitionViewDiv and programmatically add the dojox/mobile/ContentPane to it, or use declarative syntax and add the following attributes:
<div id="TransitionViewDiv" data-dojo-type="dojox/mobile/ContentPane" data-dojo-props="href: 'View1.html'"></div>
Your second scenario is differs from the first one. In the first one, you actually have 1 view with many fragments, while in your second scenario you have many views.
If you only have 1 view, you cannot transition to other views (there are none). So if you want to use transitions you cannot use dojox/mobile/ContentPane.
However, if you have seperate views, then that means you need to move the header to each view (since they're part of it). For these, more complex cases I think you should look at the dojox/app module. This covers a lot of the MVC code for you and the only thing you need to do is configure it.
If you're not interested in the dojox/app module, you can try to inherit views. You might want to look at this answer I once provided. In the comment section of that answer you can also find a more detailed JSFiddle. In this example the header is actually inherited. I also wrote a more detailed article to handle this case .
I am working with ExpressionEngine and the Structure add-on latest versions.
I am looking for some help on generating a 4 items nav bar, where two items are in a different Structure level.
My Structure rows look like this:
Services (parent)
-----Translation (child)
-----Videos (child)
Studio (parent)
About us (parent)
And my main navigation bar should be:
Translation - Videos - Studio - About us
As you can see they are at different levels. I've been working for hours trying even with Structure Entries add-on, but I simply cannot find a way to generate this navigation.
I'd appreciate any help you can give me.
Structure ultimately stores its 'pages' as regular EE entries, so you could just hardcode the entry ID's into a channel:entries loop to grab the titles:
<ul>
{exp:channel:entries dynamic="no" entry_id="3|4|6|12" disable="categories|pagination|member_data"}
<li>{title}</li>
{/exp:channel:entries}
</ul>
Alternatively if you want to give the client complete control of what appears in the navigation you could set up a custom checkbox field for 'show on primary nav' and then use something like this:
<ul>
{exp:channel:entries dynamic="no" search:options="=show on primary nav?" disable="categories|pagination|member_data"}
<li>{title}</li>
{/exp:channel:entries}
</ul>
(would be a bit nicer using P&T Checkboxes instead of the crappy built-in ones.)
No offence to the Structure module but it's this kind of thing that would really make me use Nav-EE instead, or, more often I would opt for an entirely hardcoded top level nav... The client usually thinks they need full control but in reality that top level navigation is seldom going to change (and when it does the chances are they'll need you to do other work on the site anyway!).
You are looking for Structure Entries: http://devot-ee.com/add-ons/structure-entries (which is free!)
I managed to get the Structure way to get that navigation, which is quite a code:
<ul class="myclass">
{exp:structure:nav start_from="/services" include_ul="no"}
{exp:structure:nav exclude="21|4|10" max_depth="1" include_ul="no"}
</ul>
I found the James' hard coding solution a best option. Thank you!
Since you've already spent significant time fiddling with Structure Entries, I'd say give Taxonomy a shot since it will surely get the job done with very little fuss. http://devot-ee.com/add-ons/taxonomy
For the purpose of managing a nav tree that doesn't fit the mold, the Taxonomy module is very handy and flexible.