Based on this article on MSDN: How To Determine When a Page Is Done Loading in WebBrowser Control, and from past discussions on StackOverflow, I would assume that in case of a document with multiple frames, the DocumentComplete event would fire multiple times, and the last time would be for the top level frame.
However, using the exact sample code from the above-mentioned MSDN link, I find that if there are multiple DocumentComplete events when doing a Navigate to a URL, the condition is satisfied in the following code the first time, not the last time as the article seems to indicate. Subsequent invokes of DocumentComplete seem to be for lower level frames, since the condition fails.
IUnknown* pUnk;
LPDISPATCH lpWBDisp;
HRESULT hr;
pUnk = m_webBrowser.GetControlUnknown();
ASSERT(pUnk);
hr = pUnk->QueryInterface(IID_IDispatch, (void**)&lpWBDisp);
ASSERT(SUCCEEDED(hr));
if (lpDisp == lpWBDisp )
{
// Top-level Window object, so document has been loaded
TRACE("Web document is finished downloading\n");
}
lpWBDisp->Release();
I am not sure why the behaviour I observe is exactly opposite of what it is supposed to be as per the documentation. Any pointers on this would be appreciated.
Background: I am using this code in a dialog-based VC++ / MFC application, and in the DocumentComplete event I want to get certain statistics when the document is fully loaded. So I was trying to use the above code to detect that a particular instance of DocumentComplete firing is when the page has fully loaded.
IMO, the most reliable way to get notified when the page has been fully loaded is to attach to window.onload DOM event for the top window object (IWebBrowser2::get_Document, IHTMLDocument2::get_parentWindow), when DocumentComplete is fired for the first time for a particular navigation. Then, onload event for the top web page will be fired when all inner frames have been loaded. This answer illustrates how it can be done in C# and this answer may help to get it done in C++.
The MSDN doc seems right to me: the last DISPID_DOCUMENTCOMPLETE is the one fired for the main frame.
I can't reproduce your problem for http://www.microsoft.com/ as that link gives me the final http://www.microsoft.com/fr-fr/default.aspx which is a single frame.
I don't like the way the sample code is testing for the Main Browser (equality of 2 IDispatch pointers). What I do is this:
QueryInterface the IDispatch for an IWebBrowser2
Make a true COM equality test, that is "comparing the IUnknown" from the 2 IWebBrower2 (I do it with the IsEqualObject method from the CComPtr template.
Related
I want to check the list of event listeners that are added. For example, I used the code cy.on('pan zoom resize', update); and added function called update in for loop. I do this many times. I also call cy.off('pan zoom resize', update); to remove the event listeners but I want to be sure about it.
The only think I can think of is using console.log but this method might not be helpful.
I also think that in some places people forgot to remove the event listeners and just always added. With too many repetitions this might cause problems.
There is a data field in the private cytoscape object called listeners. You can see that if you:
console.log() the cy object,
navigate to _private,
then open the emitter object
and lastly go to listeners
This is the array listing all the default and user defined event listeners with some metadata like the event, type and scope of the listener.
You can access this in your code by simply calling
cy.emitter().listeners
The question now is, why do you need this information in the first place? Normally, you should be just fine if you call cy.off('eventXY', ...) before using any cy.on('eventXY', ...). Are you sure you need this for your application to work? Maybe elaborate more on the core problem (why you want these information in the first place).
Thanks and have a great day!
I'm trying to get voiceover to work on safari however, it seems when I tab through elements it doesnt read out the aria-label of the new input box in a certain scenario.
Scenario:
When tabbing onto the next element and the on blur of the current element does something to the dom then it will not read out the aria-label of the next element.
Here is an example
http://plnkr.co/edit/x0c67oIl0wlQEguBIQVZ?p=preview
Notice if you take out the onblur function below then it works fine.
<input id="test" onblur="blurer()" onfocus="focuser()"/>
In this case, the issue isn't the presence of a blurer, but rather the contents of your blurer and corresponding focuser functions. Together these two functions are toggling the hidden state of a nearbye element. This is interupting the announcement. There's a role announcement that also occurs. The full annoucement (when text is populated in the edit text control) should be:
"The edited text" contents selected/unselected, "your aria label", edit text.
The quoted portions are parts you control, the other portions are parts controlled by the OS/VoiceOver's interaction with it, calculated automatically by the state of the control and other aria values.
The announcement we're getting is simply
"The edited text"
So, it's not an issue with the aria-label specifically. But rather, you are causing the entire announcement of the element to be interrupted.
When your blur and focus functions trigger you muck with the VoiceOver's response (or the OS's communication of) these events. Not sure what about your functions is causing this. Regardless, a trick that helps in these circumstances is to add a setTimeout to your code. By separating your function and the actual focus/blur event, you can allow the accessibility APIs to do their thing, before mucking with styles and such on the page. Here is an example that makes your little code snippet work. Just replace the contents of your javascript file with this:
function blurer(){
window.setTimeout(function() {
document.getElementById('myDiv').style.display = 'none';//
}, 0);
}
function focuser(){
window.setTimeout(function() {
document.getElementById('myDiv').style.display = 'block';//
}, 0);
}
In general I like to avoid setTimeouts because they create race conditions. However, setTimeouts of 0 are acceptable, because there is no race condition. You're just decoupling the firing event and the execution of your code by pushing your code to the end of the queue. When hacking around VoiceOver, setTimeout(someFunction, 0) works quite well for a lot of cases.
I'm using openui5. There is a constructor Function for UI control Button,unable to see the prototype properties of the Button but the same thing when executed in browser console, shows up!
sap.m.Button.prototype.Move = function(){
console.log('Move');
}
var oButton = new sap.m.Button({text:"Hello"});
oButton.Move(); // throws undefined function!
The same code when executed browser in console, it works!
jsbin --> http://jsbin.com/tepum/1/edit
After running the code I find that creating the first instance of sap.m.Button causes script to change the prototype of sap.m.Button. It's valid in JavaScript but not very smart if you ask me.
A first creation causes a synchronous request (no no as well) to fetch library-parameters.json.
If you run the code the second time it will have prototype.move because creating an instance of Button will not change the Button.prototype.
The capital M in Move would suggest a constructor function so I would advice changing it to lower case.
Since fetching the parameters is synchronous you can create the first instance and then set the prototype:
console.log("First Button creation changes Button.prototype");
var oButton = new sap.m.Button({text:"Hello"});
sap.m.Button.prototype.move = function(){
console.log('Move');
}
oButton.placeAt('content');
oButton.move(); // logs Move
My guess is that this is done to lazy load controls, if a Button is never created then the json config files are never loaded for these unused controls. It has a couple of drawbacks though.
You have to create an instance first before you can set the prototype.
The config files are synchronously loaded so when creating first instance of many controls with a slow connection would cause the app to be unresponsive.
A better way would be for a factory function to return a promise so you create the control the same way every time and the config files can be fetched asynchronously.
[update]
Looking at the config it seems to be config for the whole gui library so I can't see any reason why this is loaded only after creating a first instance. A library that changes it's object definitions when creating instances is not very easy to extend because it's unpredictable. If it only changes prototype on first creation then it should be fine but it looks like the makers of the library didn't want people to extend it or they would not make the object definition unpredictable. If there is an api documentation available then maybe try to check that.
[update]
It seems the "correct" way to extend controls is to use extend.
#HMR is right the correct way to extend a control is by using the extend function provided by UI5 managed objects, see http://jsbin.com/linob/1/edit
in the example below when debugging as mentoned by others you will notice that the control is lazy loaded when required, any changes you make prior are lost when loaded
jQuery.sap.declare("my.Button");
jQuery.sap.require("sap.m.Button");
sap.m.Button.extend("my.Button", {
renderer: {}
});
my.Button.prototype.Move = function() {
console.log('Move');
};
var oButton = new my.Button({
text: "Hello"
});
oButton.placeAt('content');
oButton.Move();
It's not hiding the prototype per se. If a constructor function exits normally then you get that function's prototype. But, if a constructor function actually returns some other object then you get that other object's prototype, so it's not valid to assume that just because you added to the Button prototype that when you call new Button() that you will see your method on whatever you get back. I'm sure if you de-obfuscate that code you'll find that the constructor you are calling has a "return new SomeOtherInstanceOfButton()" or similar at the end of it.
Edit: Ok it's a bit difficult to see what's really going on in that sap code but, it looks like they have code that overwrites the prototypes of controls to add features to them, such as: sap.ui.core.EnabledPropagator, and those things aren't run until you actually instantiate a button. So if you change your code to instantiate the button on the page, then add to it's prototype, then construct and call the method, it works fine. Like so:
http://jsbin.com/benajuko/2/edit
So I guess my answer is, when you run it from console it's finished mucking around with that prototype, whereas in your test you were adding to the prototype, then constructing the button for the first time (which changes the prototype again) then trying to call your old one, which is no longer there.
I have a seaside application with a master-detail page. The master page has a table that consists of a list of tr records. When the user clicks a particular tr element, I want to call a detail component, which'll show the individual record's data.
Since I cannot make a tr element with callback or have it contain an anchor with a callback, I want the tr's onClick property to have some JavaScript which'll call: subcomponent . When I tried this, I got an error saying call: can only be used in callbacks and tasks.
Using ajax is a workaround, however it breaks the back button.
Edit:
More generally, I'd like to know how to set callback like behaviour for various JavaScript events.
Well, you cannot render a component in a tr element, but you could add some anchor or other element in one of its td children.
For my project I did roughly the following: I added an anchor to each row with a special css class, e.g. '.dblclick-action'. This anchor has a normal Seaside callback.
Then I bound a dblclick handler to the tr element that does something like document.location=$(this).find('.dblclick.ction').get(0).href;
I am not close to a Smalltalk image now to give you source code, but I hope you get the idea: you don't use Ajax to click the link in that particular row, but instead have the browser navigate to the callback that is associated to the link in that row. You could say you use the tr.'s dblclick handler to click the link and then let the normal Seaside stuff do its work. No magic there. You can find a little bit more info here.
If you don't want the anchor to be visible you may want to experiment with making the anchor invisible (display: none) or the like.
If you are a bit more experiment friendly, you can also try saving a callback on the server and render its url with callback id as an attribute of the tr element and use the dblclick handler to follow the link from that attribute you extract the value of an attribute in query using attr().
I forgot to answer your initial question: you cannot issue a call: from javascript. But you can use the document.location trick to re/misuse an existing link to a callback on the page using the technique I described in my first answer.
Currently most of the SVG framework provides drag and dropping. I have used RaphaelJs , Extjs Drawing , and SVG.js(http://www.svgjs.com/). All of the framework provides the event handling method by binding it with the element itself. For example:
dragStart: function(event){
//'this' refers to the element itself
this.doSomething();
this.moveTo(event.x,event.y);
};
However the consequences of this is that browser performance is greatly degraded, when there are more than hundred of elements and some logic processing aside. My elements will be composite elements, meaning maybe some text or path in a rect but the rect should be the target (or this) of the event when dragged. The elements will have other event to listen to , for example onclick, onDblClick, onHover & etc.
My question is, is there any way to optimize this ? The user experience is bad after the application is showing a lot of the composite elements I mentioned above.
UPDATE:
I have built an application using the mean that i spoke of above. Binding the event handler to each of the objects. The result is not very user friendly when the element is listening to onhover and onmousedown events. I am thinking to optimize the application but not sure how. The effect is more obvious in FF, chrome is better.
Are you sure you need SVG? Fabric.js is an awesome API for scaling, rotating, drag-drop, and grouping, based on the Canvas element.