I'm developing an app with Angular and Semantic-UI. The app should be accessible, this means it should be compliant with WCAG 2.0.
To reach this purpose the modals should keep focus within the dialog and prevents users from going outside or move with "tabs" between elements of the page that lays under the modal.
I have found some working examples, like the following:
JQuery dialog: https://jqueryui.com/dialog/#modal-confirmation
dialog HTML 5.1 element: https://demo.agektmr.com/dialog
ARIA modal dialog example:
http://w3c.github.io/aria-practices/examples/dialog-modal/dialog.html
(that I have reproduced on Plunker)
Here is my try to create an accessible modal with Semantic-UI: https://plnkr.co/edit/HjhkZg
As you can see I used the following attributes:
role="dialog"
aria-labelledby="modal-title"
aria-modal="true"
But they don't solve my issue. Do you know any way to make my modal keeping focus and lose it only when user click on cancel/confirm buttons?
There is currently no easy way to achieve this. The inert attribute was proposed to try to solve this problem by making any element with the attribute and all of it's children inaccessible. However, adoption has been slow and only recently did it land in Chrome Canary behind a flag.
Another proposed solution is making a native API that would keep track of the modal stack, essentially making everything not currently the top of the stack inert. I'm not sure the status of the proposal, but it doesn't look like it will be implemented any time soon.
So where does that leave us?
Unfortunately without a good solution. One solution that is popular is to create a query selector of all known focusable elements and then trap focus to the modal by adding a keydown event to the last and first elements in the modal. However, with the rise of web components and shadow DOM, this solution can no longer find all focusable elements.
If you always control all the elements within the dialog (and you're not creating a generic dialog library), then probably the easiest way to go is to add an event listener for keydown on the first and last focusable elements, check if tab or shift tab was used, and then focus the first or last element to trap focus.
If you're creating a generic dialog library, the only thing I have found that works reasonably well is to either use the inert polyfill or make everything outside of the modal have a tabindex=-1.
var nonModalNodes;
function openDialog() {
var modalNodes = Array.from( document.querySelectorAll('dialog *') );
// by only finding elements that do not have tabindex="-1" we ensure we don't
// corrupt the previous state of the element if a modal was already open
nonModalNodes = document.querySelectorAll('body *:not(dialog):not([tabindex="-1"])');
for (var i = 0; i < nonModalNodes.length; i++) {
var node = nonModalNodes[i];
if (!modalNodes.includes(node)) {
// save the previous tabindex state so we can restore it on close
node._prevTabindex = node.getAttribute('tabindex');
node.setAttribute('tabindex', -1);
// tabindex=-1 does not prevent the mouse from focusing the node (which
// would show a focus outline around the element). prevent this by disabling
// outline styles while the modal is open
// #see https://www.sitepoint.com/when-do-elements-take-the-focus/
node.style.outline = 'none';
}
}
}
function closeDialog() {
// close the modal and restore tabindex
if (this.type === 'modal') {
document.body.style.overflow = null;
// restore or remove tabindex from nodes
for (var i = 0; i < nonModalNodes.length; i++) {
var node = nonModalNodes[i];
if (node._prevTabindex) {
node.setAttribute('tabindex', node._prevTabindex);
node._prevTabindex = null;
}
else {
node.removeAttribute('tabindex');
}
node.style.outline = null;
}
}
}
The different "working examples" do not work as expected with a screenreader.
They do not trap the screenreader visual focus inside the modal.
For this to work, you have to :
Set the aria-hidden attribute on any other nodes
disable keyboard focusable elements inside those trees (links using tabindex=-1, controls using disabled, ...)
The jQuery :focusable pseudo selector can be useful to find focusable elements.
add a transparent layer over the page to disable mouse selection.
or you can use the css pointer-events: none property when the browser handles it with non SVG elements, not in IE
This focus-trap plugin is excellent at making sure that focus stays trapped inside of dialogue elements.
It sounds like your problem can be broken down into 2 categories:
focus on dialog box
Add a tabindex of -1 to the main container which is the DOM element that has role="dialog". Set the focus to the container.
wrapping the tab key
I found no other way of doing this except by getting the tabbable elements within the dialog box and listening it on keydown. When I know the element in focus (document.activeElement) is the last one on the list, I make it wrap
"focus" events can be intercepted in the capture phase, so you can listen for them at the document.body level, squelch them before they reach the target element, and redirect focus back to a control in your modal dialog. This example assumes a modal dialog with an input element gets displayed and assigned to the variable currDialog:
document.body.addEventListener("focus", (event) => {
if (currDialog && !currDialog.contains(event.target)) {
event.preventDefault();
event.stopPropagation();
currDialog.querySelector("input").focus();
}
}, {capture: true});
You may also want to contain such a dialog in a fixed-position, clear (or low-opacity) backdrop element that takes up the full screen in order to capture and suppress mouse/pointer events, so that no browser feedback (hover, etc.) occurs that could give the user the impression that the background is active.
Don't use any solution requiring you to look up "tabbable" elements. Instead, use keydown and either click events or a backdrop in an effective manor.
(Angular1)
See Asheesh Kumar's answer at https://stackoverflow.com/a/31292097/1754995 for something similar to what I am going for below.
(Angular2-x, I haven't done Angular1 in a while)
Say you have 3 components: BackdropComponent, ModalComponent (has an input), and AppComponent (has an input, the BackdropComponent, and the ModalComponent). You display BackdropComponent and ModalComponent with the correct z-index, both are currently displayed/visible.
What you need to do is have a general window.keydown event with preventDefault() to stop all tabbing when the backdrop/modal component is displayed. I recommend you put that on a BackdropComponent. Then you need a keydown.tab event with stopPropagation() to handle tabbing for the ModalComponent. Both the window.keydown and keydown.tab could probably be in the ModalComponent but there is purpose in a BackdropComponent further than just modals.
This should prevent clicking and tabbing to the AppComponent input and only click or tab to the ModalComponent input [and browser stuffs] when the modal is shown.
If you don't want to use a backdrop to prevent clicking, you can use use click events similarly to the keydown events described above.
Backdrop Component:
#Component({
selector: 'my-backdrop',
host: {
'tabindex': '-1',
'(window:keydown)': 'preventTabbing($event)'
},
...
})
export class BackdropComponent {
...
private preventTabbing(event: KeyboardEvent) {
if (event.keyCode === 9) { // && backdrop shown?
event.preventDefault();
}
}
...
}
Modal Component:
#Component({
selector: 'my-modal',
host: {
'tabindex': '-1',
'(keydown.tab)': 'onTab($event)'
},
...
})
export class ModalComponent {
...
private onTab(event: KeyboardEvent) {
event.stopPropagation();
}
...
}
Here's my solution. It traps Tab or Shift+Tab as necessary on first/last element of modal dialog (in my case found with role="dialog"). Eligible elements being checked are all visible input controls whose HTML may be input,select,textarea,button.
$(document).on('keydown', function(e) {
var target = e.target;
var shiftPressed = e.shiftKey;
// If TAB key pressed
if (e.keyCode == 9) {
// If inside a Modal dialog (determined by attribute role="dialog")
if ($(target).parents('[role=dialog]').length) {
// Find first or last input element in the dialog parent (depending on whether Shift was pressed).
// Input elements must be visible, and can be Input/Select/Button/Textarea.
var borderElem = shiftPressed ?
$(target).closest('[role=dialog]').find('input:visible,select:visible,button:visible,textarea:visible').first()
:
$(target).closest('[role=dialog]').find('input:visible,select:visible,button:visible,textarea:visible').last();
if ($(borderElem).length) {
if ($(target).is($(borderElem))) {
return false;
} else {
return true;
}
}
}
}
return true;
});
we can use the focus trap npm package.
npm i focus-trap
This might help someone who is looking for solution in Angular.
Step 1: Add keydown event on dialog component
#HostListener('document:keydown', ['$event'])
handleTabKeyWInModel(event: any) {
this.sharedService.handleTabKeyWInModel(event, '#modal_id', this.elementRef.nativeElement, 'input,button,select,textarea,a,[tabindex]:not([tabindex="-1"])');
}
This will filters the elements which are preseneted in the Modal dialog.
Step 2: Add common method to handle focus in shared service (or you can add it in your component as well)
handleTabKeyWInModel(e, modelId: string, nativeElement, tagsList: string) {
if (e.keyCode === 9) {
const focusable = nativeElement.querySelector(modelId).querySelectorAll(tagsList);
if (focusable.length) {
const first = focusable[0];
const last = focusable[focusable.length - 1];
const shift = e.shiftKey;
if (shift) {
if (e.target === first) { // shift-tab pressed on first input in dialog
last.focus();
e.preventDefault();
}
} else {
if (e.target === last) { // tab pressed on last input in dialog
first.focus();
e.preventDefault();
}
}
}
}
}
Now this method will take the modal dialog native element and start evaluate on every tab key. Finally we will filter the event on first and last so that we can focus on appropriate elements (on first after last element tab click and on last shift+tab event on first element).
Happy Coding.. :)
I used one of the methods suggested by Steven Lambert, namely, listening to keydown events and intercepting "tab" and "shift+tab" keys. Here's my sample code (Angular 5):
import { Directive, ElementRef, Attribute, HostListener, OnInit } from '#angular/core';
/**
* This directive allows to override default tab order for page controls.
* Particularly useful for working around the modal dialog TAB issue
* (when tab key allows to move focus outside of dialog).
*
* Usage: add "custom-taborder" and "tab-next='next_control'"/"tab-prev='prev_control'" attributes
* to the first and last controls of the dialog.
*
* For example, the first control is <input type="text" name="ctlName">
* and the last one is <button type="submit" name="btnOk">
*
* You should modify the above declarations as follows:
* <input type="text" name="ctlName" custom-taborder tab-prev="btnOk">
* <button type="submit" name="btnOk" custom-taborder tab-next="ctlName">
*/
#Directive({
selector: '[custom-taborder]'
})
export class CustomTabOrderDirective {
private elem: HTMLInputElement;
private nextElemName: string;
private prevElemName: string;
private nextElem: HTMLElement;
private prevElem: HTMLElement;
constructor(
private elemRef: ElementRef
, #Attribute('tab-next') public tabNext: string
, #Attribute('tab-prev') public tabPrev: string
) {
this.elem = this.elemRef.nativeElement;
this.nextElemName = tabNext;
this.prevElemName = tabPrev;
}
ngOnInit() {
if (this.nextElemName) {
var elems = document.getElementsByName(this.nextElemName);
if (elems && elems.length && elems.length > 0)
this.nextElem = elems[0];
}
if (this.prevElemName) {
var elems = document.getElementsByName(this.prevElemName);
if (elems && elems.length && elems.length > 0)
this.prevElem = elems[0];
}
}
#HostListener('keydown', ['$event'])
onKeyDown(event: KeyboardEvent) {
if (event.key !== "Tab")
return;
if (!event.shiftKey && this.nextElem) {
this.nextElem.focus();
event.preventDefault();
}
if (event.shiftKey && this.prevElem) {
this.prevElem.focus();
event.preventDefault();
}
}
}
To use this directive, just import it to your module and add to Declarations section.
I've been successful using Angular Material's A11yModule.
Using your favorite package manager install these to packages into your Angular app.
**"#angular/material": "^10.1.2"**
**"#angular/cdk": "^10.1.2"**
In your Angular module where you import the Angular Material modules add this:
**import {A11yModule} from '#angular/cdk/a11y';**
In your component HTML apply the cdkTrapFocus directive to any parent element, example: div, form, etc.
Run the app, tabbing will now be contained within the decorated parent element.
For jquery users:
Assign role="dialog" to your modal
Find first and last interactive element inside the dialog modal.
Check if current target is one of them(depending on shift key is
pressed or not).
If target element is one of first or last interactive element of the
dialog, return false
Working code sample:
//on keydown inside dialog
$('.modal[role=dialog]').on('keydown', e => {
let target = e.target;
let shiftPressed = e.shiftKey;
// If TAB is pressed
if (e.keyCode === 9) {
// Find first and last element in the ,modal-dialog parent.
// Elements must be interactive i.e. visible, and can be Input/Select/Button/Textarea.
let first = $(target).closest('[role=dialog]').find('input:visible,select:visible,button:visible,textarea:visible').first();
let last = $(target).closest('[role=dialog]').find('input:visible,select:visible,button:visible,textarea:visible').last();
let borderElem = shiftPressed ? first : last //border element on the basis of shift key pressed
if ($(borderElem).length) {
return !$(target).is($(borderElem)); //if target is border element , return false
}
}
return true;
});
I read through most of the answers, while the package focus-trap seems like a good option. #BenVida shared a very simple VanillaJS solution here in another Stack Overflow post.
Here is the code:
const container=document.querySelector("_selector_for_the_container_")
//optional: needed only if the container element is not focusable already
container.setAttribute("tabindex","0")
container.addEventListener("focusout", (ev)=>{
if (!container.contains(ev.relatedTarget)) container.focus()
})
I have an editor application with a main editor area in SmartGwt. I handle some native events - keypresses as keyboard shortcuts. I want to disable executing these when there's one or more Window shown on the screen. How to detect if there are any Window descendants shown? /I don't mean browser windows, but SmartGwt Window descendants/
It would be perfectly enough to detect modal windows.
Is there a SmartGwt API way of doing this, or I have to create my own system for registering my windows and query their visible state?
FYI - there seems to be no other way of doing this than creating a common ancestor for my windows and keep track of the visibility changes.
Then call getVisibleCount() when needed.
public class WindowBase extends Window {
private static ArrayList<Window> visibleWindows = new ArrayList<Window>();
VisibilityChangedHandler handlervc = new VisibilityChangedHandler() {
public void onVisibilityChanged(com.smartgwt.client.widgets.events.VisibilityChangedEvent event) {
if ( event.getIsVisible() ) {
if ( !visibleWindows.contains( event.getSource() ) )
visibleWindows.add( (Window)event.getSource() );
} else
visibleWindows.remove( event.getSource() );
}
};
#Override
public void show() {
if ( !visibleWindows.contains( this ) )
visibleWindows.add( this );
addVisibilityChangedHandler( handlervc );
super.show();
}
public static int getVisibleCount() {
return visibleWindows.size();
}
}
here is the part of code:
val lsv_syns = new ListView[String]()
val scp_syns = new ScrollPane() {
listenTo(lsv_syns.mouse.moves,lsv_syns.selection)
reactions += {
case me: MouseExited => {
txf_mot.requestFocus()
}
case SelectionChanged(`lsv_syns`)=> {
println("sélection:"+lsv_syns.selection.items(0))
}
}
}
As you can see, the listView is in a scrollPane; don't pay attention to the mouseExited event, the interesting thing is the selectionChanged, which seems to be called twice when I only click on time on an other line, because the println is called two times.
thanks.
Well i also recently worked with a ListView and now that you mentioned it, it also does my calculation twice.
The answer seems to be related to mouse events. Following the stack trace SelectionChanged is called twice. One coming from the Java event MousePressed and one from MouseReleased.
When you change the selection with KeyEvents it is only called once.
My first (and I guess not nice) idea to avoid the problem would be to ignore one of the events:
reactions += {
case SelectionChanged(`lsv_syns`) if !lsv_syns.selection.adjusting => {
println("sélection:"+lsv_syns.selection.items(0))
}
}
Both ListSelection events share the same data except getValueIsAdjusting. So if you check for it you can avoid doing your stuff twice.
Warning: !lsv_syns.selection.adjusting will result in printing on key release and not on press!
If you put lsv_syns.selection.adjusting it will correspond to key press but it will also filter key events. As I said. Not nice at all...
So I have a cell table with click event selection model working fine.
I later found out you can press UP and DOWN arrows to get the highlighted row to change, but the awful thing is you have to press Space for it to actually call the SelectionChangeEvent.
I am trying to cheat my way a little, by catching the UP and DOWN events and firing the SPACE event. Sadly it doesn't work :(
Here is my code any help would be appreciated!
table.setKeyboardSelectionPolicy(KeyboardSelectionPolicy.ENABLED);
table.sinkEvents(Event.KEYUP);
table.sinkEvents(Event.KEYDOWN);
table.sinkEvents(32);
table.addHandler(new KeyUpHandler(){
#Override
public void onKeyUp(KeyUpEvent event)
{
System.out.println(event.getNativeKeyCode());
if(event.getNativeEvent().getKeyCode() == 40)
{
// down is pressed
int i = rows.getFilterList().indexOf(selectionModel.getLastSelectedObject())+1;
if(i >= 0 && i < rows.getFilterList().size())
{
// selectionModel.setSelected(selectionModel.getLastSelectedObject(), false);
// selectionModel.setSelected(rows.getFilterList().get(i), true);
// SelectionChangeEvent.fire(selectionModel);
System.out.println("firing native event space");
DomEvent.fireNativeEvent(Document.get().createKeyUpEvent(false, false, false, false, 32), table);
}
}
else if(event.getNativeEvent().getKeyCode() == 38)
{
// up is pressed
int i = rows.getFilterList().indexOf(selectionModel.getLastSelectedObject())-1;
if(i >= 0 && i < rows.getFilterList().size())
{
// selectionModel.setSelected(selectionModel.getLastSelectedObject(), false);
// selectionModel.setSelected(rows.getFilterList().get(i), true);
// SelectionChangeEvent.fire(selectionModel);
System.out.println("firing native event space");
DomEvent.fireNativeEvent(Document.get().createKeyUpEvent(false, false, false, false, 32), table);
}
}
}
}, KeyUpEvent.getType());
32 is assumingly the NativeEvent for space, my console prints something like:
40
firing native event space
32
so assumingly the event type 32 is being called for the object table.
I check if the object is selected, because on the right hand side of the screen I have additional information being pulled out from a list, since the cell table doesn't show all the information. I want it so when I press UP and DOWN the RHS information changes and I dont have to press SPACE to prompt the info change
Ideally you would poke into the selection internals. Specifically the DefaultKeyboardSelectionHandler is the default implementation of keyboard navigation and the DefaultSelectionEventManager is the default implementation of selection actions using spacebar/clicks (they are both CellPreviewEvent.Handlers).
Anyway, you can force the keyboard selection to be bound to the underlying SelectionModel by using setKeyboardSelectionPolicy(KeyboardSelectionPolicy.BOUND_TO_SELECTION). It should be fine for your use case. Much like what is done for the CellList showcase sample (the selection API is the same across cell widgets).
Is there any existing mechanism for a GtkEntry to simply position the cursor at the end of the text when focused, rather than selecting its contents to be overwritten by the next key? It seems odd to have to add a signal handler to do something this basic, but I can't find anything in the properties.
Edit: The signal handler doesn't work; whatever I do the default behaviour gets triggered after my handler runs. Here's my gtkd code; note that I am appending some text in the focus-in-event handler, and the appended text gets selected as well:
class NoteView : Entry
{
this(string text) {
if (text) {
setText(text);
}
setEditable(true);
setCanFocus(true);
addOnFocusIn(delegate bool(GdkEventFocus* f, Widget w) {
// clear selection
selectRegion(0, 0);
// test to see whether the appended text gets selected too
appendText("hello");
setPosition(-1);
// don't let any other handlers run
return 1;
}, ConnectFlags.AFTER);
}
}
The addOnFocusIn method is in the gtkd Gtk.Widget api; it calls g_signal_connect_data internally, which should in theory be honouring the G_CONNECT_AFTER flag I'm passing it, but doesn't seem to be.
Edit2: Solved - the grab-focus handler was doing the text selection, and being handled after focus-in-event
Turns out GtkEntry was selecting the text on the grab-focus signal, not focus-in-event. Working code:
class NoteView : Entry
{
this(string text) {
if (text) {
setText(text);
}
setEditable(true);
setCanFocus(true);
setHasFrame(false);
addOnGrabFocus(delegate void(Widget w) {
selectRegion(0, 0);
setPosition(-1);
}, ConnectFlags.AFTER);
}
}
Can't find anything in the docs.
I guess they figured that diverging from the default behavior is that uncommon that they just let people do it with signals, rather than provide a property for it.
Consider creating a subclass of GtkEntry that exhibits the behavior you require.