Develop Reference

Why does knockout template binding stop working after manually reordering - and reverting - dom items?

I am using a knockout foreach (more specifically, template: { foreach: items }) binding to display a list of elements.
I then proceed to take the following actions:
Swap the first and second elements of the observable array. I see the changes reflected on screen, as expected.
Repeat the previous action to revert to the initial state. Again, this works as expected.
Now, swap the first and second DOM elements. I see the changes reflected on screen, as expected.
Repeat the previous action to revert to the initial state. Again, this works as expected.
Even though we have manually tampered with the DOM, we have reverted to exactly the initial state, without invoking knockout during the DOM tampering. This means the state is restored to the last time knockout was aware of it, so it should look to knockout as if nothing ever changed to begin with.
However, if I perform the first action again, that is, swap the first two elements in the array, the changes are not reflected on screen.
Here is a jsfiddle to illustrate the problem: https://jsfiddle.net/k7u5wep9/.
I know that manually tampering with the DOM managed by knockout is a bad idea and that it can lead to undefined behaviour. This is unfortunately unavoidable in my situation due to third party code. What stumps me is that, even after reverting the manual edits to the exact initial state, knockout still does not work as expected.
My question is: what causes this behaviour?
And then, how does one work around it?

Turns out there is nothing magical happening here. The mistake I made was to only consider elements instead of all nodes. The knockout template binding keeps a record of all nodes when reordering, not just elements.
Before manually editing the DOM, the child nodes of the template binding are:
NodeList(6) [text, div, text, text, div, text].
After manually swapping the first two elements using parent.insertBefore(parent.children[1], parent.children[0]), this turns into:
NodeList(6) [text, div, div, text, text, text].
Repeating the action yields:
NodeList(6) [text, div, div, text, text, text].
Although this is identical to the initial state when only referring to elements, it is quite different when referring to all nodes.
The solution now becomes clear. One way to perform a proper manual swap is to replace
parent.insertBefore(parent.children[1], parent.children[0]);
with
let nexts = [parent.children[0].nextSibling, parent.children[1].nextSibling];
parent.insertBefore(parent.children[1], nexts[0]);
parent.insertBefore(parent.children[0], nexts[1]);
as seen in https://jsfiddle.net/k7u5wep9/2/.
Obviously more care has to be taken when there are no text nodes before/after, but the idea remains the same.

By manipulating the DOM, you have broken the binding made.
Do not manipulate directly the DOM. Knockout will not detect the changes made.

If you put a with: items around your foreach, it at least keeps working but requires double click if dom order != array order .. might get you on track atleast, maybe you can re-order the ko-array inside the dom function to keep their 'orders' in sync?
let vm = {
items: ko.observableArray(['item1', 'item2']),
reorder_array() {
vm.items([vm.items()[1], vm.items()[0]]);
},
reorder_dom() {
let parent = document.querySelector('#items');
parent.insertBefore(parent.children[1], parent.children[0]);
vm.reorder_array();
}
};
ko.applyBindings(vm);
<script src="https://cdnjs.cloudflare.com/ajax/libs/knockout/3.4.2/knockout-min.js"></script>
<div data-bind="with: items">
<div id="items" data-bind="template: { foreach: $data }">
<div data-bind="text: $data"></div>
</div>
</div>
<button data-bind="click: reorder_array">Reorder array</button>
<button data-bind="click: reorder_dom">Reorder DOM</button>
<div>
Reorder the array twice, then reorder DOM twice. This should work as expected, and end up with the initial state. Then, try to reorder the array again. It should not work. Why?
</div>

AngularJS conditional bind-once?

I'm currently working on a project that will need to display editable tables of arbitrary dimensions. I'm using AngularJS, but as the tables can get very large I expect things will get slow if I don't use some form of bind-once. The problem is, as far as I can tell, bind-once is absolute -- I want it to actually keep or at least reinstate all the watchers for rows currently being edited.
Here's the basic idea of how I'm currently doing this.
<tr ng-repeat="row in rows track by key" hackonce>
<td ng-repeat="col in cols">
<div ng-click="editStuff()">{{contents}}
directive('hackonce', function() etc
link: function(scope, elem, attrs) {
if (!scope.row.editing)
setTimeout(function(){ scope.$destroy() }, 0)
}
I'm generating a string key on the row from its various identifiers. When the item is flagged for editing, the controller switches it out for a copy and changes its key to include what will eventually be transaction lock data. This triggers a replacement in the ng-repeat, and the next time the tr is constructed with the replaced row data, the scope is left intact and updates as you edit stuff.
Somehow, it actually works (so far anyway). It drops me from over 500 watchers on a small list to under 100, and what remains is mostly top-level controls, so I hope it will scale well. It also seems to retain and properly clean up ng-click listeners and such even though the scope is gone.
However, it's the sort of hack that I cringed just typing it out in brief, and I get the feeling it may induce vomiting in sane people who know this library better than I do. I was wondering if there was a cleaner and/or safer way to do the same thing. What I'm looking for is a better way to conditionally bind once, or have an element/scope otherwise skip its $digest for itself and all its children based on some exposed flag.
Is there any way to do this that doesn't involve taking a meat cleaver to the scope or manually reattaching all the jQuery garbage I was hoping to avoid with Angular?

It's kind of difficult without more implementation detail, but I'm assuming that you're using some mechanism to show/hide the read-only/editable elements for each row? If you use ng-if to do it, it will attach the elements to the DOM and re-bind when the condition is true and detach when false.
So, something like:
<td ng-repeat="col in cols">
<span ng-if="!isEditing(row)">Read-only stuff here</span>
<span ng-if="isEditing(row)">Editable stuff here.</span>
</td>
I created a really simple example plunk here:
http://plnkr.co/edit/SkdgTuK5UvEEsZmzWat6?p=preview

Getting an Element in AngularJS

It seems that getting an element in AngularJS is a bad idea, i.e. doing something like:
$('.myElement')
in say, a controller is not an angular way of doing things.
Now my question is, how should I get something in angular?
Right now, what I'm doing (and is an accepted way of doing it) is by watching a variable, and my directive does something based on it.
scope.$watch('varToWatch', function (varToWatch) {
if(attrs.id == varToWatch)
{
//Run my Directive specific code
}
});
However, while this particular design works for most cases, watch is an expensive operation, and having lots of directives watching can really slow down your application.
TL:DR - What is an angular way of getting a directive based on a variable on the directive? (like the one above)?

If you want to get/set values you don't need to fetch the element using jQuery. Angular data binding is the way to do it.
directives is the way to go if you want to do animations or any kind of element attributes and DOM manipulation.

Your code is basically right; the directive should watch something in the $scope and perform it's logic when that thing changes. Yes, watch statements are expensive, and that is a problem once your number of watches start to approach ~2000.
Looking at your code though, I see one problem:
The variable $scope.varToWatch references an id in the template.
When this variable changes, you want something to happen to the element which has this id.
The problem here is in the first point: The controller should know nothing about the DOM, including the id of any element. You should find another way to handle this, for example:
<div my-directive="one"> ... </div>
<div my-directive="two"> ... </div>
<div my-directive="three"> ... </div>
...etc
And in your directive:
scope.$watch('varToWatch', function (varToWatch) {
if(attrs.myDirective == varToWatch)
{
// Run my Directive specific code
}
});

You are very vague as to what you're trying to achieve, but I'll try to answer in context of your last comment.
I have a lot of the same directives (therefore the code will run on all of them), but I need to get only one directive from the lot.
You talk a lot about getting the right element. The directive element is passed to the link function in the directive. If you are not using this element (or children of it) directly, but rather trying to search for the element you want somehow, you are most likely approaching the problem the wrong way.
There are several ways to solve this, I'm sure. If you're thinking about animations, there is already support for that in Angular, so please don't try reinvent the wheel yourself. For other logic, here are two suggestions:
Secondary directive
If the logic you want to apply to this directive is generic, i.e. it could be applied to other directives in your application, you could create a new directive which works together with directives. You can set prioritization in directive in order to control which directive is executed first.
<main-directive ... helper-directive="{{condition_for_applying_logic}}"></main-directive>
jsFiddle example
Expanding main directive
If the logic is tightly coupled to this directive, you can just create a new attribute, either dynamic or static, and bind to it in the directive. Instead of checking 'attrs.id == varToWatch', you check if $scope.apply-logic === 'true' and apply the logic then.
<main-directive ...></main-directive> <!-- Not applied here -->
<main-directive apply-logic="true" ...></main-directive> <!-- Applied here -->
<main-directive apply-logic="{{some.varOnScope}}"...></main-directive> <!-- Conditional -->
Please comment if something is unclear.

Change classname of div

In my page I want to change the class name of the div whose id is answer1 in div id=question. How can I do that? Thanks.
$('question1 answer1').addClassName('new_name');
<div id="question1">
<div id="answer1" class="old_name">
</div>
</div>
<div id="question2">
<div id="answer2" class="old_name">
</div>
</div>

So, as it has been pointed out, any selector method requires an iterator of some sort, so you can not just apply a method to all objects returned. You will notice in the documentation for Element.select that this is supposed to alleviate having to explicitly convert the object to an array, but I have not had luck with this on jsfiddle. However, I did try the following:
$('answer2').addClassName('new_name');
​
And it worked just fine. I don't know if the issue is that you are trying to traverse the DOM in your original element object (by using question1 answer1) and this requires the object/array iterator, or if it's just a hiccup elsewhere in the code. But in your specific example, since you know the id of the actual element you want to change the class of, the above code should work fine without specificing the parent element or using an array index of any kind.
I will admit that prototypejs threw me off because they use the same method names for the Element objects as Enumerable objects, so I saw the first instance of select and thought it looked pretty straight forward. Having now seen that almost every class/method requires you to set up a selector and convert it or manually traverse it, I will definitly say that this would be much easier with jquery and that your initial comment that they are almost the same is specifically not true in this scenario. Perhaps protojs offers some features that jquery does not, perhaps your code is tied to protojs (I worked on a project for a year that had to use YUI, which is a much bigger nightmare, trust me), but jquery is set up to play nice with prototypejs, so in cases like these, I'd consider using both. In jquery, the code would have been:
jQuery.noConflict(); // Avoids $ fight between jquery and prototypejs
jQuery( '#question1 #answer1' ).addClass('new_name');
or, to remove the old one first:
jQuery.noConflict(); // Avoids $ fight between jquery and prototypejs
jQuery( '#question1 #answer1' )removeClass('old_name').addClass('new_name');
Also, prototype has a toggleClass method that is probably also unnecessarilly esoteric, but maybe you should read up on: toggleClassName
And I wouldn't be this pissy about wasting my time on this (because I don't like to consider getting stuck learning a new framework a waste of time) except that their documentation, while attractive, has the worst examples I've ever seen. They all assume you have some fundamental idea without ever a friendly real world example or hyperlink to the difference between an instance and a class, and their examples distinguishing the two are identical. I'm definitely going to take some time out later today to find the real best answer to your question, out of spite and pride, if nothing else, but if it really comes down to having to iterator every time, just use vanilla js and use this framework when it's actually useful.

Simply $('answer1') will get you a reference to the div element that you want. The $ function returns an element with the given id string (or null if none was found). Note that you don't use a selector string when using $ - it only operates on ids.
The $$ function takes in a CSS selector string and returns an array of all matching elements, or an empty array if nothing matched. If you wanted or needed to go that route, you could locate the same div like this:
$$('#question1 .answer1')[0]
Once you have your element reference using either of the above methods, you can use addClassName or removeClassName or any other element methods available.

Javascript focus() and select() quirk

Im working on a form and getting null or not an object errors in ie.
<form action="#" method="post" name="adv_search">
<input class="inputbox" type="text" name="keyword1" value="none" id="keyword1"/>
</form>
<script>
document.adv_search.keyword1.focus();
document.adv_search.keyword1.select();
</script>
//whereas if I use
<script>
var key1 = document.getElementById('keyword1');
key1.focus();
key1.select();
</script>
//everything is fine
i would like to understand why.
i would like it to work without having the id tag for the input field
thanks in advance
shouldnt the document.formname.fieldname.focus();
and document.formname.fieldname.select();
work?

Your particular example works for me, but if I add another field with the same name:
<input type="text" name="keyword1" />
<input type="text" name="keyword1" />
Then document.adv_search.keyword1.focus() will fail with the error you specify.
The reason is that:
document.adv_search.keyword1
is a shortcut for this syntax (which goes back to DOM Level 0 and the Netscape 2 days!):
document.forms.adv_search.elements.keyword1
(Incidentally, it is better to use this full syntax, instead of relying on the behaviour of the ‘document’ and ‘form’ objects being indexed on names: if a new method is added to HTMLDocument or HTMLFormElement, that might clash with the control name you are using. This is less of an issue when you use the document.forms or form.elements collections. Also, IE mistakenly dumps all names and ids into ‘document’, so if you've got an element with id="adv_search" in addition to the form with that as a name, document.adv_search will return the wrong one.)
Anyway, the DOM Level 0 scripting methods behave slightly curiously when you access an element by name like this. If there is a single matching element, you'll get that one as a standalone object. If, on the other hand, there are more than one, you'll get a list of objects. You can't call focus() or select() on an array-like list, which is why the error appears; you'd have to do something like keyword1[0].focus() when the list was returned.
So you have to decide whether you're going to be using old-school DOM Level 0 methods to access your form controls — in which case you're going to have to cope with sniffing for single-or-multiple-controls — or move to the ID-based methods introduced by ‘DOM Level 1’:
document.getElementById('keyword1').focus();
The ID-based methods are generally a bit more typing (in the script and to add ‘id’s to all elements you wish to access this way, if they don't already have them), but they are simple and unambiguous. (Also you can then drop the name on the <form> itself.)

The ID approach really is best but if you want to go by name, use getElementsByName.
In this case, it might look like this:
<script>
// retrieves array of objects with the name 'keyword1'
// and takes the first one
var key1 = document.getElementsByName('keyword1')[0];
key1.focus();
key1.select();
</script>