Following his suggestion, I made a few tweaks to my original function (and renamed it yet again):

ko.observableArray.fn.withIndex = function (keyName) {        
    var index = ko.computed(function () {
        var list = this() || [];    // the internal array
        var keys = {};              // a place for key/value
        ko.utils.arrayForEach(list, function (v) {
            if (keyName) {          // if there is a key
                keys[v[keyName]] = v;    // use it
            } else {
                keys[v] = v;
            }
        });
        return keys;
    }, this);

    // also add a handy add on function to find
    // by key ... uses a closure to access the 
    // index function created above
    this.findByKey = function (key) {
        return index()[key];
    };

    return this;
};

To use, just tag on the withIndex function call:

var ViewModel = function () {
    this.uniqueNumber = 0;
    this.list = ko.observableArray([]).withIndex('id');
};

In the example above, the objects stored in the “list” observableArray will be indexed by the “id” property.

By adding the withIndex function call to the observableArray creation, an additional function is added to the array object, findByKey.

$("#btnAdd").on("click", function () {
    var id = vm.uniqueNumber++;
    vm.list.push({
            id: id,
            time: new Date().toLocaleTimeString()});
    var data = vm.list.findByKey(id);
});

In the above example, a new “log” object is added to the list, and then retrieved by the key a moment later (dumb, yes, but hopefully instructive).

I intentionally added the findByKey only to array instances that include the indexing functionality added by withIndex. One reason was to not have the function be available for all arrays (as it doesn’t work for non-indexed arrays) and the second was that it makes possible a bit of slick JavaScript closure value hiding. This way, the index is never stored anywhere on the original view model. Instead, it’s kept entirely within the closure.

But there’s more!

What if you want to have multiple keys or don’t like the name of the find function?

Here’s the kicked up a notch version:

ko.observableArray.fn.withIndex = function (keyName, useName) {
    /// keyName == the name of the property used as the index
    ///            value.
    /// useName == when false, a function named findByKey 
    ///            is added to the observableArray.
    ///            when true, the function is named based
    ///            on the name of the index property &
    ///            capitalized (like id becomes findById)
    var index = ko.computed(function () {
        var list = this() || [];    // the internal array
        var keys = {};              // a place for key/value
        ko.utils.arrayForEach(list, function (v) {
            if (keyName) {          // if there is a key
                keys[v[keyName]] = v;    // use it
            } else {
                keys[v] = v;
            }
        });
        return keys;
    }, this);

    var fnName = "";
    if (useName && keyName) {
        var cap = keyName.substr(0, 1).toUpperCase();
        if (keyName.length > 1) {
            fnName = cap + keyName.substring(1);
        } else {
            fnName = cap;
        }
    } else {
        fnName = "Key";
    }

    var fnName = "findBy" + fnName;
    this[fnName] = function (key) {
        return index()[key];
    };

    return this;
};

This optionally uses the name of the key parameter to build a findByXYZ function using the pattern of findBy{KeyName}. It also capitalizes the function name to follow typical JavaScript coding conventions. Here it is in when initialized:

this.list = ko.observableArray([])
                .withIndex('id', true)
                .withIndex('time', true);

(note that they’re chained above) … and in use:

var data = vm.list.findById(id);
data = vm.list.findByTime(time);

You’ll see how the “findBy” functions are named “findById” and “findByTime” as the original call to “withIndex” used the optional second parameter set to true. I suppose this follows a tiny bit of the automatic mix-in style of Ruby on Rails (but is optional).