Recursion in Angular directives

Question

There are a couple of popular recursive angular directive Q&A's out there, which all come down to one of the following solutions:

• manually incrementally 'compile' HTML based on runtime scope state
  • example 1 [stackoverflow]
  • example 2 [angular jsfiddles page]
• don't use a directive at all, but a <script> template which refers to itself
  • example 1 [google groups]

The first one has the problem that you can't remove previously compiled code unless you comprehensibly manage the manual compile process. The second approach has the problem of... not being a directive and missing out on its powerful capabilities, but more urgently, it can't be parameterised the same way a directive can be; it's simply bound to a new controller instance.

I've been playing with manually doing an angular.bootstrap or @compile() in the link function, but that leaves me with the problem of manually keeping track of elements to remove and add.

Is there a good way to have a parameterized recursive pattern that manages adding/removing elements to reflect runtime state? That is to say, a tree with a add/delete node button and some input field whose value is passed down a node's child nodes. Perhaps a combination of the second approach with chained scopes (but I have no idea how to do this)?

Answer 1

Inspired by the solutions described in the thread mentioned by @dnc253, I abstracted the recursion functionality into a service.

module.factory('RecursionHelper', ['$compile', function($compile){
    return {
        /**
         * Manually compiles the element, fixing the recursion loop.
         * @param element
         * @param [link] A post-link function, or an object with function(s) registered via pre and post properties.
         * @returns An object containing the linking functions.
         */
        compile: function(element, link){
            // Normalize the link parameter
            if(angular.isFunction(link)){
                link = { post: link };
            }

            // Break the recursion loop by removing the contents
            var contents = element.contents().remove();
            var compiledContents;
            return {
                pre: (link && link.pre) ? link.pre : null,
                /**
                 * Compiles and re-adds the contents
                 */
                post: function(scope, element){
                    // Compile the contents
                    if(!compiledContents){
                        compiledContents = $compile(contents);
                    }
                    // Re-add the compiled contents to the element
                    compiledContents(scope, function(clone){
                        element.append(clone);
                    });

                    // Call the post-linking function, if any
                    if(link && link.post){
                        link.post.apply(null, arguments);
                    }
                }
            };
        }
    };
}]);

Which is used as follows:

module.directive("tree", ["RecursionHelper", function(RecursionHelper) {
    return {
        restrict: "E",
        scope: {family: '='},
        template: 
            '<p>{{ family.name }}</p>'+
            '<ul>' + 
                '<li ng-repeat="child in family.children">' + 
                    '<tree family="child"></tree>' +
                '</li>' +
            '</ul>',
        compile: function(element) {
            // Use the compile function from the RecursionHelper,
            // And return the linking function(s) which it returns
            return RecursionHelper.compile(element);
        }
    };
}]);

See this Plunker for a demo. I like this solution best because:

1. You don't need an special directive which makes your html less clean.
2. The recursion logic is abstracted away into the RecursionHelper service, so you keep your directives clean.

Update: As of Angular 1.5.x, no more tricks are required, but works only with template, not with templateUrl