Question

I have a Django admin page for a nested category list like this:

I wrote this script to sort the list and present it hierarchically:

{% extends "admin/change_list.html" %}
{% load i18n %}
{% block footer %}

<script>
(function(){

  var rows=document.getElementById('result_list').getElementsByTagName('tr'),
      table=rows[1].parentNode||rows[1].parentElement,
      i=0, r,      // skip the first row
      data={};     // store category data

  while (r=rows[++i]) {
    var catName=r.getElementsByTagName('a')[0],
        k=catName.innerHTML,
        opts=r.getElementsByTagName('select')[0],
        j=-1, opt;
    while (opt=opts[++j]) {
      if (!opt.selected) continue;
      data[k] = {  
        title:        k,
        children:     {},
        parentName:   opt.innerHTML, 
        parentId:     opt.value, 
        catName:      catName,
        row:          r
      } 
    }
  }

  for (var sub in data) {
    if (data[sub].parentName == sub) continue;
    for (var sup in data) {
      if (sup == data[sub].parentName) {
        data[sup].children[sub]=data[sub];
        data[sub].parent = data[sup];
        break;
      }
    }
  }

  var alt = 0;
  for (var leaf in data) {
    if (data[leaf].parentName != leaf) continue;
    walk(data[leaf], leaf, function (node, nodeName) {
      var n=node, t=n.title;
      while (n=n.parent) {
        t = ' &middot; &nbsp;' + t;
      }
      node.catName.innerHTML = t;
      node.row['class']=node.row['className']='row'+alt++%2;
      table.removeChild(node.row);
      table.appendChild(node.row);
    });
  }

  function walk (leaf, leafName, cb) {
    if (cb) cb(leaf, leafName);
    leaf.ready = true;
    for (var kid in leaf.children) {
      if (leaf.children[kid].ready) continue;
      walk(leaf.children[kid], kid, cb);
    }
  }

}());
</script>

{% endblock %}

…the script runs fine and the list looks like this:

My question is: I feel like the script is prone to memory leaks in UAs with weak garbage collection because of the circular references created by the parent / child stuff. Is this something I should be worried about? Is there a better way to write the script? Should I be deleting a bunch of stuff at the end of the script, and if so, what?

Answer 1

I see some minor leaks, IE garbage collector has a problem cleaning live node references inside functions. This is because the DOM and JavaScript both have it’s own garbage collector and basically, they don’t want to pick a fight with eachother over a reference.

Since you call this script once per page? The memory leak is minute and can actually be ignored, unless people open 100+ pages in one session. Cleaning up is nicer though.

{% extends "admin/change_list.html" %}
{% load i18n %}
{% block footer %}

<script>
(function(){

  var rows=document.getElementById('result_list').getElementsByTagName('tr'),
      table={},
      i=0, r,      // skip the first row
      data={};     // store category data

  // table is now a JS object with a el reference to an element.
  table.el = rows[1].parentNode||rows[1].parentElement;

  while (r=rows[++i]) { // you skip the first row, that correct? Else use i++
    var catName=r.getElementsByTagName('a')[0],
        k=catName.innerHTML,
        opts=r.getElementsByTagName('select')[0],
        j=-1, opt;
    while (opt=opts[++j]) {
      if (!opt.selected) continue;
      data[k] = {  
        title:        k,
        children:     {},
        parentName:   opt.innerHTML, 
        parentId:     opt.value, 
        catName:      catName,
        row:          r
      } 
    }
  }
  // nullify node references
  r = catName = opt = rows =  null;

  for (var sub in data) {
    if (data[sub].parentName == sub) continue;
    for (var sup in data) {
      if (sup == data[sub].parentName) {
        data[sup].children[sub]=data[sub];
        data[sub].parent = data[sup];
        break;
      }
    }
  }

  var alt = 0;
  for (var leaf in data) {
    if (data[leaf].parentName != leaf) continue;
    walk(data[leaf], leaf, function (node, nodeName) {
      var n=node, t=n.title;
      while (n=n.parent) {
        t = ' &middot; &nbsp;' + t;
      }
      node.catName.innerHTML = t;
      node.row['class']=node.row['className']='row'+alt++%2;
      // if table wasn't a JS object, this closure would not have been cleaned up.
      // a refence to table is kept, not to a live DOM element.
      table.el.removeChild(node.row);
      table.el.appendChild(node.row);
    });
  }


  function walk (leaf, leafName, cb) {
    if (cb) cb(leaf, leafName);
    leaf.ready = true;
    for (var kid in leaf.children) {
      if (leaf.children[kid].ready) continue;
      walk(leaf.children[kid], kid, cb);
    }
  }

}());
</script>

{% endblock %}

Was a bit confusing to sort out, since your JS object names imply they are DOM elements =P but I think I figured it out correctly. But you might want to go over the code and nullify other DOM elements I might’ve overlooked (once you’re done with them ofcourse).