ArangoDB v3.4 reached End of Life (EOL) and is no longer supported.

This documentation is outdated. Please see the most recent version here: Latest Docs

Accessing Shapes Data


Documents in a collection may have different shapes associated with them. There is no way to query the shapes data directly. So how do you solve this problem?


There are two possible ways to do this.

A) The fast way with some random samplings:

  1. Ask for a random document (db.<collection>.any()) and note its top-level attribute names
  2. Repeat this for at least 10 times. After that repeat it only if you think it’s worth it.

Following is an example of an implementation:


function attributes(collection) {
  "use strict"

  var probes = 10;
  var maxRounds = 3;
  var threshold = 0.5;

  var maxDocuments = collection.count();

  if (maxDocuments < probes) {
    probes = maxDocuments;

  if (probes === 0) {
    return [ ];

  var attributes = { };

  while (maxRounds--) {
    var newDocuments = 0;
    var n = probes;
    while (n--) {
      var doc = collection.any();
      var found = false;
      var keys = Object.keys(doc);

    for (var i = 0; i < keys.length; ++i) {
      if (attributes.hasOwnProperty(keys[i])) {
      else {
        attributes[keys[i]] = 1;
        found = true;

     if (found) {

    if (newDocuments / probes <= threshold) {

  return Object.keys(attributes); 

B) The way to find all top-level attributes

If you don’t mind to make some extra inserts and you don’t care about deletion or updates of documents you can use the following:


function insert(collection, document) {
  var result =;

  try { 
    var keys = Objects.keys(document);

    for (i = 0; i < keys.length; ++i) {
      try {{ attribute: keys[i] });
        catch (err1) {
        // potential unique key constraint violations
  catch (err2) {

  return result;


A) The fast way with some random samplings:

You get some random sampling with bounded complexity. If you have a variety of attributes you should repeat the procedure more than 10 times.

The procedure can be implemented as a server side action.

B) The way to find all top-level attributes:

This procedure will not care about updates or deletions of documents. Also only the top-level attribute of the documents will be inserted and nested one ignored.

The procedure can be implemented as a server side action.

Author: Arangodb

Tags: #collection #database