Question

I have an array of nested JSON structures where they have varying depth and not the same set of keys everywhere:

[
    {
        "name":"bob",
        "salary":10000,
        "friends":[
            {
                "name": "sarah",
                "salary":10000
            },
            {
                "name": "bill",
                "salary":5000
            }
        ]
    },
    {
        "name":"marge",
        "salary":10000,
        "friends":[
            {
                "name": "rhonda",
                "salary":10000
            },
            {
                "name": "mike",
                "salary":5000,
                "hobbies":[
                    {
                        "name":"surfing",
                        "frequency":10
                    },
                    {
                        "name":"surfing",
                        "frequency":15
                    }
                ]
            }
        ]
    },
    {
        "name":"joe",
        "salary":10000,
        "friends":[
            {
                "name": "harry",
                "salary":10000
            },
            {
                "name": "sally",
                "salary":5000
            }
        ]
    }
]

I wanted to use D3 to render this as nested html tables. For example the friends column will have tables showing the name, and salary of the friends of the individual referenced in the row. Sometimes one of these tables will have another level of a sub table.

I imagine the way to do this is by recursively creating tables. I wrote a python program that takes a JSON structure like this, and renders tables within tables, and the easiest way to do that was recursively. I see on the d3.js documentation there is a .each() thing you can call, which I am sure is what I need, I just need a little boost getting there (https://github.com/mbostock/d3/wiki/Selections#wiki-each).

So is there a nice way to do this in D3? I found this great example for rendering a 2d matrix of data as a table Creating a table linked to a csv file. With that tutorial I was able to get the outer most level of this data-structure rendered as a table, but I am stuck on how to go into levels recursively as needed, as of now they just show up as “Object” in the table since I am not treating them differently from normal strings and numbers.

Also I found this other question/answer that is similar to my question, but I really don’t understand javascript well enough to see where/how the recursion is happening and readapt the solution to fit my needs: How do I process data that is nested multiple levels in D3?. Any advice or pointers to tutorials on recursively or iteratively processing nested tree like JSON data-structures in D3 would be much appreciated!

Answer 1

A recursive function would probably be good approach. See code below for one possible implementation (assuming your data is stored in jdata). See the comments in the code for some explanation and see this Gist for a live version: http://bl.ocks.org/4085017

d3.select("body").selectAll("table")
    .data([jdata])
  .enter().append("table")
    .call(recurse);

function recurse(sel) {
  // sel is a d3.selection of one or more empty tables
  sel.each(function(d) {
    // d is an array of objects
    var colnames,
        tds,
        table = d3.select(this);

    // obtain column names by gathering unique key names in all 1st level objects
    // following method emulates a set by using the keys of a d3.map()
    colnames = d                                                     // array of objects
        .reduce(function(p,c) { return p.concat(d3.keys(c)); }, [])  // array with all keynames
        .reduce(function(p,c) { return (p.set(c,0), p); }, d3.map()) // map with unique keynames as keys
        .keys();                                                     // array with unique keynames (arb. order)

    // colnames array is in arbitrary order
    // sort colnames here if required

    // create header row using standard 1D data join and enter()
    table.append("thead").append("tr").selectAll("th")
        .data(colnames)
      .enter().append("th")
        .text(function(d) { return d; });

    // create the table cells by using nested 2D data join and enter()
    // see also http://bost.ocks.org/mike/nest/
    tds = table.append("tbody").selectAll("tr")
        .data(d)                            // each row gets one object
      .enter().append("tr").selectAll("td")
        .data(function(d) {                 // each cell gets one value
          return colnames.map(function(k) { // for each colname (i.e. key) find the corresponding value
            return d[k] || "";              // use empty string if key doesn't exist for that object
          });
        })
      .enter().append("td");

    // cell contents depends on the data bound to the cell
    // fill with text if data is not an Array  
    tds.filter(function(d) { return !(d instanceof Array); })
        .text(function(d) { return d; });
    // fill with a new table if data is an Array
    tds.filter(function(d) { return (d instanceof Array); })
        .append("table")
        .call(recurse);
  });    
}