Crossfilter query

Question

Is it possible to filter a crossfilter dataset which has an array as the value?

For example, say I have the following dataset:

var data = [
  {
    bookname: "the joy of clojure",
    authors: ["Michael Fogus", "Chris Houser"],
    tags: ["clojure", "lisp"]
  },
  {
    bookname: "Eloquent Ruby",
    authors: ["Russ Olsen"],
    tags: ["ruby"]
  },
  {
    bookname: "Design Patterns in Ruby",
    authors: ["Russ Olsen"],
    tags: ["design patterns", "ruby"]
  }
];

Is there an easy way to access the books which are tagged by an particular tag? And also the books which have a particular author? The way I understand how to use crossfilter so far has me doing something like this:

var filtered_data = crossfilter(data);
var tags = filtered_data.dimension(function(d) {return d.tags});
var tag = tags.group();

And then when I access the grouping (like so):

tag.all()

I get this:

[{key: ["clojure", "lisp"], value: 1}, 
 {key: ["design patterns", "ruby"], value: 1}, 
 {key: ["ruby"], value: 1}]

When I would rather have this:

[{key: "ruby", value: 2}, 
 {key: "clojure", value: 1}, 
 {key: "lisp", value: 1},
 {key: "design patterns", value: 1}]

Answer 1

I've added comments to the code below. Big picture: use reduce function.

var data = ...
var filtered_data = crossfilter(data);
var tags = filtered_data.dimension(function(d) {return d.tags});

tags.groupAll().reduce(reduceAdd, reduceRemove, reduceInitial).value()

Notice how I've used groupAll() instead of group() b/c we want our reduce functions (defined below) to operate on one group rather than 3 groups.

Now the reduce functions should look like this:

/*
 v is the row in the dataset

 p is {} for the first execution (passed from reduceInitial). 
 For every subsequent execution it is the value returned from reduceAdd of the prev row
*/
function reduceAdd(p, v) {
  v.tags.forEach (function(val, idx) {
     p[val] = (p[val] || 0) + 1; //increment counts
  });
  return p;
}

function reduceRemove(p, v) {
   //omitted. not useful for demonstration
}

function reduceInitial() {
  /* this is how our reduce function is seeded. similar to how inject or fold 
   works in functional languages. this map will contain the final counts 
   by the time we are done reducing our entire data set.*/
  return {};  
}

Answer 2

I've never used "crossfilter" (I'm assuming this is a JS library). Here are some pure JS methods though.

This...

data.filter(function(d) {
  return d.authors.indexOf("Michael Fogus") !== -1;
})

returns this:

[{bookname:"the joy of clojure", authors:["Michael Fogus", "Chris Houser"], tags:["clojure", "lisp"]}]

This...

var res = {};
data.forEach(function(d) {
  d.tags.forEach(function(tag) {
    res.hasOwnProperty(tag) ? res[tag]++ : res[tag] = 1
  });
})

returns this:

({clojure:1, lisp:1, ruby:2, 'design patterns':1})

To either of these, you can apply d3.entries to get your {key:"ruby", value: 2} format.