Question

If I had data that looked like this:

harvest = [{type: "apple", color: "green", value: 1}, 
           {type: "apple", color: "red", value: 2}, 
           {type: "grape", color: "green", value: 3},
           {type: "grape", color: "red", value: 4 }]

I could sum it by various attributes using d3’s nest.rollup() function:

sum_by = "color";

rollup = d3.nest().key(function(d) {
  return d[sum_by];
}).rollup(function(d) {
  return d3.sum(d, function(g) {
    return g.value;
  });
}).entries(harvest);

Giving me this:

rollup = [{key: "green", values: 4},
          {key: "red", values: 6}]

Which is just what I want.

However the values in my data consist of arrays, all of equal length:

harvest = [{type: "apple", color: "green", values: [1,2,3,4]}, 
           {type: "apple", color: "red", values: [5,6,7,8]}, 
           {type: "grape", color: "green", values: [9,10,11,12]},
           {type: "grape", color: "red", values: [13,14,15,16] }]

Is it possible to combine these in a similar way? To give for example:

rollup = [{key: "green", values: [10,12,14,16]},
          {key: "red", values: [18,20,22,24]}]

I feel this is probably possible using a d3 rollup function (but it doesn’t necessarily have to be done using d3).

RESOLUTION

Thanks to the efforts of @meetamit and @Superboggly I have three solutions:

Version 1 (preferred because it uses reduce() just once and map() just once):

function sumArrays(group) {
  return group.reduce(function(prev, cur, index, arr) {
    return {
      values: prev.values.map(function(d, i) {
        return d + cur.values[i];
      })
    };
  });
}

Version 2:

function sumArrays(group) {
  return group.map(function(h) {
    return h.values;
  }).reduce(function(prev, cur, index, arr) {
    return prev.map(function(d, i) {
      return d + cur[i];
    });
  });
}

Version 3 (for interest because array length can vary):

function sumArrays(group) {
  return group.reduce(function(prev, cur, index, arr) {
    return prev.map(function(d, i) {
      return d + cur.values[i];
    });
  }, [0, 0, 0, 0]);
}

Called like this:

function rollupArrays() {
  return d3.nest().key(function(d) {
    return d[sum_by];
  }).rollup(sumArrays).entries(harvest);
}

And converted to CoffeeScript:

rollupArrays = ->
  d3.nest().key (d) ->
    d[sum_by]
  .rollup(sumArrays).entries(harvest)

sumArrays = (group) ->
  group.reduce (prev, cur, index, arr) ->
    values: prev.values.map (d,i) ->
      d + cur.values[i]

UPDATE

This method isn’t suitable if the function must run, even with one input row. See Part II

Answer 1

One solution uses [].reduce() and [].map():

// eg: sumArrays([ [1,2,3,4], [5,6,7,8] ]);// <- outputs [6, 8, 10, 12]
function sumArrays(arrays) {
  return arrays.reduce(
    function(memo, nums, i) {
      if(i == 0)
        return nums.concat();
      else
        return memo.map(
          function(memoNum, i) {
            return memoNum + nums[i];
          }
        );
    },
    [ ]// Start with empty Array for memo
  );
}

Both reduce and map are not native in old JS, so best use a module (underscore, or maybe there’s a d3 equivalent to reduce, but I haven’t seen it).

EDIT

Using it in your code:

sum_by = "color";

rollup = d3.nest().key(function(d) {
  return d[sum_by];
}).rollup(function(d) {
  var arraysToSum = d.map(function(g) { return g.values; });
  return sumArrays(arraysToSum)
}).entries(harvest);