function mg_process_scale_ticks(args, axis) {
  var accessor;
  var scale_ticks;
  var max;

  if (axis === 'x') {
    accessor = args.x_accessor;
    scale_ticks = args.scales.X.ticks(args.xax_count);
    max = args.processed.max_x;
  } else if (axis === 'y') {
    accessor = args.y_accessor;
    scale_ticks = args.scales.Y.ticks(args.yax_count);
    max = args.processed.max_y;
  }

  function log10(val) {
    if (val === 1000) {
      return 3;
    }
    if (val === 1000000) {
      return 7;
    }
    return Math.log(val) / Math.LN10;
  }

  if ((axis === 'x' && args.x_scale_type === 'log') || (axis === 'y' && args.y_scale_type === 'log')) {
    // get out only whole logs
    scale_ticks = scale_ticks.filter(function(d) {
      return Math.abs(log10(d)) % 1 < 1e-6 || Math.abs(log10(d)) % 1 > 1 - 1e-6;
    });
  }

  // filter out fraction ticks if our data is ints and if xmax > number of generated ticks
  var number_of_ticks = scale_ticks.length;

  // is our data object all ints?
  var data_is_int = true;
  args.data.forEach(function(d, i) {
    d.forEach(function(d, i) {
      if (d[accessor] % 1 !== 0) {
        data_is_int = false;
        return false;
      }
    });
  });

  if (data_is_int && number_of_ticks > max && args.format === 'count') {
    // remove non-integer ticks
    scale_ticks = scale_ticks.filter(function(d) {
      return d % 1 === 0;
    });
  }

  if (axis === 'x') {
    args.processed.x_ticks = scale_ticks;
  } else if (axis === 'y') {
    args.processed.y_ticks = scale_ticks;
  }
}

function raw_data_transformation(args) {
  'use strict';

  // dupe our data so we can modify it without adverse effect
  args.data = MG.clone(args.data);

  // we need to account for a few data format cases:
  // #0 {bar1:___, bar2:___}                                    // single object (for, say, bar charts)
  // #1 [{key:__, value:__}, ...]                               // unnested obj-arrays
  // #2 [[{key:__, value:__}, ...], [{key:__, value:__}, ...]]  // nested obj-arrays
  // #3 [[4323, 2343],..]                                       // unnested 2d array
  // #4 [[[4323, 2343],..] , [[4323, 2343],..]]                 // nested 2d array
  args.single_object = false; // for bar charts.
  args.array_of_objects = false;
  args.array_of_arrays = false;
  args.nested_array_of_arrays = false;
  args.nested_array_of_objects = false;

  // is the data object a nested array?

  if (is_array_of_arrays(args.data)) {
    args.nested_array_of_objects = args.data.map(function(d) {
      return is_array_of_objects_or_empty(d);
    }); // Case #2
    args.nested_array_of_arrays = args.data.map(function(d) {
      return is_array_of_arrays(d);
    }); // Case #4
  } else {
    args.array_of_objects = is_array_of_objects(args.data); // Case #1
    args.array_of_arrays = is_array_of_arrays(args.data); // Case #3
  }

  if (args.chart_type === 'line') {
    if (args.array_of_objects || args.array_of_arrays) {
      args.data = [args.data];
    }
  } else {
    if (!(mg_is_array(args.data[0]))) {
      args.data = [args.data];
    }
  }
  // if the y_accessor is an array, break it up and store the result in args.data
  mg_process_multiple_x_accessors(args);
  mg_process_multiple_y_accessors(args);

  // if user supplies keyword in args.color, change to arg.colors.
  // this is so that the API remains fairly sensible and legible.
  if (args.color !== undefined) {
    args.colors = args.color;
  }

  // if user has supplied args.colors, and that value is a string, turn it into an array.
  if (args.colors !== null && typeof args.colors === 'string') {
    args.colors = [args.colors];
  }

  // sort x-axis data
  if (args.chart_type === 'line' && args.x_sort === true) {
    for (var i = 0; i < args.data.length; i++) {
      args.data[i].sort(function(a, b) {
        return a[args.x_accessor] - b[args.x_accessor];
      });
    }
  }

  return this;
}

function mg_process_multiple_accessors(args, which_accessor) {
  // turns an array of accessors into ...
  if (mg_is_array(args[which_accessor])) {
    args.data = args.data.map(function(_d) {
      return args[which_accessor].map(function(ya) {
        return _d.map(function(di) {
          di = MG.clone(di);

          if (di[ya] === undefined) {
            return undefined;
          }

          di['multiline_' + which_accessor] = di[ya];
          return di;
        }).filter(function(di) {
          return di !== undefined;
        });
      });
    })[0];
    args[which_accessor] = 'multiline_' + which_accessor;
  }
}

function mg_process_multiple_x_accessors(args) {
  mg_process_multiple_accessors(args, 'x_accessor');
}

function mg_process_multiple_y_accessors(args) {
  mg_process_multiple_accessors(args, 'y_accessor');
}

MG.raw_data_transformation = raw_data_transformation;

function process_line(args) {
  'use strict';

  var time_frame;

  // do we have a time-series?
  var is_time_series = d3.sum(args.data.map(function(series) {
    return series.length > 0 && mg_is_date(series[0][args.x_accessor]);
  })) > 0;

  // are we replacing missing y values with zeros?
  if ((args.missing_is_zero || args.missing_is_hidden) && args.chart_type === 'line' && is_time_series) {
    for (var i = 0; i < args.data.length; i++) {
      // we need to have a dataset of length > 2, so if it's less than that, skip
      if (args.data[i].length <= 1) {
        continue;
      }

      var first = args.data[i][0];
      var last = args.data[i][args.data[i].length - 1];

      // initialize our new array for storing the processed data
      var processed_data = [];

      // we'll be starting from the day after our first date
      var start_date = MG.clone(first[args.x_accessor]).setDate(first[args.x_accessor].getDate() + 1);

      // if we've set a max_x, add data points up to there
      var from = (args.min_x) ? args.min_x : start_date;
      var upto = (args.max_x) ? args.max_x : last[args.x_accessor];

      time_frame = mg_get_time_frame((upto - from) / 1000);

      if (['four-days', 'many-days', 'many-months', 'years', 'default'].indexOf(time_frame) !== -1 && args.missing_is_hidden_accessor === null) {
        for (var d = new Date(from); d <= upto; d.setDate(d.getDate() + 1)) {
          var o = {};
          d.setHours(0, 0, 0, 0);

          // add the first date item, we'll be starting from the day after our first date
          if (Date.parse(d) === Date.parse(new Date(start_date))) {
            processed_data.push(MG.clone(args.data[i][0]));
          }

          // check to see if we already have this date in our data object
          var existing_o = null;
          args.data[i].forEach(function(val, i) {
            if (Date.parse(val[args.x_accessor]) === Date.parse(new Date(d))) {
              existing_o = val;

              return false;
            }
          });

          // if we don't have this date in our data object, add it and set it to zero
          if (!existing_o) {
            o[args.x_accessor] = new Date(d);
            o[args.y_accessor] = 0;
            o['_missing'] = true; //we want to distinguish between zero-value and missing observations
            processed_data.push(o);
          }

          // if the data point has, say, a 'missing' attribute set or if its
          // y-value is null identify it internally as missing
          else if (existing_o[args.missing_is_hidden_accessor] || existing_o[args.y_accessor] === null) {
            existing_o['_missing'] = true;
            processed_data.push(existing_o);
          }

          //otherwise, use the existing object for that date
          else {
            processed_data.push(existing_o);
          }
        }
      } else {
        for (var j = 0; j < args.data[i].length; j += 1) {
          var obj = MG.clone(args.data[i][j]);
          obj['_missing'] = args.data[i][j][args.missing_is_hidden_accessor];
          processed_data.push(obj);
        }
      }

      // update our date object
      args.data[i] = processed_data;
    }
  }

  return this;
}

MG.process_line = process_line;

function process_histogram(args) {
  'use strict';

  // if args.binned == false, then we need to bin the data appropriately.
  // if args.binned == true, then we need to make sure to compute the relevant computed data.
  // the outcome of either of these should be something in args.computed_data.
  // the histogram plotting function will be looking there for the data to plot.

  // we need to compute an array of objects.
  // each object has an x, y, and dx.

  // histogram data is always single dimension
  var our_data = args.data[0];

  var extracted_data;
  if (args.binned === false) {
    // use d3's built-in layout.histogram functionality to compute what you need.

    if (typeof(our_data[0]) === 'object') {
      // we are dealing with an array of objects. Extract the data value of interest.
      extracted_data = our_data
        .map(function(d) {
          return d[args.x_accessor];
        });
    } else if (typeof(our_data[0]) === 'number') {
      // we are dealing with a simple array of numbers. No extraction needed.
      extracted_data = our_data;
    } else {
      console.log('TypeError: expected an array of numbers, found ' + typeof(our_data[0]));
      return;
    }

    var hist = d3.histogram();
    if (args.bins) {
      hist.thresholds(args.bins);
    }

    var bins = hist(extracted_data);
    args.processed_data = bins.map(function(d) {
      return { 'x': d.x0, 'y': d.length };
    });
  } else {
    // here, we just need to reconstruct the array of objects
    // take the x accessor and y accessor.
    // pull the data as x and y. y is count.

    args.processed_data = our_data.map(function(d) {
      return { 'x': d[args.x_accessor], 'y': d[args.y_accessor] };
    });

    var this_pt;
    var next_pt;

    // we still need to compute the dx component for each data point
    for (var i = 0; i < args.processed_data.length; i++) {
      this_pt = args.processed_data[i];
      if (i === args.processed_data.length - 1) {
        this_pt.dx = args.processed_data[i - 1].dx;
      } else {
        next_pt = args.processed_data[i + 1];
        this_pt.dx = next_pt.x - this_pt.x;
      }
    }
  }

  // capture the original data and accessors before replacing args.data
  if (!args.processed) {
    args.processed = {};
  }
  args.processed.original_data = args.data;
  args.processed.original_x_accessor = args.x_accessor;
  args.processed.original_y_accessor = args.y_accessor;

  args.data = [args.processed_data];
  args.x_accessor = args.processed_x_accessor;
  args.y_accessor = args.processed_y_accessor;

  return this;
}

MG.process_histogram = process_histogram;

// for use with bar charts, etc.
function process_categorical_variables(args) {
  'use strict';

  var extracted_data, processed_data = {},
    pd = [];
  //var our_data = args.data[0];
  var label_accessor = args.bar_orientation === 'vertical' ? args.x_accessor : args.y_accessor;
  var data_accessor = args.bar_orientation === 'vertical' ? args.y_accessor : args.x_accessor;

  return this;
}

MG.process_categorical_variables = process_categorical_variables;

function process_point(args) {
  'use strict';

  var data = args.data[0];
  var x = data.map(function(d) {
    return d[args.x_accessor];
  });
  var y = data.map(function(d) {
    return d[args.y_accessor];
  });

  if (args.least_squares) {
    args.ls_line = least_squares(x, y);
  }

  return this;
}

MG.process_point = process_point;