Correct issues with nose dive data, NaN data on stacked graphs

For a metric with σ = 0, it would appear as having no y-axis. This is corrected.
For a metric with σ = 0 and x̄ = 0, it would appear as having no value nor axis.
This is corrected.

Logic regarding how to handle a lack of data in descartes dataset padding is
corrected, and handled in rickshaw: the nil value is used in the controller,
which is explicitly handled in rickshaw to cast back to a NaN

app/assets/javascripts/graphs/stacked.js

@@ -62,9 +62,9 @@ function renderStacked(data) {
 			}
 		} 
 		if (isRight(n)) { 
-			right_range = [ Math.min(min, right_range[0]) , Math.max(max, right_range[1]) ];
+			right_range = [ Math.min(min, right_range[0]) - 1 , Math.max(max, right_range[1]) + 1 ];
 		} else { 
-			left_range = [ Math.min(min, left_range[0]) , Math.max(max, left_range[1]) ];
+			left_range = [ Math.min(min, left_range[0]) - 1, Math.max(max, left_range[1]) + 1 ];
 		} 
 	}
 

lib/backend/descartes.rb

@@ -76,7 +76,7 @@ def get_metric m, start, stop, step
 			if points.length == 1 then
 				padded << points.first
 			elsif points.length == 0 then
-				padded << {x: i, y: (0.0/0.0)}
+				padded << {x: i, y: nil} # (0.0/0.0)}
 			end
 		end
 

vendor/assets/javascripts/rickshaw.js

@@ -550,7 +550,9 @@ Rickshaw.Graph = function(args) {
 				var seriesData = data[index];
 				if(seriesData) {
 					seriesData.forEach( function(d) {
-						d.y = series.scale(d.y);
+						// Want to handle NaN as no points - JSON won't accept these as NaN, so use null instead, and put back here.
+						if (d.y === null) { d.y = NaN; } else { 
+						d.y = series.scale(d.y); }
 					} );
 				}
 			}