[BUG] the consolidation is not aggregating correctly

[npazosmendez]

The consolidation done when marshalling to JSON does not match what GraphiteWeb does, see their tests:

• What CarbonAPI does:

  carbonapi/expr/types/types.go

  Lines 113 to 139 in 70c5ee5

  	t := r.StartTime * timestampMultiplier
  	for _, v := range r.AggregatedValues() {
  	if noNullPoints && math.IsNaN(v) {
  	t += r.AggregatedTimeStep() * timestampMultiplier
  	} else {
  	if innerComma {
  	b = append(b, ',')
  	}
  	innerComma = true
  	b = append(b, '[')
  	if math.IsNaN(v) || math.IsInf(v, 1) || math.IsInf(v, -1) {
  	b = append(b, "null"...)
  	} else {
  	b = strconv.AppendFloat(b, v, 'f', -1, 64)
  	}
  	b = append(b, ',')
  	b = strconv.AppendInt(b, t, 10)
  	b = append(b, ']')
  	t += r.AggregatedTimeStep() * timestampMultiplier
  	}
  	}

• How GraphiteWeb tests it: https://github.com/graphite-project/graphite-web/blob/ed08fa4e2f6526ae11e55f88f70d0b6d36e4052f/webapp/tests/test_render.py#L536

I haven't studied both implementations in detail, but a basic difference I see is the fact that GraphiteWeb may change the StartTime, whereas CarbonAPI always uses the original value. This test from GraphiteWeb shows such a case:

• Here a different StartTime (4 instead of 1) is expected https://github.com/graphite-project/graphite-web/blob/ed08fa4e2f6526ae11e55f88f70d0b6d36e4052f/webapp/tests/test_render.py#L559

[npazosmendez]

Fixed by #146