Graphics the hard way: Time Series in Avenue

Abstract

1. Introduction

In 1995 VKI Water Quality Institute undertook in collaboration with Danish Hydraulic Institute (DHI) an assignment to build an environmental information system for Oresundskonsortiet (OSK), who is responsible for the building of a bridge/tunnel project between Denmark and Sweden. This system EAGLE /1/ would heavily rely on time series plots to visualise trends in data such as eelgrass biomass, mussel biomass, sediment spill and the like. To this end, it was first attempted to develop a solution using a DDE connection to an external graphics program. However it soon became clear that a solution using the built-in graphics capabilities of ArcView was both possible and feasible. This paper deals with the development and application of the scripts used to view time series plots in the EAGLE system.

2. Plotting basics

2.1 Display setup

aLine = Line.Make(0.1@0.1,0.5@0.5)

gLine = GraphicShape.Make(aLine)

theGraphics.AddBatch(gLine)



figure 2: The process of adding a line to the view.

2.2 Coordinate system

By defining these constants it is easy to change the size of the plot. When we found that in a special situation it was necessary to display three time series independently, we just put up three views with a plot in each of them, and scaled the views to fit the screen. The graphs scaled according to the view size.

Width = theExtent.GetWidth

Height = theExtent.GetHeight

OriginX = Width * 0.1

OriginY = Height * 0.2

XaxisLength = Width * 0.8

YAxisLength = Height * 0.6

TickLength = XaxisLength * 0.02

MarkerSize = XaxisLength * 0.015



Figure 3: Defining "anchor points" and dimensions

2.3. Defining the interface

It was decided that these options should be passed as parameters to a central script, rather than utilizing global variables for this. In this 20-40 parameters are passed to the script, depending on how many data series there are on the plot. This makes calling TS.Graph rather complicated, but the script is meant for in-house developers, not for end-users, so clarity is less important.

3. Drawing the Time Series

3.1 Axis scaling

Scaling of the date values on the x-axis is done by the script TS.ScaleTim. Usually when scaling, it is possible to determine a start value, an end value and a step value. On a time series axis however, there is no such thing as a step value. If for instance the aim is to display a tick and a label the first day in every month, the "step" would have values such as 31, 28, 31, 30. So instead of a step the script must find "nice" dates to return. These dates will depend heavily on the time range. If the time span of the plot is a week, a tick will be shown for every day. If time span is a year, once a month or every second month is more appropriate. Therefore a scheme has been setup to let "custom logic" decide the plotting. We decided that certain ranges would return certain dates as tick marks. The ranges were tested for usability, and adjusted accordingly. This process continued through the beta testing of EAGLE.

On the Y-axis we used some code, that we have used in several other applications. The general idea, which goes back at least 15 years, is to divide the minimum and maximum values by 10 till they are in the range of 0 to 9, then set predefined values for minimum, maximum and step value and multiply them by 10 till they are back in the previous range. It was fairly simple to translate the Pascal equivalent into Avenue. The only problem was we did not know how to raise a number to a power. Not until one of the final modifications, did we learn of the ^-request

As it turned out all logarithmic axes would need to be scaled from 1 to 1000. Hence the script TS.ScaleLog returns a hard coded list of tick marks and labels. If necessary a slight modification to the TS.ScaleVal script could be made, so that it could return logarithmic values.

3.2 Drawing the Curve

4. Optimising for better performance

Next all loops were examined closely. It turned out that calling another script using av.Run was quite time consuming. So small scripts that were there for the clarity of the code, were sacrificed to optimising the loops. Another concern was that of repetitive calculation of values. Being used to programming in Pascal, were the compiler optimises repetitive expressions, I was not aware of this in the beginning. Avenue is not a compiled language, so it turned out that there was quite a difference, whether constant values were calculated before the loop, or within the loop. After these optimisations the script was able to plot 7000 points in 45 seconds, which is still a long time, but acceptable. Most time series still have less than 50 points.

5. Application

After conclusion of the time series plot script, it became clear that ArcView was not able to display profile plots. Profile plots are used to visualise a parameter through the water column, hence depth is the Y-axis and the parameter is displayed along the X-axis. They are special in the way, that there can be several Y-values to each X-value. When we found out that ArcView's built-in charting capabilities would draw lines between the points in profile plots, TS.Graph was adapted to produce profile plots as well.

Figure 4:

6. Conclusions

7. References