Add a color setting for missing data

[Robert Leach]

Currently, missing data elements (e.g. no value, "NaN", random text, missing columns, etc.) are replaced with 0s. Change the tool to store them as null (and handle all the places that that breaks) and then create a color setting for what color missing data should be (defaulted to the gray of the background window color).

[Christopher Keil]

I have looked into this. The ColorSets/-Presets/-Extractor should still contain default values for NODATA or EMPTY. Alok has also used a double array for the matrix data, and many third party libraries do as well.

Missing values will first be parsed as 'null' and then set to DataModel.NODATA or DataModel.EMPTY static doubles from the DataModel interface. Those are just really low double values like -1000000 and -2000000. If any value in the data equals any of these two variables, the specific color will be chosen instead of the color normally specified for 0.0 or either of two limits of the color gradient (red or green, for example).

Clustering with missing data is a real big issue that has prompted people to write entire papers. I am going to research this more.

[Christopher Keil]

A button has been added to the color settings page which allows to set the specific desired color for data values recognized as 'missing'. The required code for the Preferences API is also working well.