FieldData, CellData, and PointData are all objects that allow you to store data along with the geometry/topology of your data set. You can store things like colors, velocity vectors, ids, names, or any other data that is associated either with the data set as a whole or with each point/cell in the data set. In all three of these cases, you must create an array (vtk*Array, e.g. vtkUnsignedCharArray) and then add it to the data with:
Arrays attached to the FieldData of a dataset describe global properties of the data. That is, if you want to save the time at which the data were recorded, you would put that value in the FieldData. If you wanted to name the data set, you would also put that in the FieldData. There are no restrictions about the length of arrays that are added to the FieldData.
If you have a dataset which has values/data at every cell (e.g. every triangle in a mesh has a specified color), you should add these data to the CellData of the dataset. The length of arrays added to CellData must equal the number of cells in the dataset (dataset->GetNumberOfCells()).
If you have a dataset which has values/data at every point, you'll want to use PointData.
You can add as many arrays as you would like to all of the *Data. A reasonable thing to want to do is switch between different data for your display. That is, if you have two sets of colors stored in the PointData, you should be able to choose which one to display at the moment.
This concept is called ActiveScalars and ActiveVectors. Some algorithms
vtkUnsignedCharArray* colors1 = vtkUnsignedCharArray::New(); colors1->SetName("colors1"); // ... fill the colors array
vtkUnsignedCharArray* colors2 = vtkUnsignedCharArray::New(); colors2->SetName("colors2"); // ... fill the colors array
Now here is the key. To select which one is used, use this: