About the Chaco Scales package

In the summer of 2007, I spent a few weeks working through the axis ticking and labelling problem. The basic goal was that I wanted to create a flexible ticking system that would produce nicely-spaced axis labels for arbitrary sets of labels and arbitrary intervals. The chaco2.scales package is the result of this effort. It is an entirely standalone package that does not import from any other Enthought package (not even traits!), and the idea was that it could be used in other plotting packages as well.

The overall idea is that you create a ScaleSystem consisting of various Scales. When the ScaleSystem is presented with a data range (low,high) and a screen space amount, it searches through its list of scales for the scale that produces the “nicest” set of labels. It takes into account whitespace, the formatted size of labels produced by each scale in the ScaleSystem, etc. So, the basic numerical Scales defined in scales.py are:

• FixedScale: Simple scale with a fixed interval; places ticks at multiples of the resolution
• DefaultScale: Scale that tries to place ticks at 1,2,5, and 10 so that ticks don’t “pop” or suddenly jump when the resolution changes (when zooming)
• LogScale: Dynamic scale that only produces ticks and labels that work well when doing logarithmic plots

By comparison, the default ticking logic in DefaultTickGenerator (in ticks.py) is basically just the DefaultScale. (This is currently the default tick generator used by PlotAxis.)

In time_scale.py, I define an additional scale, the TimeScale. TimeScale not only handles time-oriented data using units of uniform interval (microseconds up to days and weeks), it also handles non- uniform calendar units like “day of the month” and “month of the year”. So, you can tell Chaco to generate ticks on the 1st of every month, and it will give you non-uniformly spaced tick and grid lines.

The scale system mechanism is configurable, so although all of the examples use the CalendarScaleSystem, you don’t have to use it. In fact, if you look at CalendarScaleSystem.__init__, it just initializes its list of scales with HMSScales + MDYScales:

HMSScales = [TimeScale(microseconds=1), TimeScale(milliseconds=1)] + \
           [TimeScale(seconds=dt) for dt in (1, 5, 15, 30)] + \
           [TimeScale(minutes=dt) for dt in (1, 5, 15, 30)] + \
           [TimeScale(hours=dt) for dt in (1, 2, 3, 4, 6, 12, 24)]

MDYScales = [TimeScale(day_of_month=range(1,31,3)),
            TimeScale(day_of_month=(1,8,15,22)),
            TimeScale(day_of_month=(1,15)),
            TimeScale(month_of_year=range(1,13)),
            TimeScale(month_of_year=range(1,13,3)),
            TimeScale(month_of_year=(1,7)),
            TimeScale(month_of_year=(1,))]

So, if you wanted to create your own ScaleSystem with days, weeks, and whatnot, you could do:

ExtendedScales = HSMScales + [TimeScale(days=n) for n in (1,7,14,28)]
MyScaleSystem = CalendarScaleSystem(*ExtendedScales)

To use the Scales package in your Chaco plots, just import PlotAxis from chaco2.scales_axis instead of chaco2.axis. You will still need to create a ScalesTickGenerator and pass it in. The financial_plot_dates.py demo is a good example of how to do this.