1
******************************
2
About the Chaco Scales package
3
******************************
5
In the summer of 2007, I spent a few weeks working through the axis
6
ticking and labelling problem. The basic goal was that I wanted to
7
create a flexible ticking system that would produce nicely-spaced axis
8
labels for arbitrary sets of labels *and* arbitrary intervals. The
9
chaco2.scales package is the result of this effort. It is an entirely
10
standalone package that does not import from any other Enthought
11
package (not even traits!), and the idea was that it could be used in
12
other plotting packages as well.
14
The overall idea is that you create a ScaleSystem consisting of
15
various Scales. When the ScaleSystem is presented with a data range
16
(low,high) and a screen space amount, it searches through its list of
17
scales for the scale that produces the "nicest" set of labels. It
18
takes into account whitespace, the formatted size of labels produced
19
by each scale in the ScaleSystem, etc. So, the basic numerical Scales
20
defined in scales.py are:
22
* FixedScale: Simple scale with a fixed interval; places ticks at
23
multiples of the resolution
24
* DefaultScale: Scale that tries to place ticks at 1,2,5, and 10 so that
25
ticks don't "pop" or suddenly jump when the resolution changes (when
27
* LogScale: Dynamic scale that only produces ticks and labels that work
28
well when doing logarithmic plots
30
By comparison, the default ticking logic in DefaultTickGenerator (in
31
ticks.py) is basically just the DefaultScale. (This is currently the
32
default tick generator used by PlotAxis.)
34
In time_scale.py, I define an additional scale, the TimeScale.
35
TimeScale not only handles time-oriented data using units of uniform
36
interval (microseconds up to days and weeks), it also handles non-
37
uniform calendar units like "day of the month" and "month of the
38
year". So, you can tell Chaco to generate ticks on the 1st of every
39
month, and it will give you non-uniformly spaced tick and grid lines.
41
The scale system mechanism is configurable, so although all of the
42
examples use the CalendarScaleSystem, you don't have to use it. In
43
fact, if you look at CalendarScaleSystem.__init__, it just initializes
44
its list of scales with ``HMSScales + MDYScales``::
46
HMSScales = [TimeScale(microseconds=1), TimeScale(milliseconds=1)] + \
47
[TimeScale(seconds=dt) for dt in (1, 5, 15, 30)] + \
48
[TimeScale(minutes=dt) for dt in (1, 5, 15, 30)] + \
49
[TimeScale(hours=dt) for dt in (1, 2, 3, 4, 6, 12, 24)]
51
MDYScales = [TimeScale(day_of_month=range(1,31,3)),
52
TimeScale(day_of_month=(1,8,15,22)),
53
TimeScale(day_of_month=(1,15)),
54
TimeScale(month_of_year=range(1,13)),
55
TimeScale(month_of_year=range(1,13,3)),
56
TimeScale(month_of_year=(1,7)),
57
TimeScale(month_of_year=(1,))]
59
So, if you wanted to create your own ScaleSystem with days, weeks, and
60
whatnot, you could do::
62
ExtendedScales = HSMScales + [TimeScale(days=n) for n in (1,7,14,28)]
63
MyScaleSystem = CalendarScaleSystem(*ExtendedScales)
65
To use the Scales package in your Chaco plots, just import :class:`PlotAxis` from
66
:mod:`chaco2.scales_axis` instead of :mod:`chaco2.axis`. You will still need to create a
67
:class:`ScalesTickGenerator` and pass it in. The financial_plot_dates.py demo is a
68
good example of how to do this.