This is an implementation of a "live data plotter." Think of the strip charts that are generated by a seismometer. Or those generated by the lie detectors that you see in the movies.
This is like those.
An external program generates/provides the time-series data values to
be plotted with live_plotter. The measurement values are either
"none" or a real numbers, and all the time-series values must be
present, separated by a space, on a single line for each time step.
Labels should be specified for each of the time-series data, and the
number of labels must match the number of numbers per line of input.
If the --timestamp command line argument is specified, then the
number of numbers should be one greater, with the first number
representing the timestamp value in seconds since the Unix epoch.
Unlike real strip charts with paper tape made for Turing machines, the
display area is finite, and the --viewport-width command line
argument specifies how many most-recent time values to show in the
default view. The --max-points parameter specifies how many data
points to retain per time-series data stream, and you can click-drag
to scroll the viewport to see older data. The GUI also allows zooming
into a smaller region, so it would be feasible to specify a huge
number of sample times and just zoom into the most recent. After
zooming, double clicking resets the view to show everything. A reset view button resets the display to the default viewport-width wide
view. Hovering over a data point shows the data values and the name
of the time-series to which it belongs.
A settings panel is available on the right. You can collapse it to reclaim screen real estate for the data chart. The top subpanel in the settings is a data smoothing subpanel. The default is no data smoothing. The buttons and slider allow you to set the τ value in the exponential smoothing filter. There is a command line argument to control the highest possible τ value on the slider.
See https://en.wikipedia.org/wiki/Exponential_smoothing for more info.
Below the data smoothing subpanel is the visibility subpanel. Using it, you can turn individual time-series lines off (invisible) and on to get rid of or at least attempt to control the clutter that might be hiding the particular time-series data that you're zooming in on / exploring.
$ span_circuit_info -a --abs -k instantPowerW --live 5 --timestamp | eval live_plotter --timestamp --labels $(span_circuit_info -a -k name -q)
See https://github.com/bennetyee/span_circuit_info for span_circuit_info.
The exponential smoothing for the data "works" when there's missing data points, but the display is not smoothed (data points are connected by straight lines). This is probably not the right thing to do.
There appears to be an issue with one or more of eframe, egui, and
egui_plot where even though no data is updating and you are not
"actively" interacting with the plot by clicking and dragging the
chart, you can get the program to spin and chew up most of a CPU core
by just wiggling the mouse over the chart area. As of this writing,
AI Studio doesn't seem to understand how to use a more recent version
of these graphical UI packages than 0.29 (the API isn't stable), and I
haven't tried to figure out of the latest 0.34 version of these
packages still have this performance issue.
The code in this repo is entirely vibe coded using
https://aistudio.google.com/ and https://claude.ai. The only manual thing
done other than downloading the AI generated code into the repository
is occasionally remembering to run rustfmt and reviewing the code,
mostly using git diff.