[Plot] Logarithmic axis, log plots

[tadmorgan]

Is it possible to set a plot axis to logarithmic scale?

[trusktr]

Not yet. In progress in https://github.com/nasa/openmct/pull/4710

[trusktr]

We have an initial version in the log-plots and experimental-plotly-log-plots branches. But some more work needs to be done. A few things we are thinking about:

• [ ] log ticks for Y axis: we need to provide meaningful numbers as the ticks for the Y axis in such a way that they will apear unevenly spaced in the view (due to log plot compressing space as we move upward)
• [ ] what if data has values smaller than 1 (f.e. what if most of the data is below 1 with outliers above 1)? We need a way to multiply the log formula by a fraction (f.e. 0.1) to make it apply to values greater than 0.1 instead of greater than 1, for example.
• [x] what about negative values? we would need to mirror the log formula around its Y axis and/or X axis.
  • cases:
  • positive values above 1: return log(point.y)
  • negative values with abs(value) bigger than 1: if (point.y < 0) return -log(-point.y)
  • abs(value) is smaller than 1or other sorts of ranges: there are different ways we can adjust the log formula to achieve a similar smooshing effect in different ranges:
    • for positive values, return log(point.y + (1 - n)) where n is the point at which the log formula graph (not our log plot) crosses the X axis. A value of 0.1 which means the log formula graph will be shifted left so that the intersection point is 0.1. This allows our log plot (not log formula graph) to be useful for values above 0.1 instead of above 1.
    • for negative values, similar as before with the added 1 - n, if (point.y < 0) return -log(-point.y + (1 - n))
    • we might want to shift it a different way, f.e. return log(point.y) + n
    • Examples of shifted graphs for both positive and negative values: https://www.desmos.com/calculator/bhqwpll6vb
    • I think there are other ways to modify the log formula, like elongating it, that may be useful for some cases.
• small values: we may need to multiply the above log formulas by a scalar in order to vertically stretch the log formula graph (not log plot) such that the amount of compression in the log plot (not log formula graph) increases more, or less, as we go upward in the log plot (not formula graph) depending on the multiplier.
• UI:
  • [x] user should be able to toggle between log plot and non-log plot.
  • [ ] user should to be able to change the n value, constrained to be 0 or higher, using a slider input
  • [ ] user should be able to change a multiplier to compress or uncompress the log plot vertically.

[trusktr]

Updated playground. All plots toggled off except symmetrical log functions: https://www.desmos.com/calculator/htkxrlto3a

[trusktr]

A better one showing symlog, scaled symlog, and scaled antisymlog, with a slider to adjust the scale value: https://www.desmos.com/calculator/2pcxvmsd6y

[charlesh88]

Good chat with @trusktr today:

• [x] UI should add a "Use log scale" checkbox to the Inspector "Y Axis" UI section when editing.
• [ ] Joe will look into also including a range slider that allows the user to control the log scaling directly, which only appears when the "use" checkbox is checked. For performance, this slider will probably only update the plot on mouseup, but might include a fully dynamic numeric callout that tells the user what the current setting is. This is a stretch goal.
• @charlesh88 will do any front-end sanding and shimming if needed.

[trusktr]

4931 is in. When editing an Overlay Plot, there is now a "Log mode" checkbox that can be enabled in the right sidebar.

Testing:

• make overlay plot
• add sinewave generator (or any other data)
• edit overlay and turn on log mode
• toggle log mode on/off a few times and the plot drawing should visibly toggle between both modes
• save overlay plot in log mode
• refresh the browser and the overlay should still be in log mode and render correctly
• edit overlay, try turning autoscale on/off and ensure that log mode still works
• try zooming in out and panning with permutations of any of the above actions and make sure things work

[khalidadil]

I'm seeing an issue in an overlay plot when using a sine wave generator with log mode turned on. If I turn on log mode, save, and refresh, the plot isn't displaying correctly.

I'm seeing this error:

[akhenry]

We could be doing a better job of formatting y-axis values in log mode. Instead of showing something like this:

We should be rounding the y-axis values to the nearest whole number in this case.

[akhenry]

@khalidadil Sounds like maybe a version of this issue - https://github.com/nasa/openmct/issues/5039?

I am also seeing this issue for what it's worth, with non-SWG telemetry points as well.

[khalidadil]

@akhenry I just checked, and autoscale was turned on for me.

[charlesh88]

Verified not fixed Testathon 04-25-22:

1. Log scale should not change tick values from fairly compact integers to long floats: Log scale off

Log scale on

2. Also seeing the same problem seen by others in #5039.

[akhenry]

@shefalijoshi to investigate https://github.com/nasa/openmct/issues/2297#issuecomment-1109075202

[shefalijoshi]

I think maybe we could use toFixed(2) instead of whole numbers to solve the long ticks issue? Otherwise if the range of values is small (0-1 for instance) then all the ticks would be either all 0 or all 1 which would be odd.

[akhenry]

Verified fixed, y-axis labels look awesome now!

[charlesh88]

NOT fixed Testathon 4-28-22. Ticks end up being duplicated when the plotted value is < 1.

After no incoming data for a period:

[akhenry]

Testing Instructions

This intentionally restores the behavior seen in https://github.com/nasa/openmct/issues/2297#issuecomment-1109114296 which will need to be fixed in a separate issue, but does not block release.

1. Create an overlay plot
2. Enable log scale
3. Verify that the scale is not causing a loss of precision resulting in repeating scale values as seen in https://github.com/nasa/openmct/issues/2297#issuecomment-1112724445

Important note This cannot be verified with sine wave generators. SWGs have a custom formatter which will round all values to 2 decimal places. In this case the precision is already lost before the log plots format the values.

[michaelrogers]

This change appears to be fixed in testathon 05/02. On inspecting an overlay plot with log scale enabled, the scale values do not repeat and instead show values without clipping the # of decimal places (increasing precision).

[akhenry]

Verified fixed.

[rukmini-bose]

Confirmed Fixed 05/02. Seeing the same behavior as Michael.