Population not detected error

[EricAComstock]

Hello All,

I've been getting this error when trying to use the gridspec_plot.py example to analyse some Vlasiator results, where the proton population is tried but is not detected.

My simulation has a different form of ion, which I originally thought might be the culprit, but even after renaming them to "proton," this error still shows up. Please advise.

INFO: Using LaTeX formatting
INFO: Using matplotlib version 3.6.3
INFO: Mesh is not 2-D: Use plot_colormap3Dslice instead!
INFO: V-space reduction via integration
Traceback (most recent call last):
  File "/home/LGST/Vlasiator/install_test_1/vlasiator/samples/IPShock/gridspec_plot.py", line 79, in <module>
    pt.plot.plot_vdf(vlsvobj=vf,cellids=[cid],axes=axes,center='bulk', scale=scale,
  File "/home/LGST/analysator/analysator/plot/plot_vdf.py", line 641, in plot_vdf
    raise ValueError("Unable to detect population "+pop+" in .vlsv file!")
ValueError: Unable to detect population proton in .vlsv file!

I'm using the command python3 gridspec_plot.py to run it.

Sincerely,
Eric Comstock

[alhom]

Hi,

You can pass the kwarg pop="ion" to plot_vdf to not have to rename your populations.

You also have a warning to substitute plot_colormap3dslice for plot_colormap, since the latter is a specific (and historical) function for plotting 2D runs.

However, this might be an issue with your output files not containing VDF data. You can inspect the output variable vg_f_saved for cells that have a VDF stored.

[alhom]

Looking at the outputs I I got from the still floating-point-error-y config, you should be able to see stored VDFs in the bulk file (although for the broken example, these were still empty distributions). See the system_write_distribution_stride parameters in the config to control where VDFs are saved in bulk files, btw. plot_vdf works on that file with pop=ion keyword, finding the population, so I suggest trying again with the initial ion config and pop="ion" kwargs.

Further, I suggest using the following function to find CellIDs that contain VDFs, since the example CellID array will not work here, but this is a useful helper for finding VDFs close to given coordinates to inspect.
https://fmihpc.github.io/analysator/vlsvfile/#analysator.vlsvfile.VlsvReader.get_cellid_with_vdf

[EricAComstock]

Hello Alhom,

Thank you! I tried both renaming and using the kwarg suggested, but it still threw the same error. How do I access vg_f_saved? I could not find it in the file or the documentation.

Are there any documentation files explaining the access modes in analysator's examples and what they do? This is my first time using it.

Sincerely,
Eric Comstock

[alhom]

vg_f_saved is a variable that you can choose to output in the simulation config (under [variables]). f = analysator.vlsvfile.VlsvReader(filename) is the main object for reading the data directly (via f.read_variable("vg_f_saved")), while analysator.plot.plot_colormap3dslice etc. will let you directly plot the variable from the dataset.

Here is some introductory material and examples for using Analysator: https://fmihpc.github.io/analysator/analysator/

[EricAComstock]

Hello Alhom,

Thank you!

For vg_f_saved, I got a really long vector of ones. Does that just mean that velocity was saved everywhere, or it that unexpected?

I also tried just using the plot_colormap portion by itself, but it just gave a white square with no data (though it did not throw an error). Do you know what might be causing this?

Sincerely,
Eric Comstock

[alhom]

Hi,

plot_colormap is defined for 2D datasets only, your config is a 3D dataset. You should use plot_colormap3dslice instead.

If all of your vg_f_saved values were one, then indeed you have velocity distributions (note, see ion/vg_v or ion/vg_rho for the bulk velocity moments, those are stored everywhere) stored at all cells. This is then rather strange if plot_vdf is complaining about not finding the population at all, though.

I would anyway suggest starting by using the plotting (or reading) commands in isolation to get started, along with the tutorial linked above. Try e.g. plotting the ion bulk values with plot_colormap3dslice(var="ion/vg_rho")

[EricAComstock]

Hello Alhom,

Thank you!

I'm now getting something that loos like a graph, but it just has a featureless green stripe on the side, and white for the remaining 90% of the plot. Also, is there a way to change the units to something other than R_E (which is inconvenient for my case)?

I'm using pt.plot.plot_colormap3dslice(filename=bulkfile, var='ion/vg_rho', draw=1, normal='y') to plot it

Sincerely,
Eric Comstock

[alhom]

Hi,

That sounds like it could be what the initial state of an IPShock simulation looks like - just the shock ramp. Letting it develop further should start producing some dynamics.

See the reference for available plotting options, kwarg axisunit should do what you wish:
https://fmihpc.github.io/analysator/plot/#analysator.plot.plot_colormap3dslice

[EricAComstock]

Hello Alhom,

Thank you for the recommendation for the kwarg!

May I ask what you mean be letting it develop further? I did not halt the file's execution.

Sincerely,
Eric Comstock

[alhom]

By that I mean letting the simulation run longer - but here I am indeed assuming that your outputs were still near the initialization phase. You can, of course, paste your output plot here for reference. Anyway, I would suggest inspecting the simulation outputs frame-by-frame, and trying to see how/if the simulation develops in time as expected.

Best,
Markku Alho

[EricAComstock]

Hello Markku Alho,

Ah, I see. That makes sense. Thank you!

Sincerely,
Eric Comstock

[EricAComstock]

Hello Markku Alho,

It looks like, no matter what I do, the code only plots the data from the first timestep. Is there any way to set it to any other, or to the last timestep instead?

Sincerely,
Eric Comstock

[alhom]

You need to specify, for the plotting commands, either the filename of the output you want to plot, or a corresponding VlsvReader object. Each output file contains only the data from that particular snapshot, so if you only have only one output file, please adjust the output cadence to better cover the simulation period.

[EricAComstock]

Hello Markku Alho,

Thank you! Which commend do I need to change in the .cfg file? I tried changing system_write_t_interval but it did not seem to work, and changing diagnostic_write_interval caused it to crash.

Sincerely,
Eric Comstock