usage of plot_filter_function(t,x,y,z,ux,uy,uz)

[hightower8083]

dear devs

I've a problem with the plot_filter_function(t,x,y,z,ux,uy,uz) in a 3D case (multi-GPU) with plotfile diagnostics backend.
By no means I could not make it work for a simple filter (uz>1.5) -- it writes no particles at all..
Here is a piece of the input file

diagnostics.diags_names = diag1
diag1.diag_type = Full
diag1.format = plotfile
diag1.period = 4000
diag1.fields_to_plot = Ex Ez rho
diag1.species = electrons
diag1.electrons.variables = ux uy uz w
diag1.electrons.plot_filter_function(t,x,y,z,ux,uy,uz) = (uz>1.5)

Just in case, i've also tried SI units instead of normalized, i.e. replace (uz>1.5) with (uz>4.0963867961067354e-22), but still no particles were written..

Can someone provide a working case for debugging..?

[ax3l]

Hm, I thought we had a test on this but I cannot spot it.
Generally, try to escape the function like this:

diag1.electrons.plot_filter_function(t,x,y,z,ux,uy,uz) = "(uz>1.5)"

uz should be in beta*gamma and need to validate, maybe still in SI. I think to remember it works last time I used it - nonetheless, we updated this functionality recently #1384 and it's lacking a test...

[atmyers]

The issue seems to a units bug in WarpX. What we actually store for the particle momenta is gamma*v. When we write out particles for plotfiles, we convert that to SI units. The filter function gets applied after this conversion; however, for some reason it includes another factor of (1/c), as if we were attempting to convert from gamma*v to normalized units. So the quantity that actually gets passed in to the filter function is mass*gamma*v/c.

[atmyers]

I have attempted to fix this in PR #1598

[ax3l]

@hightower8083 when you get to it, can you please confirm the latest development, which now includes #1598, fixes your reported issue?

[ax3l]

I did some independent tests on the same filter logic on openPMD files in #1632 and can confirm that this works (I filtered electrons over a certain non-zero gamma locally to see if I can spot heated electrons in the right energy range - worked well!)

Please don't hesitate to perform your own tests nonetheless.