hightower8083
commented
3 weeks ago Hi @ronandoherty1
Thank you a lot for catching this issue!
Indeed z_is_xi option has not been thoroughly tested, in particular, I've never used it with the discontinuous tracks since they are more commonly specific to the time-domain PIC, and this feature addresses the case of QSA simulations (notably HiPACE), where z coordinate is a QSA variable and doesn't reflect the propagation, and in HiPACE, the beam particles usually do not appear/disappear.
But you are right and indeed when we split the track the time array is not splitted accordingly, so this should be fixed.
However, I think it is possible make less modifications and to avoid introducing this new variable good_inds. In fact each track already has the it_start (last value in the track container)., that defines its starting position with respect to the global time array. Therefore, it_start and the track length (or fragment) should be enough to get the part of t that corresponds to this particular track. For example, if here we replace t with t[it_start:it_start+z.size], it should also fix the problem.
Can you check with your data if this single-line fix works correctly, and let me know? And of course you are welcome to submit the PR so it'll be easier to follow and finalize the fix
ronandoherty1
Hi there, hope you are well
I have been using the tracksfromOPMD function, and have been noticing that when I pass the "z_is_xi=True" argument, I sometimes get this error: "File "/home/dohertyr/.conda/envs/ff_betatron_sims/lib/python3.12/site-packages/synchrad-0.0.3-py3.12.egg/synchrad/converters.py", line 128, in tracksFromOPMD f[f'tracks/{i_tr:d}/z'] = z + c * t
^~~~~ ValueError: operands could not be broadcast together with shapes (29,) (33,) "I realised that this error stems from the fact that, while the position and momentum variables are truncated according to their validity, the t values are not truncated. Thus, if there are any invalid values of the positions or momenta, then we get this mismatching of dimension.
I wrote up a fix to this which is here
https://github.com/ronandoherty1/synchrad/blob/dev/synchrad/converters.py
The idea is just that the function "split_tracks_by_nans" now outputs a list of "good_inds" as well, ie the indices for which the position and momenta are valid. Then the t array is truncated according to these indices, so dimension mismatching between x and z no longer occurs.
I hope this is helpful in improving the software. I would be happy to submit a pull request if you think this is a good fix.