IndexError when calling emg3d.construct_mesh

[emsig]

IndexError when calling emg3d.construct_mesh

[prisae]

Thanks for reporting @jingxu94 . The entire construct_mesh()-functionality is very new, so unfortunately it is possible that there are bugs, even though I tried to avoid them ;-)

As I explained in our earlier discussion, I can currently not reproduce the problem, so I cannot do debugging myself. I will therefore ask some questions, it would be great if you could help me with that. Hopefully we will get to the bottom of it.

Could you try to run the following and post the output?

import emg3d
z0, hz = emg3d.meshes.get_origin_widths(
    frequency=1.0,
    min_width_limits=100.0,
    properties=[0.3, 2, 100],
    center=-1000,
    seasurface=0.0,
    domain=[-2500, 0],
    verb=2,
)

Just to ensure if the issue is in construct_mesh or in get_origin_widths.

[jingxu94]

Similar error as follows:

Traceback (most recent call last):
  File "c:/Users/JingXu/Desktop/code/start_emg3d/model_property_mapping/test_issue.py", line 9, in <module>
    verb=2,
  File "D:\Anaconda3\envs\tdem\lib\site-packages\emg3d\meshes.py", line 626, in get_origin_widths
    thxr = hx[-1]*sa**np.arange(nx)
IndexError: index -1 is out of bounds for axis 0 with size 0

[prisae]

That is really strange, I get the following without any issues:

Skin depth          [m] : 276 / 712 / 5033  [corresponding to `properties`]
Survey domain DS    [m] : -2500 - 0
Comp. domain DC     [m] : -6974 - 31623
Final extent        [m] : -8729 - 32628
Cell widths         [m] : 100 / 100 / 9085  [min(DS) / max(DS) / max(DC)]
Number of cells         : 48 (25 / 23 / 0)  [Total (DS/DC/remain)]
Max stretching          : 1.000 (1.000) / 1.380  [DS (seasurface) / DC]

I even recreated an environment fixing all the package versions to match yours, but I do not encounter the problem. I also tried it with pip and with conda, both work fine. The only difference that remains between your system and mine is Windows/Linux.

I will try and get hold of a Windows machine to test it.

[prisae]

OK, I found the problem, it is a Windows issue. Int's default to 64 on Linux/Mac, but on Windows they default to 32. This caused an overflow problem in emg3d.meshes.good_mg_cell_nr.

If you change line 780 (emg3d v0.14.1) from

    primes = np.array([2, 3, 5, 7, 11, 13, 17, 19])

to

    primes = np.array([2, 3, 5, 7, 11, 13, 17, 19], dtype=np.int64)

the problem will be fixed.

I will push an update shortly.

Thanks for reporting!

[jingxu94]

Thinks, it works. That's a helpful tip for programming!

[prisae]

(With #144 I will hopefully start to also test on Windows and Mac machines, which should avoid these sort of issues in the future, hopefully.)