TypeError: only 0-dimensional arrays can be converted to Python scalars

[XD1001]

After upgrading the PDSim to the newest version with Python=3.12.12, cython=3.2.3, numpy=2.4.0, an error raises when running the scroll compressor example code examples/scroll_compressor.py (error message listed below):

Traceback (most recent call last):
  File "c:\scroll\scroll_example.py", line 436, in <module>
    Compressor(Scroll)
  File "c:\scroll\scroll_example.py", line 267, in Compressor
    ScrollComp.set_disc_geo('2Arc',r2 = 0)
  File "C:\ProgramData\Anaconda3\envs\pdsim\Lib\site-packages\PDSim\scroll\core.py", line 883, in set_disc_geo
    scroll_geo.setDiscGeo(self.geo,Type,r2,**kwargs)
  File "PDSim/scroll/symm_scroll_geo.pyx", line 256, in PDSim.scroll.symm_scroll_geo.setDiscGeo
  File "PDSim/scroll/common_scroll_geo.pxd", line 108, in PDSim.scroll.common_scroll_geo.geoVals.xa_arc2.__set__
    cdef public double xa_arc2,ya_arc2,ra_arc2,t1_arc2,t2_arc2
TypeError: only 0-dimensional arrays can be converted to Python scalars

It seems that this error is because that somehow the type of xa_arc2 becomes arrays instead of double. My guess is that this is caused by some issues during the discharge tip geometry calculation process? Not sure if this is caused by new features in the new python and cython version?

[ibell]

Can you figure out which variable(s) this pertains to?

[XD1001]

Sorry for the late reply.

Unfortunately I cannot see directly which variable(s) causes this error. My best guess would be (x_os, y_os), (nx_os, ny_os) or (dx, dy) maybe, due to following reasons:

1. Noticing that the error is caused by assigning the calculated 0-dimentional array xarc2 to the preclaimed double geo.xa_arc2 (Line 256 in PDSim.scroll.symm_scroll_geo.setDiscGeo)
2. (xarc2, yarc2) is calculated in PDSim.scroll.symm_scroll_geo.setDiscGeo Line 239 - 240 by xarc2 = x_os+nx_os*r2
3. It could be either one of x_os, nx_os, and r2 (or all of them) are 0-dimentional array that leads to the 0-dimentional array xarc2.

It may need a closer look. What do you suggest?

[ibell]

Either print the variables at the call site, or use a debugger

[XD1001]

Thank you for the suggestion and apologies for the delayed reply.

Regarding printing variables or using a debugger at the call site: set_disc_geo() wraps the Cython-implemented setDiscGeo() in symm_scroll_geo. Since the Cython source is compiled into a binary extension, standard Python debugging techniques (e.g., inserting print statements) aren't feasible without modifying and recompiling the .pyx file.

That said, I verified the types of all relevant input variables immediately before invoking: ScrollComp.set_disc_geo("2Arc", r2=0.0)

All values passed to the underlying Cython function are confirmed as native Python floats:

xa_arc2 type: <class 'float'>
ya_arc2 type: <class 'float'>
ra_arc2 type: <class 'float'>
t1_arc2 type: <class 'float'>
t2_arc2 type: <class 'float'>

Thank you for the support.

[XD1001]

It seems that this issue is caused by numpy. In the current version, numpy version is not specified and will install numpy=2.4.2 as default. Once downgrade numpy=2.4.2 to any version lower than 2.4.0 (cannot be 2.4.0), the issue is then solved.

It is noticed that numpy=2.4.0 has a Backward compatibility bug. Maybe this could be the root cause? My guess is that a general update of pdsim.scroll code would be helpful to avoid similar issue in future.

Thanks much.

[ibell]

Easiest solution then would be to pin numpy<2.4 in the various env files. BTW, I have pretty much fully migrated to the uv package manager at work instead of conda, and maybe will implement that in pdsim as well; been very happy with it.

[XD1001]

Yes I would agree.

I agree that uv manager is very powerful and I also work with it for other projects. Looking forward to seeing PDSim also using uv in future.

Thank you for the efforts!

[ibell]

You would be welcome to implement uv for pdsim!