Bugs in `CompactObject/Test_Case/test_EOSgenerators.ipynb`

[ThibeauWouters]

While working on the JOSS review and looking into the CompactObject/Test_Case/test_EOSgenerators.ipynb example notebook, I encountered several issues preventing me from running the notebook

• from TOVsolver.maxium_central_density import maxium_central_density in the very first import code block throws an error
• The "Test_Case/Tolos_crust_out.txt" filename path is not found, it should be "Tolos_crust_out.txt" relative to the notebook filepath
• The first TOV solution that is shown in the documentation page here cannot be reproduced, below is the solution that I obtained after running the code. Similarly, the other TOV solutions looked different on my end compared to the solutions shown on the GitHub.

• In the DDH EOS section, the equations are not written in math mode.
• After following the installation instructions, a few packages are missing to run this example notebook: pandas, sympy, and NumbaMinpack are missing. I failed to install NumbaMinpack: please provide installation instructions in your documentation
• The code to benchmark the EOS generation runtime throws the following error:

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
Cell In[30], line 4
      1 ### Normal RMF ####
      2 
      3 first_run_time_normal = time.time()
----> 4 eps_fast, pres_fast = RMF.compute_EOS(eps_com, pres_com, theta)
      5 first_run_time_normal = time.time() - first_run_time_normal
      6 
      7 time_for_20_runs_normal = time.time()

File ~/Documents/Code/projects/joss_compactobject/CompactObject/lib/python3.11/site-packages/EOSgenerators/RMF_EOS.py:359, in compute_EOS(eps_crust, pres_crust, theta, return_tag)
    356 dt = 0.05
    357 rho_0 = 0.1505
--> 359 x_init = np.array(initial_values(0.1 * rho_0, theta))
    361 if return_tag:
    362     EoS = [[] for i in range(124)]

File ~/Documents/Code/projects/joss_compactobject/CompactObject/lib/python3.11/site-packages/EOSgenerators/RMF_EOS.py:57, in initial_values(rho, theta)
     53 m_p = 4.7583690772
     55 rho_0 = 0.1505
---> 57 sigma = g_sigma * rho / (m_sig**2)
     58 rho_03 = -g_rho * rho / (2.0 * (m_rho**2))
     59 omega = rho * (
     60     (((m_w**2) / g_omega) + (2.0 * Lambda_w * ((g_rho * rho_03) ** 2) * g_omega))
     61     ** (-1.0)
     62 )

TypeError: unsupported operand type(s) for *: 'function' and 'float'

Lastly, I would recommend an automated action in the GitHub CI/CD to execute such example notebooks, to verify whether they execute fine after installing the required packages and have correct code which is up to date with the code in the package.

[ChunHuangPhy]

Thanks for the detailed report. I was able to reproduce the notebook failures and have addressed them.

The maxium_central_density import typo has been corrected in the notebook to use maximum_central_density, and I also added a backward-compatible alias module so older notebooks fail less abruptly. The crust EOS loading is now robust from both the repository root and the Test_Case/ notebook directory.

The TOV/MR plotting issue was traced to inconsistent/nonphysical OutputMR results in some environments. I added normalization/validation around the notebook plotting path and updated OutputMR to fall back to the more robust tidal solver path when the primary solver returns nonphysical MR points. The notebook and generated docs now reproduce the mass-radius plots correctly.

The DDH markdown equations have been converted to LaTeX math mode. I also fixed a later DDH comparison issue where the built-in Char23 derivative expressions were missing the chain-rule factor; the built-in Char23 and user-defined Char23 curves now agree.

The missing dependencies have been added to the package metadata. The standard install now includes the packages used by the examples, including pandas, sympy, h5py, ultranest, and numba. NumbaMinpack remains optional because it requires a Fortran compiler. The docs now include installation notes, e.g. on macOS: brew install gcc followed by python -m pip install NumbaMinpack; on Ubuntu/Debian: install gfortran/cmake first.

The benchmark TypeError was caused by the notebook reusing a DDH-style theta containing callable coupling functions in the normal RMF benchmark. I renamed the notebook parameters so RMF/DDH sections cannot leak state into each other, and added validation so passing a DDH-style theta to RMF.compute_EOS now raises a clear error. The FastRMF benchmark is optional and is skipped with an explanatory message if NumbaMinpack is unavailable.

Finally, I added a GitHub Actions notebook-smoke job that installs the package and executes Test_Case/test_EOSgenerators.ipynb with jupyter nbconvert, so this example should not silently break again.

Hope this can address this issue!

[svisak]

I've tested the updated notebook and find that it works as expected, i.e., it produces the output shown in the current online documentation. I only encountered a couple of minor issues:

Calls like axs[0, 0].set_xlabel("Energy Density (MeV.fm⁻³)" ,fontsize=15) where there's a ⁻³ in the label, doesn't work with LaTeX rendering enabled, which I have by default. Disabling LaTeX rendering solves this.

In the section "(a) Compute DDH EOS", subsection "Function Output", the points 4-6 are displayed as "Math input error".

Concerning NumbaMinpack, the installation also fails for me due to CMake issue. I think this is an upstream issue.