Chianti collision data broken for heavier elements

[jsilvabeyer]

Describe the bug

When using ChiantiReader to get collision data for heavier elements (I tested O and Si), the resulting TARDISAtomData collision DataFrame has atomic_number and ion_number (among others) as float64, instead of int64. This results in an error when using the data for NLTE ions in TARDIS.

To Reproduce

Run the Quickstart notebook, but without using CMFGEN, and using CHIANTI as:

chianti_reader = ChiantiReader('O; Si', collisions=True, priority=30)

Screenshots

System

• OS:

  • GNU/Linux

• Environment (conda list): carsus environment

Additional context

Also, when running TARDISAtomData, the logger gave me a math warning

RuntimeWarning:

divide by zero encountered in log10

I tend to ignore those but maybe is related to my issue?

---

Edit: I found that these values can't be changed into int64 because there are nan values among them. The ChiantiReader object doesn't show any value like that, so somewhere in the TARDISAtomData.create_collisions method some numbers are becoming nan.

[andrewfullard]

This is likely to be resolved by #406