particle injection implementation + refactoring

[grzegorzbor]

No description provided.

[codecov]

Codecov Report

❌ Patch coverage is 93.72881% with 37 lines in your changes missing coverage. Please review.
✅ Project coverage is 96.89%. Comparing base (8e88445) to head (6d1b96a).
⚠️ Report is 15 commits behind head on master.
✅ All tests successful. No failed tests found.

Files with missing lines	Patch %	Lines
agnpy/time_evolution/time_evolution.py	89.81%	27 Missing ⚠️
agnpy/time_evolution/_time_evolution_utils.py	94.36%	8 Missing ⚠️
agnpy/spectra/spectra.py	83.33%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #183      +/-   ##
==========================================
- Coverage   97.22%   96.89%   -0.34%     
==========================================
  Files          44       48       +4     
  Lines        3826     4215     +389     
==========================================
+ Hits         3720     4084     +364     
- Misses        106      131      +25     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[jsitarek]

...SED to be much ...
Also I am not sure if I understand the problem here. You either way need to interpolate because the energy bins are shifting.
Does the approach for i==4 work also for i<4? If so you could just use it for testing all the time moments

[grzegorzbor]

The i==4 is the moment when we start to have a high peak at the last bean where acceleration and loss rates get equal, and intepolation error is high in this single point, so it needs some sort of a hack/workaround to make the sed_flux method work correctly. I'm not sure if it makes sense to use the same workaround for earlier steps when it's not needed. Using it in the last step only, we don't need to reset the n_e on the blob back to the original distribution.

[jsitarek]

strange to have it in 3 points, or are those problematic points show up for different times?

[grzegorzbor]

Yes they are for different times... i think for just 3 points it's not worth the effort to use different points for different times. Added the exaplnation

[jsitarek]

if you have a function with so many parameters it is more clear and less error prone not to use position arguments but execute it in this style ...(AAA=aaa, BBB=bbb, ...)
especially that some of the functions take first rel_inejction and some abs_injection

[grzegorzbor]

I cleaned it up to have the same order (rel first, abs second) everywhere.
I started also converting it to the AAA=aaa, BBB=bbb, ... style, but because parameter names are quite long, the code started getting quite obscure and difficult to read with these repetitions. So i think I would keep it as is, especially that every parameter uses a different unit, so the chances for an unnoticed error is rather low.

[jsitarek]

with 'absolute' escape you might also get negative density that would be unphysical as well, so you can change the condition from == 0 to <=0

[grzegorzbor]

Correct, but in such case I'd prefer to catch it earlier, for easier debugging. So I just added a check in the place where density is calculated. So here, it will not be possible to get negative value.