Update fitting functions

[erwanp]

Description

• add Spectrum.sort() method
• add 1-T fitting example
• just-in-time tabulation of partition functions for multi-Tvib with Treanor distribution
• update 3-T fitting example, updating fitting method
• add download-HITEMP example

[codecov-commenter]

Codecov Report

Merging #336 (4dc8f64) into develop (781fb52) will increase coverage by 0.07%.
The diff coverage is 80.17%.

@@             Coverage Diff             @@
##           develop     #336      +/-   ##
===========================================
+ Coverage    75.17%   75.24%   +0.07%     
===========================================
  Files          144      146       +2     
  Lines        15957    16050      +93     
===========================================
+ Hits         11995    12077      +82     
- Misses        3962     3973      +11     

[lgtm-com]

This pull request fixes 1 alert when merging ef47b76 into 781fb52 - view on LGTM.com

fixed alerts:

• 1 for Redundant assignment

[minouHub]

• There are multiple warnings about the iso column. I know you want to get rid of it in the near future, but just to remind you.

e:\python\radis\radis\misc\warning.py:341: PerformanceWarning: There shouldn't be a Column 'iso' with a unique value
  warnings.warn(WarningType(message))

• In your example at Add fit spectrum method #334 (comment) : replace maxiter = 300 by solver_options={"maxiter": 300}. Although, I don't find this example anywhere in the code.
• What is the difference between the fit_spectrum in factory.py and fitting.py. Seems odd to me to have the same function define twice.
• Finally, this is what I get when I run the example at Add fit spectrum method #334 (comment). Not a great fit

Init ['T12', 'T3', 'Trot'] = [1150. 2650. 1150.]['', '', '']
Final ['T12', 'T3', 'Trot'] = [1150. 2650. 1150.]['', '', '']
Best ['T12', 'T3', 'Trot'] = [1150. 2670. 1150.]['', '', ''] reached at iteration 4/4

[minouHub]

The example doesn't show anything on my side

[erwanp]

yes ! It will only be generated on Readthedocs (once merged)

[lgtm-com]

This pull request fixes 1 alert when merging f57bb84 into 781fb52 - view on LGTM.com

fixed alerts:

• 1 for Redundant assignment

[erwanp]

Thanks for the review!

• fit_spectrum and fitting.py --> the 1st is just a wrapper to the 2nd. The idea is to make it easier to use fitting.py with other codes; as long as they have a Factory with similar syntax. Think of a SpecairFactory for pyspecair : )
• I fixed all the rest. Even allowed more iterations in multi-temperature fit eventually converge to a nice fit. My only concern was that it was very long ; but your feedback convinced me it's better to show a good fit than save 1min of computation on Readthedocs.

Note : fitting is very long because parsum_mode="tabulation" cannot be activated as it is stuck with vaex. I hope to fix that (will make it 10x faster!). It's also long because the fitting algorithms (low level scipy.optimize) are not necessarily optimized. I want to switch to astropy.modelling or lmfit to have higher-level; more optimized control of the fits

[lgtm-com]

This pull request fixes 1 alert when merging ce26cc1 into 781fb52 - view on LGTM.com

fixed alerts:

• 1 for Redundant assignment