Start refactoring pyIRF

[maxnoe]

Implemented in the first commit:

• reading of EventDisplay DL2 fits files
• Calculating event weights
• Functions for binning events
• Li&Ma Significance

Definition of the internal data format, for now astropy.table.QTables
with several columns expected for DL2 event lists.

Co-authored-by: Michele Peresano michele.peresano@cea.fr
Co-authored-by: Lukas Nickel lukas.nickel@tu-dortmund.de

More to come

[codecov]

Codecov Report

Merging #36 into master will increase coverage by 71.39%.
The diff coverage is 72.00%.

@@             Coverage Diff             @@
##           master      #36       +/-   ##
===========================================
+ Coverage    0.72%   72.12%   +71.39%     
===========================================
  Files           9       29       +20     
  Lines         685      696       +11     
===========================================
+ Hits            5      502      +497     
+ Misses        680      194      -486     

Impacted Files	Coverage Δ
pyirf/cut_optimization.py	0.00% <0.00%> (ø)
pyirf/sensitivity.py	0.00% <0.00%> (ø)
pyirf/spectral.py	0.00% <0.00%> (ø)
pyirf/statistics.py	0.00% <0.00%> (ø)
pyirf/io/eventdisplay.py	40.00% <40.00%> (ø)
pyirf/utils.py	61.90% <61.90%> (ø)
pyirf/io/gadf.py	84.37% <84.37%> (ø)
pyirf/simulations.py	91.66% <91.66%> (ø)
pyirf/benchmarks/energy_bias_resolution.py	93.10% <93.10%> (ø)
pyirf/binning.py	94.73% <94.73%> (ø)
... and 45 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 0bf3f40...7c802a5. Read the comment docs.

[maxnoe]

Short update here:

With the current version of the script in exaples/calculate_eventdisplay_irfs.py, we calculate a sensitivity that looks correctly calculated but which is different from the event display one due to different cut optimization, especially theta cuts are very different as what EventDisplay calculates does not seem to be a simple, bin-wise 68% containment.

At the moment, we do:

• Calculate theta cut as 68 % containment radius in very fine reco energy bins with a soft gamma hadron cut applied
• Optimize the gh cut in the bins of the sensitivity

[HealthyPear]

At the moment, we do:

* Calculate theta cut as 68 % containment radius in very fine reco energy bins with a soft gamma hadron cut applied

* Optimize the gh cut in the bins of the sensitivity

As far as I understood we are then only missing the 3rd step, which is to re-calculate the bin-wise 68% containment angular cut after the g/h bin-wise optimization

[HealthyPear]

we could add the codacy badga again though, for me it was working

[maxnoe]

If it works, great. Maybe codacy was just down for a while when I tried.

[HealthyPear]

If it works, great. Maybe codacy was just down for a while when I tried.

It should: the badge was pointing to this url- though, it's the master, the PR should be automatic I think

I will re-add it by copy-pasting the original RST command from their website just to be sure there wasn't a typo

what about the GitHub pages build? what are the equivalent RST links for image and target?
I couldn' find a specific github pages deployment badge, and using the one of travis just doubled the information.
I added a specific link for the docs, even though I think it's still easier for people to use the much easier to see link in the about page (I personally never used a readme to go to the docs...)

I think we are good to merge now

[HealthyPear]

I just realized that we should probably add cython to extras_require["tests"], otherwise gammapy makes the installation crash for that section.

Modifying setup.py in this way worked for me. Can you confirm?

bug in installation