Problem about load_absorption_table in class EBL

[huangylei]

I've found differences in the EBL absorption(with Saldana-Lopez model) between agnpy and the data released by Saldana-Lopez.

Then I checked the function load_absorption_table and found that all the models use np.unique() to return the sorted unique elements of an array. However, we don't need to sort data in Saldana-Lopez model.

[cosimoNigro]

@huangylei, thanks a lot for reporting this.
I will make a fix soon.

Do you have a snippet to show the issue you are reporting?
The interpolated values are checked against the original table values in the tests, so I am surprised by this.

Thanks again.

[huangylei]

I used the table that Saldana-Lopez provided, and compared it with the result by agnpy. They are different. (Unfortunately
I cannot find the code)

I checked the function load_absorption_table into:

def load_absorption_table(self):
        """load the reference values from the table file to be interpolated later"""
        f = fits.open(self.model_file)
        self.energy_ref = (
            np.sqrt(f["ENERGIES"].data["ENERG_LO"] * f["ENERGIES"].data["ENERG_HI"])
            * u.keV
        )
        # Franceschini file has two columns repeated, eliminate them
        if self.model == "franceschini":
            self.z_ref = np.unique(f["SPECTRA"].data["PARAMVAL"])
            self.values_ref = np.unique(f["SPECTRA"].data["INTPSPEC"], axis=0)
        else:
            self.z_ref = f["SPECTRA"].data["PARAMVAL"]
            self.values_ref = f["SPECTRA"].data["INTPSPEC"]

It worked and perfectly match the result that Saldana-Lopez provided. After this change, the aborption should be: