Issue when TS_vaue and Variability_Index being empty strings

[sfegan]

          We have had the same problem a few times at LLR when analyzing ROIs that contain the Crab Nebula. I understood it to be due to the fact that the Crab is modelled by multiple components, one of which is fixed in the catalog, so there is no TS_value given for it and the code dies. The same problem could also occur for sources that don't have "Variability_Index".

We ended up fixing it by changing the code to test explicitly that these two attributes aren't a null string before trying to make them into float :

        #evaluate variability and significance threshold info
        self.sources[name].update([('variable',source.getAttribute('Variability_Index')=="" or float(source.getAttribute('Variability_Index'))>=self.variability_threshold),
            ('significant',source.getAttribute('TS_value')=="" or float(source.getAttribute('TS_value'))>=self.sigma_to_free)])

We default to keeping them in the model if the attribute isn't present.

Originally posted by @sfegan in #68 (comment)

[sfegan]

Sorry to open another issue here.. github won't let me reopen the other issue as I wasn't the original author

[physicsranger]

Thanks for opening this. I should be able to make an update this weekend unless you want to try and do a PR.

[physicsranger]

Okay, it took me a little bit longer to find time to address this issue, but that is now done with release v1.10.11 (on PyPI and created as a tagged release on GitHub). I did this creating a function LATSourceModel.utilities.float_conversion to make things a bit more compact. I tested building a region centered near the Crab, using the XML catalog, and was able to make the model successfully.