[AstroPy] Wrong earth mass in astropy.constants

Eric Winter ewinter at stsci.edu
Mon Nov 28 11:00:58 EST 2016

I think the consideration of versioning the constants (not the same as versioning the code) is the right approach. Someone else suggested creating subsections of the package based on the source and year of the constants. I think that’s a good approach. The “latest” version would be the default, and alternative versions could be made available with the appropriate import statements.


On 11/28/16, 10:56 AM, "AstroPy on behalf of Demitri Muna" <astropy-bounces at scipy.org on behalf of demitri.muna at gmail.com> wrote:

    On Nov 28, 2016, at 10:44 AM, Thomas Robitaille <thomas.robitaille at gmail.com> wrote:
    > The main problem we are facing in Astropy is that if we update constants from one release to the next, results might then change, which might not be acceptable for some users. Therefore, we need to make sure we have a mechanism for users to import a 'fixed/frozen' version of constants that will not change in future releases of Astropy, and this is why there are delays currently in updating some of the constants.
    Taking the devil's advocate position-
    1) Those that are worried about things like the Earth’s mass to the precision of the edge of current research (or similar constants) probably are putting in their own value and not relying on a library. That seems like the right thing to do in that case.
    2) While a tiny of change of a constant at the most precise level would change a person's calculation from one release to another, they should be checking equivalence to a certain precision, not an exact value when using constants like these.
    I would argue for always adopting the most current "trusted" value (where "trusted" can be defined as JPL, IAU, etc.). Changes in a user's output because of changes in constants like these is a reflection of a bad software test, not the library.
    Demitri Muna
    Center for Cosmology and AstroParticle Physics
    & Department of Astronomy
    Le Ohio State University
    My Projects:
    AstroPy mailing list
    AstroPy at scipy.org

More information about the AstroPy mailing list