« Changelog History


astropy v1.0.6 Release Notes

Release Date: 2015-10-22 // over 8 years ago

  • 🐛 Bug Fixes

    astropy.analytic_functions

    • 🛠 Fixed blackbody analytic functions to properly support arrays of temperatures. [#4251]

    astropy.coordinates

    • 🛠 Fixed errors in transformations for objects within a few AU of the Earth. Included substantive changes to transformation machinery that may change distances at levels ~machine precision for other objects. [#4254]

    astropy.io.fits ^

    • 👷 fitsdiff and related functions now do a better job reporting differences between values that are different types but have the same representation (ex: the string '0' versus the number 0). [#4122]

    • 🛠 Miscellaneous fixes for supporting Numpy 1.10. [#4228]

    • 🛠 Fixed an issue where writing a column of unicode strings to a FITS table resulted in a quadrupling of size of the column (i.e. the format of the FITS column was 4 characters for every one in the original strings). [#4228]

    • ➕ Added support for an obscure case (but nonetheless allowed by the FITS standard) where a column has some TDIMn keyword, but a repeat count in the TFORMn column greater than the number of elements implied by the TDIMn. For example TFORMn = 100I, but TDIMn = '(5,5)'. In this case the TDIMn implies 5x5 arrays in the column, but the TFORMn implies a 100 element 1-D array in the column. In this case the TDIM takes precedence, and the remaining bytes in the column are ignored. [#4228]

    astropy.io.votable

    • 🛠 Fixed crash with Python compiler optimization level = 2. [#4231]

    astropy.vo ^

    • Fixed check_conesearch_sites with parallel=True on Python >= 3.3 and on Windows (it was broken in both those cases for separate reasons). [#2970]

    Other Changes and Additions

    • ✅ All tests now pass against Numpy v1.10.x. This implies nominal support for Numpy 1.10.x moving forward (but there may still be unknown issues). For example, there is already a known performance issue with tables containing large multi-dimensional columns--for example, tables that contain entire images in one or more of their columns. This is a known upstream issue in Numpy. [#4259]