You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
There are some regions of RM-Tools which now deprecated, it would be nice to explicitly mark all code which is deprecated and make plans to remove it at some point. The deprecation package offers a very nice way to handle with via decorators. I've got a PR incoming that makes the obviously deprecated code as such.
I think we should also consider deprecating the large and unwieldy code that really should be handled by external packages. Namely: emcee, nestle, corner, and mpfit. The first three can probably be immediately replaced with their properly maintained versions. mpfit should probably be replaced by scipy tools (like curve_fit) but this might take some refactoring.
Finally, we need to catch up on some Python updates. Version 3.7 has reached end-of-life, we should consider sunsetting support. Further, we should add up to version 3.12 for testing. Further, Numpy version 2.0 is on the horizon with major, potentially breaking, changes. We should restrict our reqs to numpy < 2 until we decide to migrate.
I've wrapped up most of these suggestions into a PR, with the exception of removing the larger copies of external packages.
The text was updated successfully, but these errors were encountered:
Merged. I also fixed the issues with the tests that appeared (not as a result of the changes in the code, but as a result of testing with other version of Python). There were two problems:
higher versions of numpy have removed the compatibility of np.percentile() with complex numbers (which affected some plotting code for QU fitting). Fixed this by splitting the real/imaginary components (which were already plotted separately).
higher versions of Python caused very minor (<1 part in 10^7) differences in outputs, which was causing some badly designed tests to fail (because it was doing string comparison instead of float comparison on the polynomial fit coefficients). I've cleaned up the broken test.
There are some regions of RM-Tools which now deprecated, it would be nice to explicitly mark all code which is deprecated and make plans to remove it at some point. The deprecation package offers a very nice way to handle with via decorators. I've got a PR incoming that makes the obviously deprecated code as such.
I think we should also consider deprecating the large and unwieldy code that really should be handled by external packages. Namely:
emcee
,nestle
,corner
, andmpfit
. The first three can probably be immediately replaced with their properly maintained versions.mpfit
should probably be replaced byscipy
tools (likecurve_fit
) but this might take some refactoring.Finally, we need to catch up on some Python updates. Version 3.7 has reached end-of-life, we should consider sunsetting support. Further, we should add up to version 3.12 for testing. Further, Numpy version 2.0 is on the horizon with major, potentially breaking, changes. We should restrict our reqs to
numpy < 2
until we decide to migrate.I've wrapped up most of these suggestions into a PR, with the exception of removing the larger copies of external packages.
The text was updated successfully, but these errors were encountered: