Repository navigation
Conversation
2450955 to
554cf0a
Compare
|
I am not sure if we have a |
|
I just looked on the Team page. I see there isn’t anyone leading the cosmology module. That would make this difficult.
Much of it is just moving stuff from `core` to `builtin`, but the enhancement to `Cosmology`, the registry, and the entry points definitely need more careful attention.
I've just been cleaning house on all my uncommitted branches...
|
4537419 to
ea493fe
Compare
1ce02fb to
d934aaa
Compare
|
@pllim. Can you add a milestone, even if it's 6.0, so I can try to achieve all passing checks. |
d934aaa to
af56056
Compare
|
Done. @nstarman , for future reference, to ensure your work is not wasted, it is preferable to discuss your development plan for something that takes significant effort (such as refactoring a subpackage) with the core maintainers first before doing the work. Not all PRs are merged for a reason or another. Thanks for your effort! |
|
Thanks! Completely fair and I definitely should have done that. I had most of this code done for personal research and sitting around since last summer. I just added some of the cleanup from #10624 before submitting this PR. Most of this PR is just moving code from |
|
@pllim What's the best way to reach out? There are a number of convenience functions, such as to calculate z matter-radiation equality, that I would like to add (as a separate PR from this one). |
|
There are a few channels listed at https://www.astropy.org/help.html . Maybe try the "developer email list"? |
|
For this PR, whom can I contact or with whom can I work? |
|
😬 |
|
I think there should also be a way to provide new cosmology realizations without having to do anything with entrypoints. I'm not really up on how this works currently though (I used to be but I haven't looked at it in a looong time). |
@embray
On further thought, I'm kind of leaning towards an ECSV instead of JSON because of the built-in unit and metadata support and the cool possibility of converting an MCMC chain to an ECSV file and built able to work with sets of cosmology realizations. Also, I think the astropy config folder would be a good place to auto-import user-provided realizations. Anything else would be done by the user. I would love to implement this... #11345 is the first step... 🚀 |
builtin cosmology entrypoints codecov RTD and CircleCI fixes fix typo in test_builtin change handing of read-only states update milestone Signed-off-by: Nathaniel Starkman <[email protected]>
a6d6421 to
219bf34
Compare
What this PR does already and how it should be split into smaller PRs for review. In no particular order:
default_cosmology) to a separate file —realizationsI've currently called thisbuiltin, butrealizationmight be better.default_cosmologyto be a registry (a lagalactocentric_frame_defaults) so users can add their own cosmology realizations and set them as the default.4.1 Add entrypoints to the realizations, so that third-party packages can add specific cosmology realizations and users can access them straight through astropy.
4.2 Also allow auto-import of user-defined realizations from a file(s).
Enhance the
Cosmologyclass so it initializes the default cosmology, with any user-passed modifications. This centralizes the cosmology creation process for any "perturbations" about the default cosmology. Finally a use for the base class.[Not in this PR] Add entrypoints to
core, where theCosmologyclasses are defined, so that third-party packages can add cosmologies and have them integrate well with the realizations machinery.At the end, astropy will provide a core set of cosmology classes and realizations, but be highly extensible to allow user-defined realizations, and aggregate third-party cosmologies and realizations.
This PR might also might partially alleviate the lack-of-maintainer problem. Instead of programming more
Cosmologyclasses, we let third-party packages demonstrate user need. If a Cosmology class becomes stable & popular enough then it could be added as a built-in.migrated to #11777 since this will never be a Pr