Python: Download bridgestan source if needed #80
Conversation
|
I'd love to get some opinions on this @bob-carpenter @roualdes |
|
Re location: Re "should Julia also look in this location". I guess I'll vote no, even though on first thought it does sound somewhat appealing. Julia sharing Python's location would effectively build in an asymmetry across Python and Julia. Julia would be able to look in both |
|
Re: minor versions Yes, if a user was really good about following updates to bridgestan they’d eventually end up with a 1.0.2, 1.0.3, … etc. I think this is actually a good thing with our versioning scheme, since then someone saying “I’m using version x.y.z” is completely unambiguous. The only downside I see is the potential for bloat of that directory over time. If we wanted to we could print out a message if older versions exist telling the user they’re free to delete those older ones? I don’t want to delete them automatically, since it’s possible a user could have several different versions installed e.g. in different virtual environments |
There was a problem hiding this comment.
The immediate verification of the BridgeStan path after a user tries to set it is nice.
I'm waffling about a friendly reminder that's printed out whenever get_bridgestan_src() is called, reminding the user about all the versions of BridgeStan that exist in ~/.bridgestan. On the one hand, I see such a message as helpful. On the other hand, I don't know how to do a similar thing in Julia. So I guess we should put this off for now, and move forward with the nice feature contained in this PR.
This looks good. Thanks.
| @@ -1,4 +1,5 @@ | |||
| from .compile import compile_model, set_bridgestan_path | |||
| from .model import StanModel | |||
| from .__version import __version__ | |||
There was a problem hiding this comment.
Are you asking if we should be doing more, or less, in the init file? Having bridgestan.__version__ defined is pretty common (e.g. https://github.com/numpy/numpy/blob/28bce82c82ba1151baeb7399abf454418b33eefe/numpy/__init__.py#L440) but I don't personally like adding it to __all__ because I don't think from bridgestan import * should import __version__ into global scope
There was a problem hiding this comment.
I was asking if we should add __version__ to __all__, with the thought (from memory) that numpy does this. But maybe my memory is incorrect here.
Your opinion makes sense to me.
And a user could still from bridgestan import __version__ correct?
There was a problem hiding this comment.
Correct, a user can still manually import it or access it qualified (import bridgestan; bridgestan.__version__)
It seems the "big" packages are inconsistent with whether or not they put it in __all__. numpy and scipy do, pandas and matplotlib do not. I can't find anything "official" which states a preference.
Would close #79
This is marked as a draft due to the following requirements not yet being met:
~/.bridgestan/bridgestan-VERSION