Backward compatibility is not that convincing. If the items should not be imported (that is, if it was a mistake in some sense to import them), then we should fix that. Is there a good way to deprecate imports? (That is, print a warning the first time someone uses sys...., for example?)
Re Nils' remarks: I think that we should have good reasons every time we deviate from standard Python. From that point of view, maybe (?) "math" or "operator" would be sensible things to import, although I don't remember the last time I used either. Do others use these frequently? "sys" and "os" are not directly tied to anything Sage-specific; they don't seem like they belong at all (although I do use them). -- John On Tuesday, September 8, 2020 at 8:31:49 PM UTC-7, William wrote: > > +1 to Nils remarks. > > Please also consider backward compatibility. If you remove a bunch of > things, I'm probably just going to have to add them back on cocalc to avoid > all the headaches of people's code breaking. There might be a ton of random > code out there that will break if you remove a bunch of imports. > > It would be nice to have a different entry point to Sage that has > *dramatically* less imported by default. By I don't think it should be the > default, just because of backward compatibility. > > William > > On Tuesday, September 8, 2020 at 12:35:38 AM UTC-7 Nils Bruin wrote: > >> On Tuesday, September 1, 2020 at 9:50:07 PM UTC-7, John H Palmieri wrote: >> >>> There is a ticket (https://trac.sagemath.org/ticket/25383) about >>> removing some Sage functions from the global namespace, which I think is a >>> good idea. But Sage also imports some Python modules: >>> >>> - os >>> - sys >>> - operator >>> - math >>> - warnings >>> >> o u >>> >> Should we keep all of these? Remove all? Keep some? >>> >>> >> While I understand the drive to clean up the global namespace and in >> principle agree with the aesthetics of doing so, I've never been >> disappointed by something being available in the global namespace and on >> occasion am slightly annoyed by having to import something in python prior >> to use that I for some reason thought would there be by default (e.g., many >> of the functions in "math"). I've sometimes have had mildly surprising >> results by encountering something I didn't expect (R being the R interface >> is the main one there), but that was always quickly diagnosed. So, are we >> actually solving a practical problem by cleaning up the global namespace? >> If it measurably improves start-up time then we do have a win, but I >> suspect that the time-consuming start-up imports are necessary anyway (or >> are already lazy). >> > -- You received this message because you are subscribed to the Google Groups "sage-devel" group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/sage-devel/be25c103-3181-48ab-b35c-31de145c1901o%40googlegroups.com.
