| 18 Sep 2019 |
| 02:30:43 |  Nick Murphy | Hm...we did mean to make them public, though I had also been wondering if it would be better to put them into the Particle class as attributes in some way. I don't think they're going to be used very often. |
| 19:05:14 | Nick Murphy | The poster on PlasmaPy is going to be on Wednesday morning at the APS DPP meeting in October. Anyone else who can be at the poster so I don't lose my voice as quickly would be appreciated! |
| 19:05:57 | Nick Murphy | I'm also giving a talk on writing clean code for plasma simulation later that afternoon, so I'm going to do my best to not lose my voice. |
| 19 Sep 2019 |
| 16:48:38 | Nick Murphy | Should be about another month before Python 3.8 is released! |
| 17:02:34 | Nick Murphy | I don't remember if we've decided on how long we want to remain 3.6 compatible (or more generally, how far back in terms of Python releases do we want to retain compatibility). My thought is that while we are very much in the development phase, it should be fine to be compatible with just the two most recent versions of Python (e.g., it would be reasonable for v0.3.0 to be Python 3.7+). Once we get more users and have more stability, then probably the two most recent versions would be necessary. |
| 17:08:40 |  Cadair | https://numpy.org/neps/nep-0029-deprecation_policy.html |
| 17:16:05 | Nick Murphy | Thanks, Cadair! That looks really useful. I can't remember if that went into the PyHC guidelines, but it probably should. |
| 17:16:40 | Nick Murphy | I guess that means it'll be ~26 months before I get to use walrus operators := |
| 17:44:24 | Nick Murphy | * I guess that means it'll be ~26 months before I get to use walrus operators := |
| 18:04:44 |  Ben Alterman | In reply to @namurphy:matrix.org
I guess that means it'll be ~26 months before I get to use walrus operators := The position only argument identifier is also interesting. |
| 18:53:28 | Nick Murphy | The reason I was looking up Python 3.8 was because I had a use case for a position only argument identifier. I was hoping it was in 3.7 so I might be able to justify using it, but it was in 3.8. |
| 18:54:48 | Cadair | The best thing in 3.8 is the debug f-strings. |
| 18:55:11 |  SolarDrew | All hail f-strings |
| 18:55:17 | Nick Murphy | It may be kind of like the feeling of 3.6 coming out and having f-strings, and then having to wait a really long time before being able to use them in core packages. Or like having to wait three years after The Empire Strikes Back came out before Return of the Jedi. |
| 18:55:18 | Cadair | >>> a = 1234
>>> f"{a=}"
'a=1234'
|
| 18:55:42 | Cadair | Prints the variable name and value in 1 extra character |
| 18:55:43 | Nick Murphy | The debug f-strings look fantastic too! The walrus operator just sounds cooler. |
| 18:56:21 | Nick Murphy | Can't we get a time machine and go to 26 months from now so we can use debug f-strings? |
| 18:57:40 | Cadair | The best thing about the debug f-strings is they are for debugging. |
| 18:58:06 | Cadair | You are unlikely to commit them to a project with a supported python version, they are just a developer quality of life upgrade |
| 18:58:08 | Nick Murphy | In [1]: from astropy.time import time_machine
---------------------------------------------------------------------------
ImportError Traceback (most recent call last)
<ipython-input-1-a98d67ba0710> in <module>
----> 1 from astropy.time import time_machine
ImportError: cannot import name 'time_machine' from 'astropy.time' (/media/Scratch/namurphy/envs/py37/lib/python3.7/site-packages/astropy/time/__init__.py)
|
| 18:58:14 | Cadair | * The best thing about the debug f-strings is they are for debugging. |
| 18:59:09 | Nick Murphy | This is true, though they could be used a lot in the test suite for coming up with useful error messages |
| 19:00:17 | Cadair | Sure, but they are more useful than the walrus to me without bumping sunpys versions lol |
| 19:01:33 | Nick Murphy | Very true! The data classes in 3.7 probably would get a lot of use from us in the meantime. |
| 19:05:07 | Nick Murphy | Going back to NEP 29 that Cadair mentioned... Does it sound reasonable to everyone to abide by their recommendations for support of old versions of Python and NumPy? If so, we can put it in the developer docs. |
| 19:06:22 | Nick Murphy | We could possibly run into exceptions, like if there's a new dtype in NumPy that really helps with units stuff. |
| 19:14:09 |  Dominik Stańczak | As a general rule barring exceptional cases like HOLY SPITZER, UNITS WORK WITH CROSS PRODUCTS NOW - sure! |
| 20:46:02 | Cadair | We are pretty much at NEP 29 apart from astropy |
| 20:46:17 | Cadair | For similar reasons |
