6 May 2022 |
tepickering | In reply to@slack_astropy_U774ARTKR:openastronomy.org I really don't remember why .org is in mirror, not primary. So pls PR it so Marten and Erik can have a look. Thanks! yes, flip the IERS_A URLs so that IERS's datacenter is primary and switch the IERS_B URL to point to datacenter instead of obspm.fr. | 17:15:37 |
pllim | In reply to@slack_astropy_U86349G3C:openastronomy.org yes, flip the IERS_A URLs so that IERS's datacenter is primary and switch the IERS_B URL to point to datacenter instead of obspm.fr. Yes, I think that is acceptable for v5.1 unless saimn objects. | 17:18:13 |
pllim | In reply to@slack_astropy_U774ARTKR:openastronomy.org Yes, I think that is acceptable for v5.1 unless saimn objects. Two trailing null characters lead to over 100 comments here. Angry fist at round Earth! | 17:19:26 |
tepickering | In reply to@slack_astropy_U774ARTKR:openastronomy.org Two trailing null characters lead to over 100 comments here. Angry fist at round Earth! people get annoyed when mission critical software falls over due to two trailing null characters... | 17:23:01 |
pllim | In reply to@slack_astropy_U86349G3C:openastronomy.org people get annoyed when mission critical software falls over due to two trailing null characters... I got a formal reply and posted it at https://github.com/astropy/astropy/issues/13223#issuecomment-1119841197 FYI | 17:36:04 |
tepickering | In reply to@slack_astropy_U774ARTKR:openastronomy.org I got a formal reply and posted it at https://github.com/astropy/astropy/issues/13223#issuecomment-1119841197 FYI https://github.com/astropy/astropy/pull/13226 | 18:07:07 |
9 May 2022 |
| Jesús Fuentes joined the room. | 10:35:43 |
| Jesús Fuentes changed their display name from slack_astropy_U03EP9JAE1J to Jesús Fuentes. | 10:36:28 |
| Jesús Fuentes set a profile picture. | 10:36:30 |
10 May 2022 |
hamogu | dev telecon in 72 min:
https://zoom.us/j/94465601266?pwd=aldieHR3NGp2RlRnWEtZanlCaVgzZz09
Meeting ID: 944 6560 1266
Passcode: astropy!
agenda is here: https://docs.google.com/document/d/15JSFh3OMF9Iz6ov3q_xxGO_BL8hRnuse4IMUrqEIvcg/edit?usp=sharing | 13:48:57 |
hamogu | All are welcome, this is an open telecon for everyone doing anything on the project! | 13:49:37 |
JD Smith | In reply to@slack_astropy_U0LE97SBD:openastronomy.org Cool, looks like you have something that is pretty much there. I spent a lot more time than I expected trying to close the loop for making sure the format functions can round-trip. There are some weird issues deep in Table that I don't understand right now. Anyway, check out this gist:
https://gist.github.com/taldcroft/44a88f079f4afe85ae2a14c75ac2795d
Another unfortunate thing is that I uncovered a bug in serializing structured columns, namely that you need to have the structured names in alphabetical order for things to round-trip. This is because YAML helpfully puts dict entries into alphabetical order. Thanks for working on this and sorry for the slow response. Picking up on this, one other little thought I had was to simply slot fmt_func into info.format “just in time”, temporarily, as the table is being _printed,_ using a closure incorporating the original info.format. This would make it seem like formatting “just works” on the entire column, and relieves the need to remove/restore fmt_function on round-tripping to file. Here’s my approach.
Since the structured arrays are only supported in 5.1, I’d still prefer a “real” (n,3) array backing. Also I can’t seem to mask a structured array column. Any further thoughts on that approach, whereas it’s trivial to mask any parts of a multi-dimensional column. | 21:43:38 |
tomaldcroft | In reply to@slack_astropy_U01CPK6VCMP:openastronomy.org Thanks for working on this and sorry for the slow response. Picking up on this, one other little thought I had was to simply slot fmt_func into info.format “just in time”, temporarily, as the table is being _printed,_ using a closure incorporating the original info.format. This would make it seem like formatting “just works” on the entire column, and relieves the need to remove/restore fmt_function on round-tripping to file. Here’s my approach.
Since the structured arrays are only supported in 5.1, I’d still prefer a “real” (n,3) array backing. Also I can’t seem to mask a structured array column. Any further thoughts on that approach, whereas it’s trivial to mask any parts of a multi-dimensional column. Masked structured arrays (for pure numpy) are a thing that work IIRC. Are you finding a problem within astropy? | 22:07:09 |
11 May 2022 |
JD Smith | In reply to@slack_astropy_U0LE97SBD:openastronomy.org Masked structured arrays (for pure numpy) are a thing that work IIRC. Are you finding a problem within astropy? For normal arrays in a column within a Table with masked=True, you can just set the value to np.ma.masked. For structured array columns, this results in:
UserWarning: Warning: converting a masked element to nan | 01:56:34 |
tomaldcroft | In reply to@slack_astropy_U01CPK6VCMP:openastronomy.org For normal arrays in a column within a Table with masked=True, you can just set the value to np.ma.masked. For structured array columns, this results in:
UserWarning: Warning: converting a masked element to nan Ah. The problem is due to the issue that is fixed in https://github.com/astropy/astropy/pull/13236, namely that the way you wrote it the np.ma.array is getting auto-converted to NdarrayMixin which views the array as ndarray and loses the masking. But with 13236 this mostly works, though the column repr is not perfect.
In [2]: PAR_DTYPE = np.dtype([("a", "f"), ("b", "f"), ("c", "f")])
In [3]: a = np.ma.array(
...: [(np.pi, 2, 4.5), (np.pi / 2, 1, 3.1)], dtype=PAR_DTYPE, mask=[(0, 0, 0), (0, 0, 1)]
...: )
In [5]: t = Table()
In [6]: t["a"] = a
In [7]: t
Out[7]:
Table length=2
a [a, b, c]
(float32, float32, float32)
------------------------------
(3.1415927410125732, 2.0, 4.5)
--
In [8]: t["a"]["c"]
Out[8]:
MaskedColumn name='a' dtype='float32' length=2
4.5
--
In [9]: t["a"]["b"]
Out[9]:
MaskedColumn name='a' dtype='float32' length=2
2.0
1. | 10:23:35 |
tomaldcroft | In reply to@slack_astropy_U0LE97SBD:openastronomy.org Ah. The problem is due to the issue that is fixed in https://github.com/astropy/astropy/pull/13236, namely that the way you wrote it the np.ma.array is getting auto-converted to NdarrayMixin which views the array as ndarray and loses the masking. But with 13236 this mostly works, though the column repr is not perfect.
In [2]: PAR_DTYPE = np.dtype([("a", "f"), ("b", "f"), ("c", "f")])
In [3]: a = np.ma.array(
...: [(np.pi, 2, 4.5), (np.pi / 2, 1, 3.1)], dtype=PAR_DTYPE, mask=[(0, 0, 0), (0, 0, 1)]
...: )
In [5]: t = Table()
In [6]: t["a"] = a
In [7]: t
Out[7]:
Table length=2
a [a, b, c]
(float32, float32, float32)
------------------------------
(3.1415927410125732, 2.0, 4.5)
--
In [8]: t["a"]["c"]
Out[8]:
MaskedColumn name='a' dtype='float32' length=2
4.5
--
In [9]: t["a"]["b"]
Out[9]:
MaskedColumn name='a' dtype='float32' length=2
2.0
1. BTW, you should not use masked=True in the Table creation. That is a relic of when all columns had to be either masked or unmasked, so with masked=True every column will be masked. | 10:25:36 |
JD Smith | In reply to@slack_astropy_U0LE97SBD:openastronomy.org BTW, you should not use masked=True in the Table creation. That is a relic of when all columns had to be either masked or unmasked, so with masked=True every column will be masked. Thanks; looks like 5.1 is going to be a big release for Table. I wonder how this will interact with your new structure formatter PR. Will it be able to correctly show -- for sub-units inside each record (i.e. just 3.1 masked above)? | 15:46:03 |
12 May 2022 |
hamogu | I just saw this (we started seeing failures in sherpa): https://github.com/scipy/oldest-supported-numpy/issues/53
I don’t know if it affects astropy (have not tried with new pip yet), but thought I’ll put in here in case someone runs into that. | 14:07:23 |
saimn | It's only when using --no-build-isolation so should not affect "normal" users. | 14:41:26 |
13 May 2022 |
| Krafty Pengu joined the room. | 14:12:49 |
| Krafty Pengu changed their display name from slack_astropy_U03FCNP5FBL to Krafty Pengu. | 14:13:15 |
| Krafty Pengu set a profile picture. | 14:13:18 |
| Krafty Pengu changed their profile picture. | 14:13:18 |
| Krafty Pengu changed their profile picture. | 14:14:05 |
| Krafty Pengu changed their display name from Krafty Pengu to Polaris. | 14:14:29 |
| Krafty Pengu changed their display name from Polaris to Krafty Pengu. | 14:17:24 |
William Jamieson | If any other maintainers would like to be added to the CODEOWNERS file so that they get automatically added as reviewers please comment on: https://github.com/astropy/astropy/pull/13241 | 20:40:46 |
14 May 2022 |
| @davis_:matrix.org joined the room. | 12:22:52 |
| Moderator banned @davis_:matrix.org (spam). | 12:22:56 |
Matt Craig | Astropy is working with NumFOCUS on a research project funded by the Gordon & Betty Moore Foundation to understand the barriers to participation that contributors, particularly those from historically underrepresented groups, face in the open-source software community. The research team would like to talk to new contributors, project developers and maintainers, and those who have contributed in the past about their experiences joining and contributing to Astropy.
Interested in sharing your experiences?
Please complete this brief “Participant Interest” form which contains additional information on the research goals, privacy, and confidentiality considerations. Your participation will be valuable to the growth and sustainability of diverse and inclusive open-source software communities. Accepted participants will participate in a 30-minute interview with a research team member.
Thank you for your interest! | 23:13:41 |