Updates on organizing the Astropy Roadmap
Erik Tollerud
Dear Astropy community,
The Strategic Planning Committee (SPOC) has made progress over the past year and we’re excited to tell you about some proposed tweaks to the Astropy Project’s roadmap process. We are open to feedback on all aspects of this plan. In one sentence, our proposed new process would frame roadmap items as concrete and clearly defined “Objectives” that have measurable results and phrased in a way to make it clear when the item has been achieved. This is inspired by the framework of “Objectives and key results.”
With that in mind, we have reorganized the existing roadmap items in a Draft New Roadmap document. There are tabs and subtabs: The tabs are “Objectives” (organized by theme), taken from the prior roadmap document. Some subtabs include proposed metrics that could be used to measure progress on these objectives. Some prior roadmap items were split into multiple objectives that contain multiple metrics/results, so there is not an exact one-to-one mapping from the prior roadmap to the new roadmap objective items. While actionable objectives have been drafted in every category, more work is needed to identify realistic and meaningful target key results for them.
A new aspect of this process is that we would like to have a clear timeframe in mind for objectives and measurable results. This will help drive prioritization based on need and opportunity (e.g., available work effort). We would like to set and review this prioritization with a 6 month cadence.
At the Coordination Meeting, we'd like community input on:
1) this proposed new structure,
2) modifications and updates to existing roadmap items,
3) new objectives to add, and
4) key results to use to measure progress on the objectives.
If you have time to review the document before the coordination meeting, that would be helpful, but we will also review it all at the meeting.
The exact mechanics of this process are still open for discussion and we would love to hear your input. The linked Draft New Roadmap document is a temporary Google Doc to be used during this phase of rapid iteration on the content and structure. We have not yet made any decisions about the tooling we will use to implement or track this proposed new framework.
As always, not all Astropy work needs to be on the roadmap: As a volunteer or in-kind effort, everyone is still free to contribute in the way that makes sense to them and honors their own personal or institutional priorities. This proposed new process and structure is primarily meant to give us a clearer framework for prioritizing, assessing, and reporting on work supported by Astropy’s grants.
We are looking forward to hearing your thoughts!
Sincerely,
The Strategic Planning and Organization Committee (SPOC)
Kelle Cruz
Adrian Price-Whelan
Tom Robitaille
Nathanial Starkman
Erik Tollerud
Marten van Kerkwijk
That's a great improvement -- very much in favour of actionable
objectives and key results!
I made a few comments on the document, but realized that what I'm
thinking about is probably bigger than small additions: if we want full
support of array formats other than ndarray in astropy, that needs more
than just having a new Quantity class that supports them. I would
suggest to make that support its own tab, which lists the reasons why
supporting other formats is useful, and then has sub-objectives that are
needed to fully do so (still objectives, not results, since interesting
on their own). Basically, those boil down to being able to replace our
current main dependencies, numpy and erfa.
For this, quantity support of anything that supports the Array API would
the first of the required sub-objectives.
A second sub-objective would be that Masked also supports such other
array types. I think that would almost certainly use the same procedure
as for Quantity, so probably fairly straightforward as a follow-up.
A third one is to move non-quantity code in astropy (e.g., stats) from
using np.XXX to xp.XXX (with xp = input_array.__array_namespace__()).
Easy in principle, likely quite a lot of work in practice, since it
also involves updating the test suites properly.
A final and arguably most difficult sub-objective would be support for
ERFA, which is required for Time and SkyCoord. The problem is that the
PyERFA wrappers are (g)ufuncs and thus only support ndarray. It seems
one would either need to rewrite ERFA's C code into python (allowing
only Array-API routines), or have more generic wrappers of the ERFA C
code (not sure to what level that's possible for GPUs). Neither is
trivial, and this may well be the trickiest part of supporting other
array formats.
Anyway, looking forward to having more discussion about that in the
coordination meeting.
All the best,
Marten
Erik Tollerud <erik.t...@gmail.com> writes:
> Dear Astropy community,
>
> The Strategic Planning Committee (SPOC) has made progress over the past year and we’re excited to tell you about some
> proposed tweaks to the Astropy Project’s roadmap process. We are open to feedback on all aspects of this plan. In one
> sentence, our proposed new process would frame roadmap items as concrete and clearly defined “Objectives” that have
> measurable results and phrased in a way to make it clear when the item has been achieved. This is inspired by the
> framework of “Objectives and key results.”
>
> With that in mind, we have reorganized the existing roadmap items in a Draft New Roadmap document. There are tabs
> and subtabs: The tabs are “Objectives” (organized by theme), taken from the prior roadmap document. Some subtabs
> include proposed metrics that could be used to measure progress on these objectives. Some prior roadmap items were
> split into multiple objectives that contain multiple metrics/results, so there is not an exact one-to-one mapping from the
> prior roadmap to the new roadmap objective items. While actionable objectives have been drafted in every category,
> more work is needed to identify realistic and meaningful target key results for them.
>
> A new aspect of this process is that we would like to have a clear timeframe in mind for objectives and measurable
> results. This will help drive prioritization based on need and opportunity (e.g., available work effort). We would like to set
> and review this prioritization with a 6 month cadence.
>
> At the Coordination Meeting, we'd like community input on:
>
>
> * 2) modifications and updates to existing roadmap items,
>
> * 3) new objectives to add, and
>
> * 4) key results to use to measure progress on the objectives.