Open gpfreitas opened 6 years ago
@gbrener I can work with you on the docs overhaul here and elsewhere. I'll start with the notebooks in xarray_filters and see how those may be adjusted as needed for auto-docs generation / testing as described in Elm issue 188 (Documentation structure overhaul).
@PeterDSteinberg , do we want to keep the docs independent between elm
, earthio
, and xarray_filters
, or do we want to keep them all consolidated to the elm
repo? If we plan on making each repo independent, it might make sense to keep them separate, but there might also be a convenience aspect to having everything consolidated on ensemble-learning-models.readthedocs.io.
That's a good question, @gbrener - I'm not sure what is optimal approach on docs (separate docs for 3 repos vs one docs page). My thought at this point is we should have 3 separate documentation pages, if that can be done somewhat easily with the auto-docs / notebook testing harness we are developing. So each repo has its own docs folder, but the Elm docs link to xarray_filters and Earthio as needed. For the near term until December primarily focus on usages of Elm that work with xarray_filters. Later in Phase II, we can work on more Elm docs that show machine learning for non-Earth science data structures. More complete explanations and descriptions of use cases can be given in Elm docs and if xarray_filters/earthio are more of an API-description than complete story, I think that is fine for a while.
Earthio - I think the docs for the next 6+ months should treat Earthio as mainly an environment that installs "everything" needed for climate/satellite data. It would be nice if earthio issue 31 could be resolved in a way that effectively moves the functionality of what is now load_array
to xarray
, using optional package dependencies if necessary. I think that would be nice because it would simplify the overall documentation story for Earthio and we can explain Earthio is an environment one can choose to use with Elm optionally and that later Earthio will be developed as an Earth Data Catalog.
Best way to move forward between now and November 14 is to
./notebooks
directory are a starting point for the docs.
Regarding docstrings and how many / much of them to make, I think we should follow the dask level of docstring usage, where some internal functions have shorter docstrings than public interface functions.
Link to the Elm-docs-overhaul issue.