Implement the component

[mdpiper]

I'd like to make the planning and development of this project as open and as public as possible. GitHub provides a good place for this.

Project planning is described in #1. This is a record of what I actually did.

Tasks

• Create GitHub repository (csdms/bmi-geotiff) with README and license
• Make directories: bmi_geotiff, tests, examples, docs
• Collect build files (copy from bmi-topography and tailor, using PyPA guidelines)
  • pyproject.toml
  • setup.cfg (include flake8 and zest.releaser config because they don't yet work in pyproject.toml)
  • Makefile
  • environment.yaml
  • autogenerate a setup.py file through my Makefile
• Set up package definition file and iteratively write library (io.py) and tests (test_io.py)
  • make test
  • make pretty
• Make example for library
• BMI the library (bmi.py)
  • generate skeleton with bmipy-render
  • fill in the methods of BmiGeoTiff
  • make a config file (place in examples and link in tests/data)
  • test iteratively with bmi-tester (make bmi-test)
• Make example for BMI
• Add content to README
  • overview -- uses experimental xarray.open_rasterio
  • install instructions
  • example(s)
  • acknowledgment
• Add package metadata files (copy from bmi-topography and modify)
  • CHANGES.md
  • CONTRIBUTING.md
  • use setuptools-scm to include files in the sdist instead of using a MANIFEST.in file
• Generate distribution archives (make build)
  • check content in the dist directory
  • test upload to test.pypi.org (make testpypi)
• Create v0.1 release with zest.releaser (make fullrelease)
  • upload to PyPI
• Write documentation
  • run sphinx-quickstart from the docs directory
  • include README
  • build API docs
  • automate versioning for docs
  • deploy to Read the Docs
    • use .readthedocs.yaml
• Format and lint (make pretty and make lint)
• CI with GitHub Actions
  • build and test
  • format
  • lint
  • build docs
• Create v0.2 release and upload to PyPI
• Make a conda-forge package
  • use PyPI for source tarball instead of GitHub
• Add a DOI through Zenodo
  • turn repo "on" in Zenodo
  • make a release from a tag on GitHub
  • edit title and license on Zenodo
  • make a CITATION document
• Babelize
• Make a pymt package (pymt_geotiff)
  • add component metadata
  • include script and notebook examples
  • create brief documentation (including notebook)
  • deploy docs to Read the Docs (use .readthedocs.yaml)
  • make a conda-forge package
• Add to CSDMS Model Repository (this will populate Data Components page)

[mdpiper]

It's tricky to get the crs using get_value:

>>> var_name = "coordinate_reference_system__string"
>>> nbytes = bmi.get_var_nbytes(var_name)
>>> dest = np.ndarray((1,), dtype="U{}".format(nbytes))
>>> r = bmi.get_value(var_name, dest)

>>> print(r)
['+init=epsg:32618']

>>> print(dest)
['+init=epsg:32618']

[mdpiper]

matplotlib,pyplot.imshow can display an RGB image, but it needs band sequential ordering:

rasterBSQ = np.transpose(rasterRGB, axes=[1, 2, 0])
plt.imshow(rasterBSQ)

[mdpiper]

I'm using setuptools-scm to include repository files outside the package, like Makefile, CHANGES.md, etc. It conflicts slightly with zest.releaser, which shows a dirty repository when running fullrelease:

INFO: Starting release.
Tag needed to proceed, you can use the following command:
git tag v0.2.dev5+ged6292c -m 'Tagging 0.2.dev5+ged6292c'
Run this command (Y/n)? n

The workaround is to tag the repo before running fullrelease:

$ git tag v0.2 -m 'Tagging 0.2'

[mdpiper]

Release text (using v0.2 tag):

CSDMS GeoTIFF data component

The public release of the GeoTIFF data component, with examples, CI, docs, and installation through pip and conda.