Update failing doctests with upstream dataset changes

[seisman]

Description of proposed changes

Address #2961 (comment).

[seisman]

There are still two failures:

===================================FAILURES===================================_______________ [doctest] pygmt.datasets.tile_map.load_tile_map________________105Raises106------107ImportError108If``contextily``isnotinstalledorcan'tbeimported. Follow109 :doc:`install instructions for contextily <contextily:index>`, (e.g.
110via``python-mpipinstallcontextily``) beforeusingthisfunction.
111112Examples113--------114>>>importcontextilyUNEXPECTEDEXCEPTION: ModuleNotFoundError("No module named 'contextily'")
Traceback (mostrecentcalllast):
File"/home/runner/micromamba/envs/pygmt/lib/python3.9/doctest.py", line1334, in__runexec(compile(example.source, filename, "single",
File"<doctest pygmt.datasets.tile_map.load_tile_map[0]>", line1, in<module>ModuleNotFoundError: Nomodulenamed'contextily'/home/runner/work/pygmt/pygmt/pygmt/datasets/tile_map.py:114: UnexpectedException__________________ [doctest] pygmt.src.grdproject.grdproject___________________179>>>importpygmt180>>># Load a grid of @earth_relief_30m data, with a longitude range of181>>># 10° E to 30° E, and a latitude range of 15° N to 25° N182>>>grid=pygmt.datasets.load_earth_relief(
183 ... resolution="30m", region=[10, 30, 15, 25]
184 ... )
185>>># Create a new grid from the input grid, set the projection to186>>># Mercator, and set inverse to "True" to change from "geographic"187>>># to "rectangular"188>>>new_grid=pygmt.grdproject(grid=grid, projection="M10c", inverse=True)
Error: TEDEXCEPTION: GMTCLibError("Module 'grdproject' failed with status code 72:\ngrdproject [ERROR]: Option -J: If map width is given you must also specify a region with -R")
Traceback (mostrecentcalllast):
File"/home/runner/micromamba/envs/pygmt/lib/python3.9/doctest.py", line1334, in__runexec(compile(example.source, filename, "single",
File"<doctest pygmt.src.grdproject.grdproject[2]>", line1, in<module>File"/home/runner/work/pygmt/pygmt/pygmt/helpers/decorators.py", line603, innew_modulereturnmodule_func(*args, **kwargs)
File"/home/runner/work/pygmt/pygmt/pygmt/helpers/decorators.py", line776, innew_modulereturnmodule_func(*bound.args, **bound.kwargs)
File"/home/runner/work/pygmt/pygmt/pygmt/src/grdproject.py", line122, ingrdprojectlib.call_module(
File"/home/runner/work/pygmt/pygmt/pygmt/clib/session.py", line624, incall_moduleraiseGMTCLibError(
pygmt.exceptions.GMTCLibError: Module'grdproject'failedwithstatuscode72:
Error: ect [ERROR]: Option-J: Ifmapwidthisgivenyoumustalsospecifyaregionwith-R/home/runner/work/pygmt/pygmt/pygmt/src/grdproject.py:188: UnexpectedException-----------------------------Capturedstderrcall-----------------------------Error: ect [ERROR]: Option-J: Ifmapwidthisgivenyoumustalsospecifyaregionwith-R

[seisman]

The pygmt.datasets.tile_map.load_tile_map failure is because contextily is an optional dependency and is not installed in the Python 3.9 job run. It also means we can't run full tests in the Tests workflow, but fortunately we can still run them in GMT Legacy Tests and Dev Tests.

The pygmt.src.grdproject.grdproject is likely an upstream bug, reported in GenericMappingTools/gmt#8273.

[weiji14]

The pygmt.datasets.tile_map.load_tile_map failure is because contextily is an optional dependency and is not installed in the Python 3.9 job run. It also means we can't run full tests in the Tests workflow, but fortunately we can still run them in GMT Legacy Tests and Dev Tests.

This doctest should be skipped with the __doctest_requires__ = {("load_tile_map"): ["contextily"]} line (see https://github.com/scientific-python/pytest-doctestplus/tree/v1.1.0#doctest-dependencies and #2125 (comment)). I thought we had pytest-doctestplus installed here:

pygmt/.github/workflows/ci_doctests.yaml

Line 68 in 2d61b54

pytest-doctestplus

Is that doctest still being run somehow?

[seisman]

The pygmt.datasets.tile_map.load_tile_map failure is because contextily is an optional dependency and is not installed in the Python 3.9 job run. It also means we can't run full tests in the Tests workflow, but fortunately we can still run them in GMT Legacy Tests and Dev Tests.

This doctest should be skipped with the __doctest_requires__ = {("load_tile_map"): ["contextily"]} line (see https://github.com/scientific-python/pytest-doctestplus/tree/v1.1.0#doctest-dependencies and #2125 (comment)). I thought we had pytest-doctestplus installed here:

pygmt/.github/workflows/ci_doctests.yaml

Line 68 in 2d61b54

pytest-doctestplus

Is that doctest still being run somehow?

The doctest will be skipped only if the --doctest-plus option is given. So, the doctest is skipped for make test but will be run for make fulltest (see

pygmt/Makefile

Line 45 in 621d6da

test: PYTEST_ARGS=--doctest-plus $(PYTEST_COV_ARGS) ${PYTEST_EXTRA}

).

[weiji14]

The doctest will be skipped only if the --doctest-plus option is given. So, the doctest is skipped for make test but will be run for make fulltest (see

pygmt/Makefile

Line 45 in 621d6da

test: PYTEST_ARGS=--doctest-plus $(PYTEST_COV_ARGS) ${PYTEST_EXTRA}

Hmm, you mentioned at #1790 (comment) that we were using --doctest-modules --doctest-plus for make fulltest, but looking at the code, it looks like only --doctest-modules is applied (from pyproject.toml).

It also means we can't run full tests in the Tests workflow, but fortunately we can still run them in GMT Legacy Tests and Dev Tests.

Do we need to run the full tests on GMT Legacy/Dev? The ci_doctests.yaml should cover the doctests, and ci_tests.yaml should cover the non-doctest tests already right?

[seisman]

The doctest will be skipped only if the --doctest-plus option is given. So, the doctest is skipped for make test but will be run for make fulltest (see

pygmt/Makefile

Line 45 in 621d6da

test: PYTEST_ARGS=--doctest-plus $(PYTEST_COV_ARGS) ${PYTEST_EXTRA}

Hmm, you mentioned at #1790 (comment) that we were using --doctest-modules --doctest-plus for make fulltest, but looking at the code, it looks like only --doctest-modules is applied (from pyproject.toml).

Hmm, I think the description in #1790 (comment) is incorrect. make test uses --doctest-modules --doctest-plus and make fulltest uses --doctest-modules. I've updated the comment.

It also means we can't run full tests in the Tests workflow, but fortunately we can still run them in GMT Legacy Tests and Dev Tests.

Do we need to run the full tests on GMT Legacy/Dev? The ci_doctests.yaml should cover the doctests, and ci_tests.yaml should cover the non-doctest tests already right?

You're right, I almost forgot this workflow. I was looking at the checklist at #2843, which only mentions "GMT Legacy" and "GMT Tests" but not "Doctests".

I think we should mention the "DocTests" workflow in both checklists (#2843 and #2961).

[weiji14]

It also means we can't run full tests in the Tests workflow, but fortunately we can still run them in GMT Legacy Tests and Dev Tests.

Do we need to run the full tests on GMT Legacy/Dev? The ci_doctests.yaml should cover the doctests, and ci_tests.yaml should cover the non-doctest tests already right?

You're right, I almost forgot this workflow. I was looking at the checklist at #2843, which only mentions "GMT Legacy" and "GMT Tests" but not "Doctests".

Yeah, could you trigger the "Doctests" workflow in this PR? I did try it locally and it seems to work fine without contextily if you add --doctest-plus here:

pygmt/Makefile

Lines 52 to 54 in 621d6da

	# run doctests only
	doctest: PYTEST_ARGS=--ignore=../pygmt/tests ${PYTEST_EXTRA}
	doctest: _runtest

I think we should mention the "DocTests" workflow in both checklists (#2843 and #2961).

Sounds good.

[weiji14]

Yeah, could you trigger the "Doctests" workflow in this PR? I did try it locally and it seems to work fine without contextily if you add --doctest-plus here:

Owh, but using --doctest-plus only runs 34 tests, compared to 74 doctests if using only --doctest-modules. Maybe don't add that --doctest-plus line to make doctest then, as we agreed on in #1790 (comment).

[weiji14]

Oh wow, did we write the wrong description before. Default (when inverse=False) should have been "geographic" to "rectangular" 😅