Adds `conftest.py` of common fixtures for integration tests by spjuhel · Pull Request #1273 · CLIMADA-project/climada_python

spjuhel · 2026-03-26T10:50:52Z

This is based on data test used for the integration tests of the static and interpolated trajectories.

The objective is to provide:

"as comprehensible as possible and as minimalistic as possible data (exposure, hazard, impfset)"
Factories

Changes proposed in this PR:

Adds fixtures and factories for minimalistic objects for integration testing
Adds an quick example of how to use them

We need to decide if this is something we actually want.
Otherwise, I will move the content to the trajectory integration test directly.

PR Author Checklist

PR Reviewer Checklist

…or-common-fixtures

spjuhel · 2026-03-26T15:16:41Z

Notes from discussion:

shift the examples to a (possibly existing) page in the documentation

spjuhel · 2026-03-31T09:34:34Z

Note for this PR (or rather for future improvements):

Have also fixtures with the DEMO files
Plan (a hackathon?) to move tests to use the fixtures

spjuhel · 2026-03-31T11:49:13Z

I've added a documentation page,
I chose to make a new one because direct integration in the "Writing tests" one did not feel right, but I'm happy to reconsider. Note that there are two mentions/references in the "Writing tests" one!

I tried to find the balance between being explicit and concise (had help from Claude to be honest).

I am wondering about putting actual values in the documentation as this implies that we keep them up to date with the file (although the values should technically not change).

Quick access: https://climada-python--1273.org.readthedocs.build/en/1273/development/Guide_test_fixtures.html

peanutfun

This is a good first step towards using common fixtures for all our unit tests! I found a few issues in the code. Other than that, I think we should add an API reference for fixtures to the docs (this might require loading the conftest.py as well. Not sure how to do that atm).

I think most of the functions defined here should not be fixtures because they are not intended for use outside this file. The only "real" fixtures I see:

exposures_factory
exposures
hazard_factory
hazard
impf_factory
linear_impact_function
impfset_factory
impfset

Comments on the tutorial:

Add a target at the top of the new tutorial file and use it for linking with a ref directive. This will keep links valid if file paths change. See https://myst-parser.readthedocs.io/en/latest/syntax/cross-referencing.html https://myst-parser.readthedocs.io/en/latest/syntax/cross-referencing.html#reference-roles
To make the advantage of a fixture more obvious, add another test to the example that uses sample_user
Should be attention: "Never import fixtures directly" https://www.sphinx-doc.org/en/master/usage/restructuredtext/directives.html#directive-attention
For use of directives in MyST, see https://myst-parser.readthedocs.io/en/latest/syntax/roles-and-directives.html#directives-a-block-level-extension-point
Remove explicit information on fixture contents, like "There are 4 events..." (should be in docstring, as mentioned above)
"Tip" should use tip, "Note" should use note directive https://www.sphinx-doc.org/en/master/usage/restructuredtext/directives.html#directive-tip
Errors:
- "Usefull references"
- from climada.entity.exposures ?
- stronger_hazard = ...; # Intensities are halved ... 😂

peanutfun · 2026-04-02T12:37:25Z

climada/test/conftest.py

+# Sanity checks
+for const in [VALUES, CATEGORIES, EXP_LONS, EXP_LATS]:
+    assert len(const) == len(
+        VALUES
+    ), "VALUES, REGIONS, CATEGORIES, EXP_LONS, EXP_LATS should all have the same lengths."
+
+for const in [EVENT_IDS, EVENT_NAMES, DATES, FREQUENCY]:
+    assert len(const) == len(
+        EVENT_IDS
+    ), "EVENT_IDS, EVENT_NAMES, DATES, FREQUENCY should all have the same lengths."


If these assertions fail, importing this file will result in an error. I think this is dangerous, as the resulting errors are hard to understand an occur at weird points in time. Are those assertions needed at all? Shouldn't the constructed objects like Exposures complain?

peanutfun · 2026-04-02T12:41:57Z

climada/test/conftest.py

+    ), "EVENT_IDS, EVENT_NAMES, DATES, FREQUENCY should all have the same lengths."
+
+
+@pytest.fixture(scope="session")


Other scopes than function are primarily intended to avoid reconstructing "expensive" objects. The way I see it, this applies to none of the objects here and can have serious side effects, as modifications will persist throughout the test session. I suggest using the default (smallest) scope everywhere, at least in this file.

peanutfun · 2026-04-02T12:43:00Z

climada/test/conftest.py

+
+
+@pytest.fixture(scope="session")
+def exposure_values():


Do you foresee this being used as a fixture to an actual test? I don't, so I would keep this a "regular" function and only declare fixtures that are intended for use.

peanutfun · 2026-04-02T12:44:44Z

climada/test/conftest.py

+A set of reusable fixtures for testing purpose.
+
+The objective of this file is to provide minimalistic, understandable and consistent
+default objects for unit and integration testing.
+
+Values are chosen such that:
+    - Exposure value of the first points is 0. (First location should always have 0 impacts)
+    - Category / Group id of all points is 1, except for third point, valued at 2000 (Impacts on that category are always a share of 2000)
+    - Hazard centroids are the exposure centroids shifted by `HAZARD_JITTER` on both lon and lat.
+    - There are 4 events, with frequencies == 0.03, 0.01, 0.006, 0.004, 0,
+      such that impacts for RP250, 100 and 50 and 20 are at_event,
+      (freq sorted cumulate to 1/250, 1/100, 1/50 and 1/20).
+    - Hazard intensity is:
+        * Event 1: zero everywhere (always no impact)
+        * Event 2: max intensity at first centroid (also always no impact (first centroid is 0))
+        * Event 3: half max intensity at second centroid (impact == half second centroid)
+        * Event 4: quarter max intensity everywhere (impact == 1/4 total value)
+        * Event 5: max intensity everywhere (but zero frequency)
+      With max intensity set at 100
+    - Impact function is the "identity function", x intensity is x% damages
+    - Impact values should be:
+        * AAI = 18 = 1000*1/2*0.006+(1000+2000+3000+4000+5000)*0.25*0.004
+        * RP20 = event1 = 0
+        * RP50 = event2 = 0
+        * RP100 = event3 = 500 = 1000*1/2
+        * RP250 = event4 = 3750 = (1000+2000+3000+4000+5000)*0.25


Fixtures are Python functions, so I suggest adding this information to fixture docstrings and creating an API reference. Individual fixtures can be documented with autofunction. See https://docs.pytest.org/en/stable/reference/reference.html#fixtures and https://github.com/pytest-dev/pytest/blob/f93656d611f3c9cd22f346dfeda52cddb778c589/doc/en/reference/reference.rst?plain=1#L325

We should be able to import docstrings from conftest in Sphinx, as the test code is part of the climada package. However, imports may fail (since packages for testing are required).

I asked ChatGPT and it said this:

peanutfun · 2026-04-02T12:49:29Z

climada/test/conftest.py

+
+@pytest.fixture(scope="session")
+def exposure_geometry():
+    return [Point(lon, lat) for lon, lat in zip(EXP_LONS, EXP_LATS)]


https://geopandas.org/en/stable/docs/reference/api/geopandas.points_from_xy.html#geopandas.points_from_xy ?

peanutfun · 2026-04-02T12:51:24Z

climada/test/conftest.py

+
+# Frequency are choosen so that they cumulate nicely
+# to correspond to 250, 100, 50, and 20y return periods (for impacts)
+FREQUENCY = np.array([0.03, 0.01, 0.006, 0.004, 0.0])


Do the math for clarity?

Suggested change

FREQUENCY = np.array([0.03, 0.01, 0.006, 0.004, 0.0])

FREQUENCY = np.array([1/20, 1/50, 1/100, 1/200, 0.0, dtype="float"])

peanutfun · 2026-04-02T12:52:57Z

climada/test/conftest.py

+
+# Exposure coordinates
+EXP_LONS = np.array([4, 4.25, 4.5, 4, 4.25, 4.5])
+EXP_LATS = np.array([45, 45, 45, 45.25, 45.25, 45.25])


Hard to read. As values don't matter, choose digits that are easier to distinguish (also against the LON values)

Suggested change

EXP_LATS = np.array([45, 45, 45, 45.25, 45.25, 45.25])

EXP_LATS = np.array([33, 33, 33, 33.25, 33.25, 33.25])

peanutfun · 2026-04-02T12:59:02Z

climada/test/conftest.py

+EXPOSURE_REF_YEAR = 2020
+EXPOSURE_VALUE_UNIT = "USD"
+VALUES = np.array([0, 1000, 2000, 3000, 4000, 5000])
+CATEGORIES = np.array([1, 1, 2, 1, 1, 3])


Good first step, but I think we need to parameterize asap: Use multiple CRS, build exposures of size 1 and >1, etc.

spjuhel added 2 commits February 11, 2026 11:31

adds conftest and how to use file

0cda1e2

update

648a00e

spjuhel requested review from chahank, emanuel-schmid and peanutfun as code owners March 26, 2026 10:50

Merge remote-tracking branch 'origin/develop' into feature/conftest-f…

4d2a9c3

…or-common-fixtures

fixture tutorial, first iteration

23f9e9b

Clearer values in tutorial

890c020

peanutfun requested changes Apr 2, 2026

View reviewed changes

		), "EVENT_IDS, EVENT_NAMES, DATES, FREQUENCY should all have the same lengths."


		@pytest.fixture(scope="session")

	FREQUENCY = np.array([0.03, 0.01, 0.006, 0.004, 0.0])
	FREQUENCY = np.array([1/20, 1/50, 1/100, 1/200, 0.0, dtype="float"])

	EXP_LATS = np.array([45, 45, 45, 45.25, 45.25, 45.25])
	EXP_LATS = np.array([33, 33, 33, 33.25, 33.25, 33.25])

Conversation

spjuhel commented Mar 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

PR Author Checklist

PR Reviewer Checklist

Uh oh!

spjuhel commented Mar 26, 2026

Uh oh!

spjuhel commented Mar 31, 2026

Uh oh!

spjuhel commented Mar 31, 2026

Uh oh!

peanutfun left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

spjuhel commented Mar 26, 2026 •

edited

Loading

peanutfun left a comment •

edited

Loading