Fix AttributeError in facebudget when front is omitted (fixes #1097)

[gaoflow]

Summary

imod.evaluate.facebudget documents and enforces that at least one of
front, lower, right must be supplied (it raises ValueError if all three
are None). However two code paths dereferenced front unconditionally, so
omitting front crashed:

def facebudget(budgetzone, front=None, lower=None, right=None, netflow=True):
    ...
    if "time" in dims:
        for itime in range(front.coords["time"].size):   # front may be None
    ...
    else:                                                 # no zones defined
        dask_front = dask.array.full(front.shape, ...)    # front may be None
        dask_lower = dask.array.full(lower.shape, ...)    # lower may be None
        dask_right = dask.array.full(right.shape, ...)    # right may be None

So a perfectly valid call such as facebudget(budgetzone, lower=lower) raised:

AttributeError: 'NoneType' object has no attribute 'coords'   # transient case
AttributeError: 'NoneType' object has no attribute 'shape'    # no-zones case

This is exactly the null-dereference Coverity flagged in #1097
(REVERSE_INULL: front is dereferenced on all paths before the
if front is not None guard inside the loop). The other location named in
#1097 (package.py:292) no longer exists — that code was refactored into
imod/evaluate/budget.py, which is the surviving occurrence fixed here.

Fix

Pick the first supplied DataArray as a reference for the time axis and the
output shape. All three inputs are already validated to share budgetzone's
spatial shape, so any of them is a valid reference:

reference = next(da for da in (front, lower, right) if da is not None)
...
for itime in range(reference.coords["time"].size):
...
dask_front = dask.array.full(reference.shape, np.nan, chunks=chunks)
dask_lower = dask.array.full(reference.shape, np.nan, chunks=chunks)
dask_right = dask.array.full(reference.shape, np.nan, chunks=chunks)

Reproduction / verification

import numpy as np, xarray as xr, imod
da = xr.DataArray(np.zeros((3, 2, 1)),
                  {"layer": [1, 2, 3], "y": [0.5, 1.5], "x": [0.5]},
                  ("layer", "y", "x"))
lower = xr.full_like(da, 0.0); lower[:2] = 1.0
zone = xr.full_like(da, np.nan); zone[:2] = 1
imod.evaluate.facebudget(zone, lower=lower)   # before: AttributeError; after: works

Added test_omit_front and test_omit_front_transient to
imod/tests/test_evaluate/test_budget.py. The transient test exercises both
crash sites (the time loop and the no-zones branch) via a partial zone and an
all-inactive zone. Verified locally: both tests fail on master
(AttributeError) and pass with this change; all pre-existing budget tests
(9) still pass.

Fixes #1097.

[JoerivanEngelen]

@gaoflow

Looks good to me, could you also add your fix to the changelog? Write about it with users in mind, who are not software developers.

[gaoflow]

Done — added a Fixed entry to the changelog under [Unreleased], phrased for users: it notes that facebudget previously errored when you left out front, even though only one of front/lower/right is required, and that omitting front now works as documented.

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud