[bhuvan-somisetty]

Fixes #1890

Summary of Changes

This PR fixes issues in r.local.relief in bicubic spline interpolation mode (-v) and when generating shaded relief output:

1. Fix TypeError: 'dict_keys' object is not subscriptable and small-unit spline_step estimation:

   • The spline step was taken by indexing dict.keys() from gs.parse_command("v.surf.bspline", flags="e", ...), which fails in Python 3. Parsing that output is also fragile because v.surf.bspline -e prints a translatable message, not key=value pairs.
   • The spline step is now computed directly from gs.vector_info() of the contour points as sqrt(extent_area / n_points) * 2 (falling back to the computational region area if the contour extent area is zero), matching v.surf.bspline -e (P_estimate_splinestep). The module fails with a clear message if there are no contour points.
   • Removed round() from the spline_step calculation so small-unit projects (such as latitude-longitude in degrees) do not round spline_step down to 0.

2. Pass computed spline_step to v.surf.bspline:

   • The spline step was computed and reported with gs.info, but never passed to v.surf.bspline. Added ew_step=spline_step and ns_step=spline_step.

3. Fix color table name in shaded output:

   • Changed the default color table from "difference" to "differences", the valid GRASS color table name.

4. Add tests:

   • Added testsuite/test_r_local_relief.py covering the default r.fillnulls mode, -v bspline mode, and shaded output.
   • Added tests/r_local_relief_latlon_test.py verifying -v bspline mode in a latitude-longitude project.

Signed-off-by: bhuvan-somisetty somisettybhuvan5@gmail.com

[wenzeslaus]

G_message are user messages, so they may be translated. This then fails. Other options?

[bhuvan-somisetty]

Good catch. Replaced it with v.info -t and gs.region() to compute the mean distance directly (sqrt(area / points)), so it no longer relies on parsing localized messages.

[wenzeslaus]

Please, keep the PR desc in sync with the code.

[wenzeslaus]

This is low quality code

[wenzeslaus]

Overly complicated and a general exception. See gs.vector_info_topo

[bhuvan-somisetty]

thanks for the review. cleaned up the spline step part, it now just uses gs.vector_info on the contour points and computes sqrt(extent area / points) * 2, same as what v.surf.bspline -e does internally. no more broad except or hardcoded fallbacks, it fails with a clear message if there are no contour points. also updated the PR description to match the current code.

[wenzeslaus]

With round() the spline step is 0 when the mean point distance is below 0.5 map units, e.g., in a latitude-longitude project, and v.surf.bspline then never finishes. Please add a test for this, such as the one below (pytest, tests/r_local_relief_latlon_test.py), and make it pass.

"""Test r.local.relief in a latitude-longitude project"""

import os

import pytest

import grass.script as gs
from grass.tools import Tools


@pytest.fixture
def latlon_session(tmp_path):
    """Latitude-longitude project with a synthetic wavy elevation raster"""
    project = tmp_path / "project"
    gs.create_project(project, epsg="4326")
    with gs.setup.init(project, env=os.environ.copy()) as session:
        tools = Tools(session=session)
        tools.g_region(n=36, s=35.95, w=-79, e=-78.95, res="0:00:03")
        tools.r_mapcalc(
            expression=(
                "elevation = 100 + 20 * sin(row() * 12) * cos(col() * 9)"
                " + 3 * sin(row() * 47 + col() * 31)"
            )
        )
        yield session


def test_bspline_latlon(latlon_session):
    """Check that -v gives a full local relief raster when map units are degrees"""
    tools = Tools(session=latlon_session)
    tools.r_local_relief(
        input="elevation", output="local_relief", neighborhood_size=5, flags="v"
    )
    stats = tools.r_univar(map="local_relief", format="json")
    assert stats["null_cells"] == 0
    assert stats["min"] < 0 < stats["max"]

[bhuvan-somisetty]

Good catch, removed round() so small units like degrees don't collapse to 0 and added tests/r_local_relief_latlon_test.py. Updated the PR description as well.