Skip to content

Turn any AI agent into an AI Scientist. The #1 Agent Skills library for science, used by 250,000+ scientists worldwide. 177 ready-to-use validated skills plus 100+ scientific databases covering biology, chemistry, medicine, and drug discovery. Compatible with Cursor, Claude Code, Codex, Pi, Antigravity, and the open Agent Skills standard.

6.0 KB
Raw Blame
from pathlib import Path
import importlib.util
import numpy as np
import pytest
import rasterio
from rasterio.transform import from_origin
import geopandas as gpd
from shapely.geometry import box, Point

SKILL_ROOT = Path(__file__).resolve().parents[2] / 'skills' / 'geomaster'
spec = importlib.util.spec_from_file_location('geomaster_raster_workflows', SKILL_ROOT / 'scripts' / 'raster_workflows.py')
recipes = importlib.util.module_from_spec(spec)
spec.loader.exec_module(recipes)


def raster(path, data, *, crs='EPSG:32610', nodata=None):
    if data.ndim == 2:
        data = data[None]
    with rasterio.open(path, 'w', driver='GTiff', count=data.shape[0], height=data.shape[1], width=data.shape[2],
                       transform=from_origin(500000, 4200000, 10, 20), crs=crs, dtype=data.dtype, nodata=nodata) as ds:
        ds.write(data)


def test_unsigned_and_masked_difference():
    a=np.ma.array([1,3,0,5],mask=[0,0,0,1],dtype='uint16')
    b=np.array([3,1,0,4],dtype='uint16')
    np.testing.assert_allclose(recipes.normalized_difference(a,b),[-.5,.5,np.nan,np.nan],equal_nan=True)


def test_shapes_and_scales():
    with pytest.raises(ValueError): recipes.normalized_difference(np.zeros((2,1)),np.zeros(2))
    with pytest.raises(ValueError): recipes.spectral_indices(np.zeros(2),*[np.zeros(3)]*5)
    v=recipes.spectral_indices(*[np.array([x]) for x in [.1,.2,.2,.6,.3,.1]])
    assert v['NDVI'][0] == pytest.approx(.5)
    assert v['EVI'][0] == pytest.approx(2.5*.4/(.6+1.2-.75+1))
    assert v['NBR'][0] == pytest.approx(.5/.7)


@pytest.mark.parametrize('dx,dy,aspect',[(1,0,270),(0,1,180),(-1,0,90),(0,-1,0)])
def test_terrain_known_planes(dx,dy,aspect):
    rows,cols=np.mgrid[:5,:6]
    z=cols*10*dx + (-rows*20)*dy
    slope, direction, shade=recipes.terrain_metrics(z,from_origin(0,0,10,20),'EPSG:32610')
    np.testing.assert_allclose(slope,45)
    np.testing.assert_allclose(direction,aspect)
    assert np.isfinite(shade).all() and (shade>=0).all() and (shade<=1).all()


def test_flat_shade_and_mask():
    z=np.ma.array(np.zeros((5,5)),mask=False)
    z.mask[2,2]=True
    slope,aspect,shade=recipes.terrain_metrics(z,from_origin(0,0,10,20),'EPSG:32610')
    assert np.isnan(aspect).all()
    assert np.isnan(slope[2,2]) and np.isnan(slope[1,2])
    assert shade[0,0]==pytest.approx(np.sqrt(.5))


@pytest.mark.parametrize('crs,transform', [('EPSG:4326',from_origin(0,0,1,1)),('EPSG:2227',from_origin(0,0,1,1)),('EPSG:32610',rasterio.Affine(10,1,0,0,-10,0))])
def test_terrain_rejects_invalid_units_or_grid(crs,transform):
    with pytest.raises(ValueError): recipes.terrain_metrics(np.ones((2,2)),transform,crs)


def test_ndvi_roundtrip(tmp_path):
    src,dst=tmp_path/'s2.tif',tmp_path/'ndvi.tif'
    raster(src,np.array([[[2000,0],[5000,4000]],[[6000,0],[1000,4000]]],dtype='uint16'),nodata=0)
    result=recipes.write_ndvi(src,dst,red_band=1,nir_band=2,scale=.0001,offset=-.1)
    assert result[0,0]==pytest.approx(2/3)
    assert np.isnan(result[0,1]) and result[1,1]==0
    with rasterio.open(dst) as ds:
        assert ds.count==1 and ds.dtypes==('float32',)
        assert ds.crs.to_epsg()==32610 and ds.transform.e==-20
        assert ds.read(1,masked=True).mask[0,1]
    with pytest.raises(ValueError): recipes.write_ndvi(src,src,red_band=1,nir_band=2,scale=1,offset=0)


@pytest.mark.parametrize('band,scale',[(0,1),(3,1),(True,1),(1,0),(1,float('nan'))])
def test_ndvi_rejects_bad_parameters(tmp_path,band,scale):
    p=tmp_path/'in.tif';raster(p,np.ones((2,2,2),dtype='uint16'))
    with pytest.raises(ValueError): recipes.write_ndvi(p,tmp_path/'out.tif',red_band=band,nir_band=2,scale=scale,offset=0)


def training():
    return gpd.GeoDataFrame({'class_id':[1,300]},geometry=[box(500000,4199920,500020,4200000),box(500020,4199920,500040,4200000)],crs=32610)


def test_classification_preserves_large_labels_mask_and_reprojects(tmp_path):
    p=tmp_path/'in.tif';q=tmp_path/'out.tif'
    data=np.tile([10,10,100,100],(4,1)).astype('uint16');data[0,0]=0
    raster(p,data,nodata=0)
    model=recipes.classify_imagery(p,training().to_crs(4326),q)
    assert list(model.classes_)==[1,300]
    with rasterio.open(q) as ds:
        out=ds.read(1,masked=True)
        assert out.mask[0,0] and ds.nodata==0 and ds.dtypes==('uint16',)
        assert (out[:,2:]==300).all() and out[1,0]==1


@pytest.mark.parametrize('failure',['missing_crs','bad_label','overlap','no_pixels','point','empty'])
def test_classification_rejects_bad_training(tmp_path,failure):
    p=tmp_path/'in.tif';raster(p,np.ones((4,4),dtype='uint16'))
    t=training()
    if failure=='missing_crs': t=t.set_crs(None,allow_override=True)
    if failure=='bad_label': t.loc[0,'class_id']=0
    if failure=='overlap': t.loc[1,'geometry']=t.geometry.iloc[0]
    if failure=='no_pixels': t.loc[0,'geometry']=box(0,0,1,1)
    if failure=='point': t.loc[0,'geometry']=Point(500000,4200000)
    if failure=='empty': t=t.iloc[:0]
    with pytest.raises(ValueError): recipes.classify_imagery(p,t,tmp_path/'out.tif')


def test_writers_refuse_existing_output_and_same_band(tmp_path):
    src=tmp_path/'source.tif';dst=tmp_path/'output.tif'
    raster(src,np.ones((2,4,4),dtype='uint16'))
    dst.write_bytes(b'preserve existing result')
    with pytest.raises(ValueError,match='already exists'):
        recipes.write_ndvi(src,dst,red_band=1,nir_band=2,scale=1,offset=0)
    with pytest.raises(ValueError,match='already exists'):
        recipes.classify_imagery(src,training(),dst)
    assert dst.read_bytes()==b'preserve existing result'
    with pytest.raises(ValueError,match='distinct'):
        recipes.write_ndvi(src,tmp_path/'new.tif',red_band=1,nir_band=1,scale=1,offset=0)


def test_writers_refuse_hardlink_alias(tmp_path):
    import os
    src=tmp_path/'source.tif';alias=tmp_path/'alias.tif'
    raster(src,np.ones((2,4,4),dtype='uint16')); before=src.read_bytes()
    os.link(src,alias)
    with pytest.raises(ValueError): recipes.write_ndvi(src,alias,red_band=1,nir_band=2,scale=1,offset=0)
    with pytest.raises(ValueError): recipes.classify_imagery(src,training(),alias)
    assert src.read_bytes()==before