Harden v0.1 diagnostics and policy

.geodebug.toml.example

@@ -0,0 +1,18 @@
+# GeoDebug project policy.
+profile = "default"
+fail_on = "error"
+
+[rules]
+disable = []
+
+# Explicit severity overrides take precedence over the selected profile.
+[rules.severity]
+# GEO101 = "error"
+
+# Suppressions require an explanation. An optional expiry date keeps temporary
+# exceptions from becoming permanent policy.
+# [[suppress]]
+# rule = "GEO103"
+# path = "legacy/*.geojson"
+# reason = "Known upstream coordinate convention"
+# expires = 2027-01-01

README.md

@@ -38,7 +38,7 @@ Run dataset checks:
 
 ```bash
 geodebug check roads.geojson
-geodebug check roads.geojson --deep
+geodebug check landcover.tif --deep
 ```
 
 Compare two datasets:
@@ -47,29 +47,68 @@ Compare two datasets:
 geodebug compare dem-a.tif dem-b.tif
 ```
 
+Check an operation before execution:
+
+```bash
+geodebug preflight roads.geojson --operation buffer --distance 500
+geodebug preflight parcels.geojson --operation area
+```
+
 Emit the canonical JSON report:
 
 ```bash
 geodebug check roads.geojson --format json
 ```
 
-The default path is metadata-first. Full vector geometry scans are opt-in with `--deep`.
+The default path is metadata-first. Full geometry or raster scans are opt-in with `--deep`.
+
+## Project policy
+
+GeoDebug searches the current directory and its parents for `.geodebug.toml`. An explicit
+config path can be supplied with `--config`.
+
+```toml
+profile = "default"
+fail_on = "error"
+
+[rules]
+disable = []
+
+[rules.severity]
+GEO101 = "error"
+
+[[suppress]]
+rule = "GEO103"
+path = "legacy/*.geojson"
+reason = "Known upstream coordinate convention"
+expires = 2027-01-01
+```
+
+Profiles do not change rule truth values. `strict` promotes warnings to errors;
+`exploratory` demotes warnings to notes. Explicit severity overrides take precedence.
+Suppressions require a reason and may expire.
 
 ## Built-in rules
 
-The v0.1 usable core currently includes:
+The current v0.1 rule set includes:
 
 - `GEO101` — missing CRS
+- `GEO103` — geographic coordinate out of range
+- `GEO105` — projected data outside CRS area of use
 - `GEO201` — invalid geometry
+- `GEO301` — invalid or singular raster transform
+- `GEO304` — NoData value conflicts with the validity mask
 - `GEO402` — no spatial overlap
 - `GEO404` — raster grid misalignment
 - `GEO501` — metric buffer on a geographic CRS
+- `GEO502` — planar area on a geographic CRS
+- `GEO503` — planar distance or length on a geographic CRS
 
 Inspect rule contracts with:
 
 ```bash
 geodebug rules list
-geodebug rules show GEO404
+geodebug rules show GEO304
 ```
 
 ## Architecture
@@ -94,3 +133,6 @@ ruff check .
 mypy src/geodebug
 pytest
 ```
+
+Golden rule cases live in `tests/golden/cases.toml`. Each case declares both expected
+diagnostics and rules that must not be reported.

src/geodebug/__init__.py

@@ -1,5 +1,13 @@
-from geodebug.api import check, compare, evaluate, inspect
+from geodebug.api import check, compare, evaluate, inspect, preflight
 from geodebug.models.report import Report
 from geodebug.version import __version__
 
-__all__ = ["Report", "__version__", "check", "compare", "evaluate", "inspect"]
+__all__ = [
+    "Report",
+    "__version__",
+    "check",
+    "compare",
+    "evaluate",
+    "inspect",
+    "preflight",
+]

src/geodebug/adapters/rasterio.py

@@ -15,12 +15,13 @@
     RASTER_DTYPES,
     RASTER_HEIGHT,
     RASTER_NODATA,
+    RASTER_NODATA_COLLISION_BANDS,
     RASTER_RESOLUTION,
     RASTER_TRANSFORM,
     RASTER_WIDTH,
     SPATIAL_BOUNDS,
 )
-from geodebug.models.enums import SubjectKind
+from geodebug.models.enums import FactState, SubjectKind
 from geodebug.models.facts import FactProvenance, FactRecord, FactStore
 from geodebug.models.subjects import DatasetSnapshot, SubjectRef
 
@@ -35,7 +36,6 @@ def supports(self, target: Any) -> int:
         return 90 if Path(target).suffix.casefold() in self._suffixes else 0
 
     def inspect(self, target: Any, options: InspectOptions) -> DatasetSnapshot:
-        del options
         try:
             import rasterio
         except ImportError as exc:
@@ -97,6 +97,13 @@ def inspect(self, target: Any, options: InspectOptions) -> DatasetSnapshot:
                         ),
                     )
                 )
+                records.append(
+                    _nodata_collision_fact(
+                        dataset,
+                        deep=options.deep,
+                        provenance=provenance,
+                    )
+                )
         except Exception as exc:
             raise AdapterError(f"unable to inspect raster dataset: {path}") from exc
 
@@ -110,3 +117,42 @@ def inspect(self, target: Any, options: InspectOptions) -> DatasetSnapshot:
             ),
             facts=FactStore(records),
         )
+
+
+def _nodata_collision_fact(
+    dataset: Any,
+    *,
+    deep: bool,
+    provenance: FactProvenance,
+) -> FactRecord:
+    nodata = dataset.nodata
+    if nodata is None:
+        return FactRecord(
+            key=RASTER_NODATA_COLLISION_BANDS,
+            state=FactState.NOT_APPLICABLE,
+            provenance=provenance,
+        )
+    if not deep:
+        return FactRecord.unknown(
+            RASTER_NODATA_COLLISION_BANDS,
+            provenance=provenance,
+        )
+
+    collisions: list[int] = []
+    for band_index in range(1, dataset.count + 1):
+        found = False
+        for _, window in dataset.block_windows(band_index):
+            values = dataset.read(band_index, window=window)
+            valid_mask = dataset.dataset_mask(window=window) != 0
+            nodata_values = values != values if nodata != nodata else values == nodata
+            if (nodata_values & valid_mask).any():
+                found = True
+                break
+        if found:
+            collisions.append(band_index)
+
+    return FactRecord.known(
+        RASTER_NODATA_COLLISION_BANDS,
+        tuple(collisions),
+        provenance=FactProvenance(origin="rasterio", method="deep.nodata-mask-scan"),
+    )

src/geodebug/api.py

@@ -1,5 +1,6 @@
 from __future__ import annotations
 
+from collections.abc import Mapping
 from typing import Any
 
 from geodebug.adapters.base import InspectOptions
@@ -9,6 +10,7 @@
 from geodebug.engine.policy import Policy
 from geodebug.facts.relations import build_relation_facts
 from geodebug.models.context import EvaluationContext
+from geodebug.models.operations import OperationContext
 from geodebug.models.report import Report
 from geodebug.models.subjects import DatasetSnapshot
 
@@ -50,6 +52,26 @@ def compare(
     return evaluate(context, policy=policy)
 
 
+def preflight(
+    target: Any,
+    *,
+    operation: str,
+    parameters: Mapping[str, Any] | None = None,
+    deep: bool = False,
+    policy: Policy | None = None,
+) -> Report:
+    """Evaluate operation-aware diagnostics before an operation is executed."""
+    snapshot = inspect(target, deep=deep)
+    context = EvaluationContext(
+        subjects=(snapshot,),
+        operation=OperationContext(
+            name=operation,
+            parameters=parameters or {},
+        ),
+    )
+    return evaluate(context, policy=policy)
+
+
 def evaluate(context: EvaluationContext, *, policy: Policy | None = None) -> Report:
     """Evaluate built-in rules against an already-normalized context."""
     return Evaluator(build_default_registry(), policy=policy).evaluate(context)