Fix some grid_spec warnings

Make a private function that
extracts the same product information
that was previously gotten from
the now deprecated grid_spec,
and then use that helper instead
of grid_spec.

The helper should be removed once
there are suitable replacements
in some part of the opendatacube-API.

Author: pjonsson

cubedash/summary/_extents.py

@@ -4,7 +4,7 @@
 from dataclasses import dataclass
 from datetime import date, datetime
 from pathlib import Path
-from typing import Dict, Iterable, List, Optional, TypeAlias
+from typing import Any, Dict, Iterable, List, Optional, TypeAlias
 
 import fiona
 import structlog
@@ -15,6 +15,7 @@
 from datacube.model import Dataset, Field, MetadataType, Product
 from geoalchemy2 import Geometry, WKBElement
 from geoalchemy2.shape import to_shape
+from odc.geo import CRS
 from psycopg2._range import Range as PgRange
 from shapely.geometry import shape
 from sqlalchemy import (
@@ -287,12 +288,9 @@ def _select_dataset_extent_columns(
     # Some time-series-derived products have seemingly-rectangular but *huge* footprints
     # (because they union many almost-indistinguishable footprints)
     # If they specify a resolution, we can simplify the geometry based on it.
-    if (
-        footprint_expression is not None
-        and product.grid_spec
-        and product.grid_spec.resolution
-    ):
-        resolution = min(abs(r) for r in product.grid_spec.resolution)
+    info = _product_info(product)
+    if footprint_expression is not None and info is not None:
+        resolution = min(abs(r) for r in info.resolution)
         footprint_expression = func.ST_SimplifyPreserveTopology(
             footprint_expression, resolution / 4
         )
@@ -442,15 +440,15 @@ def for_product(
             "region_code"
         )
 
-        grid_spec = product.grid_spec
         # Ingested grids trump the "region_code" field because they've probably sliced it up smaller.
         #
         # hltc has a grid spec, but most attributes are missing, so grid_spec functions fail.
         # Therefore: only assume there's a grid if tile_size is specified.
         if region_code_field is not None:
             # Generic region info
             return RegionInfo(product, known_regions)
-        elif grid_spec is not None and grid_spec.tile_size:
+        info = _product_info(product)
+        if info is not None and info.get("tile_size") is not None:
             return GridRegionInfo(product, known_regions)
         elif "sat_path" in product.metadata_type.dataset_fields:
             return SceneRegionInfo(product, known_regions)
@@ -523,7 +521,6 @@ def alchemy_expression(self):
 
         """
         product = self.product
-        grid_spec = product.grid_spec
 
         doc = jsonb_doc_expression(product.metadata_type)
         projection_offset = _projection_doc_offset(product.metadata_type)
@@ -536,9 +533,10 @@ def alchemy_expression(self):
             )
         )
 
-        # todo: look at grid_spec crs. Use it for defaults, conversion.
-        size_x, size_y = grid_spec.tile_size or (1000.0, 1000.0)
-        origin_x, origin_y = grid_spec.origin
+        info = _product_info(product) or {}
+        # todo: use the CRS for defaults, conversion.
+        size_x, size_y = info.get("tile_size") or (1000.0, 1000.0)
+        origin_x, origin_y = info.get("origin") or (0.0, 0.0)
         return func.concat(
             func.floor((func.ST_X(center_point) - origin_x) / size_x).cast(String),
             "_",
@@ -561,6 +559,24 @@ def dataset_region_code(self, dataset: Dataset) -> Optional[str]:
         return f"{x}_{y}"
 
 
+def _product_info(product: Product) -> dict[str, Any] | None:
+    storage = product.definition.get("storage")
+    if storage is None:
+        return None
+    storage_crs = storage.get("crs")
+    if storage_crs is None:
+        return None
+    crs = CRS(str(storage_crs).strip())
+
+    def extract_point(name: str) -> tuple[str] | None:
+        xx = storage.get(name)
+        return None if xx is None else tuple(xx[dim] for dim in crs.dimensions)
+
+    info = {name: extract_point(name) for name in ("tile_size", "resolution", "origin")}
+    complete = all(info[k] is not None for k in ("tile_size", "resolution"))
+    return info if complete else None
+
+
 def _from_xy_region_code(region_code: str):
     """
     >>> _from_xy_region_code('95_3')