Drop read_vector

CHANGELOG.md

@@ -14,6 +14,10 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 ### Changed
 
 - `MultiBackendJobManager`: costs has been added as a column in tracking databases ([[#588](https://github.com/Open-EO/openeo-python-client/issues/588)])
+- When passing a path/string as `geometry` to `DataCube.aggregate_spatial()`, `DataCube.mask_polygon()`, etc.:
+  this is not translated automatically anymore to deprecated, non-standard `read_vector` usage.
+  Instead, if it is a local GeoJSON file, the GeoJSON data will be loaded directly client-side.
+  ([#104](https://github.com/Open-EO/openeo-python-client/issues/104), [#457](https://github.com/Open-EO/openeo-python-client/issues/457))
 
 ### Removed
 

docs/cookbook/tricks.rst

@@ -80,3 +80,51 @@ For example:
 
     # `create_job` with URL to JSON file
     job = connection.create_job("https://jsonbin.example/my/process-graph.json")
+
+
+.. _legacy_read_vector:
+
+
+Legacy ``read_vector`` usage
+----------------------------
+
+In versions up to 0.35.0 of the openEO Python client library,
+there was an old, deprecated feature in geometry handling
+of :py:class:`~openeo.rest.datacube.DataCube` methods like
+:py:meth:`~openeo.rest.datacube.DataCube.aggregate_spatial()` and
+:py:meth:`~openeo.rest.datacube.DataCube.mask_polygon()`
+where you could pass a *backend-side* path as ``geometries``, e.g.:
+
+.. code-block:: python
+
+    cube = cube.aggregate_spatial(
+        geometries="/backend/path/to/geometries.json",
+        reducer="mean"
+    )
+
+The client would handle this by automatically adding a ``read_vector`` process
+in the process graph, with that path as argument, to instruct the backend to load the geometries from there.
+This ``read_vector`` process was however a backend-specific, experimental and now deprecated process.
+Moreover, it assumes that the user has access to (or at least knowledge of) the backend's file system,
+which violates the openEO principle of abstracting away backend-specific details.
+
+In version 0.36.0, this old deprecated ``read_vector`` feature has been *removed*,
+to allow other and better convenience functionality
+when providing a string in the ``geometries`` argument:
+e.g. load from a URL with standard process ``load_url``,
+or load GeoJSON from a local clientside path.
+
+If your workflow however depends on the old, deprecated ``read_vector`` functionality,
+it is possible to reconstruct that by manually adding a ``read_vector`` process in your workflow,
+for example as follows:
+
+.. code-block:: python
+
+    from openeo.processes import process
+
+    cube = cube.aggregate_spatial(
+        geometries=process("read_vector", filename="/backend/path/to/geometries.json"),
+        reducer="mean"
+    )
+
+Note that this is also works with older versions of the openEO Python client library.

openeo/rest/datacube.py

@@ -59,7 +59,7 @@
 from openeo.rest.service import Service
 from openeo.rest.udp import RESTUserDefinedProcess
 from openeo.rest.vectorcube import VectorCube
-from openeo.util import dict_no_none, guess_format, normalize_crs, rfc3339
+from openeo.util import dict_no_none, guess_format, load_json, normalize_crs, rfc3339
 
 if typing.TYPE_CHECKING:
     # Imports for type checking only (circular import issue at runtime).
@@ -609,7 +609,8 @@ def filter_spatial(
               (also see :py:func:`Connection.list_file_formats() <openeo.rest.connection.Connection.list_file_formats>`),
               e.g. GeoJSON, GeoParquet, etc.
               A ``load_url`` process will automatically be added to the process graph.
-            - a path (that is valid for the back-end) to a GeoJSON file.
+            - a path (:py:class:`str` or :py:class:`~pathlib.Path`) to a local, client-side GeoJSON file,
+              which will be loaded automatically to get the geometries as GeoJSON construct.
             - a :py:class:`~openeo.rest.vectorcube.VectorCube` instance.
             - a :py:class:`~openeo.api.process.Parameter` instance.
 
@@ -619,6 +620,12 @@ def filter_spatial(
 
         .. versionchanged:: 0.36.0
             Support passing a URL as ``geometries`` argument, which will be loaded with the ``load_url`` process.
+
+        .. versionchanged:: 0.36.0
+            Support for passing a backend-side path as ``geometries`` argument was removed
+            (also see :ref:`legacy_read_vector`).
+            Instead, it's possible to provide a client-side path to a GeoJSON file
+            (which will be loaded client-side to get the geometries as GeoJSON construct).
         """
         valid_geojson_types = [
             "Point", "MultiPoint", "LineString", "MultiLineString",
@@ -1053,7 +1060,7 @@ def _merge_operator_binary_cubes(
 
     def _get_geometry_argument(
         self,
-        geometry: Union[
+        argument: Union[
             shapely.geometry.base.BaseGeometry,
             dict,
             str,
@@ -1065,19 +1072,19 @@ def _get_geometry_argument(
         crs: Optional[str] = None,
     ) -> Union[dict, Parameter, PGNode]:
         """
-        Convert input to a geometry as "geojson" subtype object.
+        Convert input to a geometry as "geojson" subtype object or vectorcube.
 
         :param crs: value that encodes a coordinate reference system.
             See :py:func:`openeo.util.normalize_crs` for more details about additional normalization that is applied to this argument.
         """
-        if isinstance(geometry, Parameter):
-            return geometry
-        elif isinstance(geometry, _FromNodeMixin):
-            return geometry.from_node()
+        if isinstance(argument, Parameter):
+            return argument
+        elif isinstance(argument, _FromNodeMixin):
+            return argument.from_node()
 
-        if isinstance(geometry, str) and re.match(r"^https?://", geometry, flags=re.I):
+        if isinstance(argument, str) and re.match(r"^https?://", argument, flags=re.I):
             # Geometry provided as URL: load with `load_url` (with best-effort format guess)
-            url = urllib.parse.urlparse(geometry)
+            url = urllib.parse.urlparse(argument)
             suffix = pathlib.Path(url.path.lower()).suffix
             format = {
                 ".json": "GeoJSON",
@@ -1086,18 +1093,20 @@ def _get_geometry_argument(
                 ".parquet": "Parquet",
                 ".geoparquet": "Parquet",
             }.get(suffix, suffix.split(".")[-1])
-            return self.connection.load_url(url=geometry, format=format)
-
-        if isinstance(geometry, (str, pathlib.Path)):
-            # Assumption: `geometry` is path to polygon is a path to vector file at backend.
-            # TODO #104: `read_vector` is non-standard process.
-            # TODO: If path exists client side: load it client side?
-            return PGNode(process_id="read_vector", arguments={"filename": str(geometry)})
+            return self.connection.load_url(url=argument, format=format)
 
-        if isinstance(geometry, shapely.geometry.base.BaseGeometry):
-            geometry = mapping(geometry)
-        if not isinstance(geometry, dict):
-            raise OpenEoClientException("Invalid geometry argument: {g!r}".format(g=geometry))
+        if (
+            isinstance(argument, (str, pathlib.Path))
+            and pathlib.Path(argument).is_file()
+            and pathlib.Path(argument).suffix.lower() in [".json", ".geojson"]
+        ):
+            geometry = load_json(argument)
+        elif isinstance(argument, shapely.geometry.base.BaseGeometry):
+            geometry = mapping(argument)
+        elif isinstance(argument, dict):
+            geometry = argument
+        else:
+            raise OpenEoClientException(f"Invalid geometry argument: {argument!r}")
 
         if geometry.get("type") not in valid_geojson_types:
             raise OpenEoClientException("Invalid geometry type {t!r}, must be one of {s}".format(
@@ -1147,7 +1156,8 @@ def aggregate_spatial(
               (also see :py:func:`Connection.list_file_formats() <openeo.rest.connection.Connection.list_file_formats>`),
               e.g. GeoJSON, GeoParquet, etc.
               A ``load_url`` process will automatically be added to the process graph.
-            - a path (that is valid for the back-end) to a GeoJSON file.
+            - a path (:py:class:`str` or :py:class:`~pathlib.Path`) to a local, client-side GeoJSON file,
+              which will be loaded automatically to get the geometries as GeoJSON construct.
             - a :py:class:`~openeo.rest.vectorcube.VectorCube` instance.
             - a :py:class:`~openeo.api.process.Parameter` instance.
 
@@ -1177,6 +1187,12 @@ def aggregate_spatial(
 
         .. versionchanged:: 0.36.0
             Support passing a URL as ``geometries`` argument, which will be loaded with the ``load_url`` process.
+
+        .. versionchanged:: 0.36.0
+            Support for passing a backend-side path as ``geometries`` argument was removed
+            (also see :ref:`legacy_read_vector`).
+            Instead, it's possible to provide a client-side path to a GeoJSON file
+            (which will be loaded client-side to get the geometries as GeoJSON construct).
         """
         valid_geojson_types = [
             "Point", "MultiPoint", "LineString", "MultiLineString",
@@ -1502,7 +1518,8 @@ def apply_polygon(
               (also see :py:func:`Connection.list_file_formats() <openeo.rest.connection.Connection.list_file_formats>`),
               e.g. GeoJSON, GeoParquet, etc.
               A ``load_url`` process will automatically be added to the process graph.
-            - a path (that is valid for the back-end) to a GeoJSON file.
+            - a path (:py:class:`str` or :py:class:`~pathlib.Path`) to a local, client-side GeoJSON file,
+              which will be loaded automatically to get the geometries as GeoJSON construct.
             - a :py:class:`~openeo.rest.vectorcube.VectorCube` instance.
             - a :py:class:`~openeo.api.process.Parameter` instance.
 
@@ -1519,6 +1536,12 @@ def apply_polygon(
 
         .. versionchanged:: 0.36.0
             Support passing a URL as ``geometries`` argument, which will be loaded with the ``load_url`` process.
+
+        .. versionchanged:: 0.36.0
+            Support for passing a backend-side path as ``geometries`` argument was removed
+            (also see :ref:`legacy_read_vector`).
+            Instead, it's possible to provide a client-side path to a GeoJSON file
+            (which will be loaded client-side to get the geometries as GeoJSON construct).
         """
         # TODO drop support for legacy `polygons` argument:
         #      remove `kwargs, remove default `None` value for `geometries` and `process`
@@ -2011,7 +2034,8 @@ def mask_polygon(
               (also see :py:func:`Connection.list_file_formats() <openeo.rest.connection.Connection.list_file_formats>`),
               e.g. GeoJSON, GeoParquet, etc.
               A ``load_url`` process will automatically be added to the process graph.
-            - a path (that is valid for the back-end) to a GeoJSON file.
+            - a path (:py:class:`str` or :py:class:`~pathlib.Path`) to a local, client-side GeoJSON file,
+              which will be loaded automatically to get the geometries as GeoJSON construct.
             - a :py:class:`~openeo.rest.vectorcube.VectorCube` instance.
             - a :py:class:`~openeo.api.process.Parameter` instance.
 
@@ -2024,6 +2048,12 @@ def mask_polygon(
 
         .. versionchanged:: 0.36.0
             Support passing a URL as ``geometries`` argument, which will be loaded with the ``load_url`` process.
+
+        .. versionchanged:: 0.36.0
+            Support for passing a backend-side path as ``geometries`` argument was removed
+            (also see :ref:`legacy_read_vector`).
+            Instead, it's possible to provide a client-side path to a GeoJSON file
+            (which will be loaded client-side to get the geometries as GeoJSON construct).
         """
         valid_geojson_types = ["Polygon", "MultiPolygon", "GeometryCollection", "Feature", "FeatureCollection"]
         mask = self._get_geometry_argument(mask, valid_geojson_types=valid_geojson_types, crs=srs)

tests/data/1.0.0/aggregate_zonal_path.json

@@ -21,11 +21,11 @@
       }
     }
   },
-  "readvector1": {
-    "process_id": "read_vector",
+  "loadurl1": {
+    "process_id": "load_url",
     "arguments": {
-      "filename": "/some/path/to/GeometryCollection.geojson"
-    }
+      "url": "https://example.com/geometries.geojson",
+      "format": "GeoJSON"}
   },
   "aggregatespatial1": {
     "process_id": "aggregate_spatial",
@@ -34,7 +34,7 @@
         "from_node": "filterbbox1"
       },
       "geometries": {
-        "from_node": "readvector1"
+        "from_node": "loadurl1"
       },
       "reducer": {
         "process_graph": {

tests/data/geojson/polygon02.json

@@ -0,0 +1,23 @@
+{
+  "type": "Polygon",
+  "coordinates": [
+    [
+      [
+        3,
+        50
+      ],
+      [
+        4,
+        50
+      ],
+      [
+        4,
+        51
+      ],
+      [
+        3,
+        50
+      ]
+    ]
+  ]
+}