Xpublish routers for the OGC EDR API.

For conda users you can

conda install --channel conda-forge xpublish_edr

or, if you are a pip users

pip install xpublish_edr

import xarray as xr
import xpublish
from xpublish_edr import CfEdrPlugin

ds = xr.open_dataset("dataset.nc")

rest = xpublish.Rest(
    {"ds": ds},
    plugins={"edr": CfEdrPlugin()},
)

For /position, /area, and /cube, xpublish-edr needs to know the dataset's native CRS and which dimensions are X and Y. There are three supported ways to provide that information.

Use a grid_mapping attribute on each data variable and CF attrs on the X/Y coordinates:

import pyproj
import xarray as xr

crs = pyproj.CRS.from_epsg(3857)

ds["spatial_ref"] = xr.DataArray(0, attrs=crs.to_cf())
ds["temperature"].attrs["grid_mapping"] = "spatial_ref"
ds["x"].attrs.update(
    axis="X",
    standard_name="projection_x_coordinate",
)
ds["y"].attrs.update(
    axis="Y",
    standard_name="projection_y_coordinate",
)

For native longitude/latitude grids, use CF longitude/latitude coordinate attrs instead of projected X/Y attrs:

ds["lon"].attrs["standard_name"] = "longitude"
ds["lat"].attrs["standard_name"] = "latitude"

Datasets that already have a scalar spatial_ref or crs variable with CF CRS attrs are also accepted.

Use proj: attrs for the CRS and spatial:dimensions for the Y/X dimensions:

ds.attrs["proj:code"] = "EPSG:3857"
ds.attrs["spatial:dimensions"] = ["y", "x"]  # [Y, X] order

Use proj:wkt2 instead of proj:code if the CRS is stored as WKT.

For raster-style datasets with x/y dimensions but no explicit X/Y coordinate arrays, provide CRS and affine transform metadata. rasterix detects the raster dimensions and materializes regular 1D X/Y coordinates before selection.

CF/GDAL form:

ds["spatial_ref"] = xr.DataArray(
    0,
    attrs={
        **crs.to_cf(),
        "GeoTransform": "0 1000 0 3000 0 -1000",
    },
)
ds["temperature"].attrs["grid_mapping"] = "spatial_ref"

GeoZarr form:

ds.attrs["proj:code"] = "EPSG:3857"
ds.attrs["spatial:transform"] = [1000, 0, 0, 0, -1000, 3000]
ds.attrs["spatial:dimensions"] = ["y", "x"]

datetime and z queries also require indexed CF T and Z coordinates.

Spatial selection currently expects regular 1D X/Y coordinate grids, or an affine transform that can be materialized into regular 1D X/Y coordinates. 2D curvilinear spatial selection and proj:projjson CRS attrs are not currently supported.

This package attempts to follow the spec where reasonable, adding functionality where the value is demonstrable.

Note: POST is supported on /position and /area as a non-spec extension so that requests with large geometries (many points, complex polygons) can submit them in the request body instead of being limited by URL length. All selection parameters (datetime, z, parameter-name, crs, f, method) are still passed as query string parameters. See the per-query tables below for supported body content types.

On its own, xpublish-edr serves queries at /{dataset_id}/edr/{query} rather than the /collections/{collectionId}/{query} path template described in the spec, because of the path structure of xpublish. Collection metadata is available at the dataset level through the /{dataset_id}/edr/ resource.

When composed with xpublish-ogc-core, the spec compliant resource paths are also served.

If xpublish-ogc-core is installed alongside xpublish-edr (both load automatically through their xpublish.plugin entry points), this plugin implements its OGC hookspecs so that the composed app serves:

• /collections/{collection_id}/position, /collections/{collection_id}/area, and /collections/{collection_id}/cube — every supported geometry query at the spec compliant resource paths, sharing the query handling with the dataset level routes.
• /collections/{collection_id} carrying the full EDR collection metadata (extent, parameter_names, crs, output_formats) and a data_queries member describing each query with collection scoped hrefs.
• /conformance declaring the EDR 1.1 conformance classes (core, collections, json, covjson, and queries).

The integration is validated end to end in tests/test_ogc_*.py against the official OGC schemas vendored by xpublish-ogc-core, Schemathesis fuzz of the composed app's and OGCs OpenAPI description, plus running most of the OGC TeamEngine Dockerized test suite.

8.2.1 Position query

Any additional query parameters are assumed to be additional selections to make on the dimensions/coordinates. These queries will use the specified selections method.

8.2.3 Area query

method is not applicable for the coordinates of area queries, only for selecting datetime, z, or additional dimensions.

For POLYGON coordinates, points that are located within OR on the polygons boundary are included in the response.

8.2.4 Cube query

method is not applicable for the coordinates of cube queries, only for selecting datetime, z, or additional dimensions.

Cube queries are not flattened like area queries, so the response is returned as sliced by xarray. This is particularly useful for subsetting regular grids.

Report bugs, suggest features or view the source code on GitHub.

xpublish-edr is licensed under BSD 3-Clause "New" or "Revised" License (BSD-3-Clause).

Development occurs on GitHub at https://github.com/gulfofmaine/xpublish-edr/issues.