CHANGES.md

Changes in 1.8.0 (in development)

Enhancements

• The method xcube.core.GridMapping.transform now supports lazy execution. If computations based on actual data are required—such as determining whether the grid mapping is regular or estimating the resolution in the x or y direction—only a single chunk is accessed whenever possible, ensuring faster performance.

• The function xcube.core.resampling.rectify_dataset now supports xarray.Datasets containing multi-dimensional data variables structured as var(..., y_dim, x_dim). The two spatial dimensions (y_dim and x_dim) must occupy the last two positions in the variable's dimensions.

• Added a new preload API to xcube data stores:

  • Enhanced the xcube.core.store.DataStore class to optionally support preloading of datasets via an API represented by the
    new xcube.core.store.DataPreloader interface.
  • Added handy default implementations NullPreloadHandle and ExecutorPreloadHandle to be returned by implementations of the prepare_data() method of a given data store.

• A xy_res keyword argument was added to the transform() method of xcube.core.gridmapping.GridMapping, enabling users to set the grid-mapping resolution directly, which speeds up the method by avoiding time-consuming spatial resolution estimation. (#1082)

• The behaviour of the function xcube.core.resample.resample_in_space() has been changed if no tile_size is specified for the target grid mapping. It now defaults to the tile_size of the source grid mapping, improving the user-friendliness of resampling and reprojection. (#1082)

• The "https" data store (store = new_data_store("https", ...)) now allows for lazily accessing NetCDF files. Implementation note: For this to work, the DatasetNetcdfFsDataAccessor class has been adjusted. (#1083)

• Added new endpoint /viewer/state to xcube Server that allows for xcube Viewer state persistence. (#1088)

  The new viewer API operations are:

  • GET /viewer/state to get a keys of stored states or restore a specific state;
  • PUT /viewer/state to store a state and receive a key for it.

  Persistence is configured using new optional Viewer/Persistence setting:

  Viewer:
   Persistence:
     # Any filesystem. Can also be relative to base_dir.
     Path: memory://states
     # Filesystem-specific storage options   
     # StorageOptions: ...

Fixes

• The function xcube.core.resample.resample_in_space() now always operates lazily and therefore supports chunk-wise, parallel processing. (#1082)
• Bux fix in the has_data method of the "https" data store (store = new_data_store("https", ...)). (#1084)
• Bux fix in the has_data method of all filesystem-based data store ("file", "s3", "https"). data_type can be any of the supported data types, e.g. for .tif file, data_type can be either dataset or mldataset. (#1084)
• The explanation of the parameter xy_scale in the method xcube.core.gridmapping.GridMapping.scale has been corrected. (#1086)
• The spurious tileserver/viewer warning "no explicit representation of timezones available…" (formerly "parsing timezone aware datetimes is deprecated…") is no longer generated. (#807)

Other changes

• Added experimental feature that allows for extending the xcube Viewer user interface with server-side panels. For this to work, users can now configure xcube Server to load one or more Python modules that provide xcube.webapi.viewer.contrib.Panel UI-contributions. Panel instances provide two decorators layout() and callback() which are used to implement the UI and the interaction behaviour, respectively. The functionality is provided by the Chartlets Python library. A working example can be found in examples/serve/panels-demo.

• The xcube test helper module test.s3test has been enhanced to support testing the experimental server-side panels described above:

  • added new decorator @s3_test() for individual tests with timeout arg;
  • added new context manager s3_test_server() with timeout arg to be used within tests function bodies;
  • S3Test, @s3_test(), and s3_test_server() now restore environment variables modified for the Moto S3 test server.

Changes in 1.7.1

Enhancements

• Level creation now supports aggregation method mode to aggregate to the value which is most frequent. (#913)

Fixes

• The time query parameter of the /statistics endpoint of xcube server has now been made optional. (#1066)
• The /statistics endpoint now supports datasets using non-WGS84 grid systems, expanding its compatibility with a wider range of geospatial datasets. (#1069)
• Bug fix in resampling_in_space when projecting from geographic to non-geographic projection. (#1073)
• Bug fix of the extent field in the single item collection published by the xcube server STAC API so that it follows the collection STAC specifications. (#1077)

Changes in 1.7.0

Enhancements

• Bundled xcube-viewer 1.3.0.

• xcube server can now deal with "user-defined" variables. Endpoints that accept a {varName} path parameter in their URL path can now be called with assignment expressions of the form <var_name>=<var_expr> where <var_name> is the name user defined variable and <var_expr> is an arbitrary band-math expression, see xcube-dev/xcube-viewer#371.

• xcube server now allows for configuring new dataset properties GroupTitle and Tags . This feature has been added in order to support grouping and filtering of datasets in UIs, see xcube-dev/xcube-viewer#385.

• Added server endpoint GET /statistics/{varName} with query parameters lon, lat, time which is used to extract single point data. This feature has been added in order to support xcube-dev/xcube-viewer#404.

• The xcube server STAC API now publishes all fields available via the /datasets endpoint. This includes colormap information for each asset such as colorBarName, colorBarNorm, colorBarMin, colorBarMax, tileLevelMin, tileLevelMax. (#935, #940)

• xcube server now allows for configuring custom color maps via the configuration file. It supports continuous, stepwise and categorical colormaps, which may be configured as shown in the section CustomColorMaps of the xcube serve documentation (#1055)

Fixes

• Migrated the .github/workflows/xcube_build_docker.yaml and the corresponding Dockerfile from setup.py to pyproject.toml. Additionally, updated the relevant documentation in doc/source to reflect this change from setup.py to pyproject.toml. (related to #992)
• Normalisation with xcube.core.normalize.normalize_dataset fails when chunk encoding must be updated (#1033)
• The open_data method of xcube's default xcube.core.store.DataStore implementations now supports a keyword argument data_type, which determines the data type of the return value. Note that opener_id includes the data_type at its first position and will override the data_type argument. To preserve backward compatibility, the keyword argument data_type has not yet been literally specified as open_data() method argument, but may be passed as part of **open_params. (#1030)
• The xcube.core.store.DataDescriptor class now supports specifying time ranges using both datetime.date and datetime.datetime objects. Previously, only datetime.date objects were supported.
• The xcube server STAC API has been adjusted so that the data store parameters and data ID, which are needed to open the data referred to by a STAC item, are now included with the item's analytic asset. Furthermore, a second assert called analytic_multires will be published referring to the multi-resolution data format levels (#1020).
• Improved the way color mapping works in xcube server to support simplified color bar management in xcube viewer, see xcube-dev/xcube-viewer#390. (#1043)
• The xcube server's dataset configuration extraction methodology has been updated. When the data resource ID is provided in the Path field, xcube will attempt to access the dataset using the given ID. If wildcard patterns are used, the server will crawl through the data store to find matching data IDs. This process may result in a long setup time if the data store contains numerous data IDs. A UserWarning will be issued for the "stac" data store.
• Corrected extent object of a STAC collection issued by xcube server, following the collection STAC specifications (#1053)
• When opening a GeoTIFF file using a file system data store, the default return value is changed from MultiLevelDataset to xr.Dataset, if no data_type is assigned in the open_params of the store.open_data() method. (#1054) xcube server has been adapted to always open MultiLevelDatasets from a specified data store, if that data type is supported.
• Adjustments to resample_in_time() in xcube/core/resampling/temporal.py so that xcube now supports xarray=2024.7.

Other changes

• Renamed internal color mapping types from "node", "bound", "key" into "continuous", "stepwise", "categorical".

Changes in 1.6.0

Enhancements

• Added new statistics API to xcube server. The service computes basic statistical values and a histogram for given data variable, time stamp, and a GeoJSON geometry. Its endpoint is: /statistics/{datasetId}/{varName}?time={time}. Geometry is passed as request body in form of a GeoJSON geometry object.

• xcube server's tile API can now handle user-defined colormaps from xcube viewer. Custom color bars are still passed using query parameter cmap to endpoint /tiles/{datasetId}/{varName}/{z}/{y}/{x}, but in the case of custom color bars it is a JSON-encoded object with the following format: {"name": <str>, "type": <str>, "colors": <list>}. (#975) The object properties are

  • name: a unique name.
  • type: optional type of control values.
  • colors: a list of pairs [[<v1>,<c1>], [<v2>,<c2>], [<v3>,<c3>], ...] that map a control value to a hexadecimal color value using CSS format "#RRGGBBAA".

  The type values are

  • "node": control points are nodes of a continuous color gradient.
  • "bound": control points form bounds that map to a color, which means the last color is unused.
  • "key": control points are keys (integers) that identify a color.

• xcube server's tile API now allows specifying the data normalisation step before a color mapping is applied to the variable data to be visualized. This affects endpoint /tiles/{datasetId}/{varName}/{z}/{y}/{x} and the WMTS API. The possible normalisation values are

  • lin: linear mapping of data values between vmin and vmax to range 0 to 1 (uses matplotlib.colors.Normalize(vmin, vmax)).
  • log: logarithmic mapping of data values between vmin and vmax to range 0 to 1 (uses matplotlib.colors.LogNorm(vmin, vmax)).
  • cat: categorical mapping of data values to indices into the color mapping. (uses matplotlib.colors.BoundaryNorm(categories)). This normalisation currently only works with user-defined colormaps of type key or bound (see above).

  The normalisation can be specified in three different ways (in order):

  1. As query parameter norm passed to the tile endpoint.
  2. Property Norm in the Styles/ColorMapping element in xcube server configuration.
  3. Data variable attribute color_norm.

• xcube server can now read SNAP color palette definition files (*.cpd) with alpha values. (#932)

• The class xcube.webapi.viewer.Viewer now accepts root paths or URLs that will each be scanned for datasets. The roots are passed as keyword argument roots whose value is a path or URL or an iterable of paths or URLs. A new keyword argument max_depth defines the maximum subdirectory depths used to search for datasets in case roots is given. It defaults to 1.

• The behaviour of function resample_in_space() of module xcube.core.resampling changed in this version. (#1001)

  1. A new keyword argument ref_ds can now be used to provide a reference dataset for the reprojection. It can be passed instead of target_rm. If ref_ds is given, it also forces the returned target dataset to have the same spatial coordinates as ref_ds.
  2. In the case of up-sampling, we no longer recover NaN values by default as it may require considerable CPU overhead. To enforce the old behaviour, provide the var_configs keyword-argument and set recover_nan to True for desired variables.

• The class MaskSet() of module xcube.core.maskset now correctly recognises the variable attributes flag_values, flag_masks, flag_meanings when their values are lists (ESA CCI LC data encodes them as JSON arrays). (#1002)

• The class MaskSet() now provides a method get_cmap() which creates a suitable matplotlib color map for variables that define the flag_values CF-attribute and optionally a flag_colors attribute. (#1011)

• The Api.route decorator and ApiRoute constructor in xcube.server.api now have a slash argument which lets a route support an optional trailing slash.

Fixes

• When using the xcube.webapi.viewer.Viewer class in Jupyter notebooks multi-level datasets opened from S3 or from deeper subdirectories into the local filesystem are now fully supported. (#1007)

• Fixed an issue with xcube server /timeseries endpoint that returned status 500 if a given dataset used a CRS other geographic and the geometry was not a point. (#995)

• Fixed broken table of contents links in dataset convention document.

• Web API endpoints with an optional trailing slash are no longer listed twice in the automatically generated OpenAPI documentation (#965)

• Several minor updates to make xcube compatible with NumPy 2.0.0 (#1024)

Incompatible API changes

• The get_cmap() method of util.cmaps.ColormapProvider now returns a Tuple[matplotlib.colors.Colormap, Colormap] instead of Tuple[str, matplotlib.colors.Colormap].

• The signatures of functions resample_in_space(), rectify_dataset(), and affine_transform_dataset() of module xcube.core.resampling changed:

  • Source dataset must be provided as 1st positional argument.
  • Introduced keyword argument ref_ds that can be provided instead of target_gm. If given, it forces the returned dataset to have the same coordinates as ref_ds.

• Removed API deprecated since many releases:

  • Removed keyword argument base from function xcube.core.resampling.temporal.resample_in_time().
  • Removed option base from CLI command xcube resample.
  • Removed keyword argument assert_cube from xcube.core.timeseries.get_time_series().
  • Removed property xcube.core.xarray.DatasetAccessor.levels.
  • Removed function xcube.core.tile.parse_non_spatial_labels().
  • Removed keyword argument tag from context manager xcube.util.perf.measure_time().
  • Removed function xcube.core.geom.convert_geometry().
  • Removed function xcube.core.geom.is_dataset_y_axis_inverted().
  • Removed function xcube.util.assertions.assert_condition().
  • Removed function xcube.util.cmaps.get_cmaps().
  • Removed function xcube.util.cmaps.get_cmap().
  • Removed function xcube.util.cmaps.ensure_cmaps_loaded().
  • Removed endpoint /datasets/{datasetId}/vars/{varName}/tiles2/{z}/{y}/{x} from xcube server.

Other changes

• Make tests compatible with PyTest 8.2.0. (#973)

• Addressed all warnings from xarray indicating that Dataset.dims will be replaced by Dataset.sizes. (#981)

• NUMBA_DISABLE_JIT set to 0 to enable numba.jit in github workflow. (#946)

• Added GitHub workflow to perform an automatic xcube release on PyPI after a GitHub release. To install xcube via the pip tool use pip install xcube-core,
  since the name "xcube" is already taken on PyPI by another software. (#982)

• Added project URLs and classifiers to setup.py, which will be shown in the left sidebar on the PyPI xcube-core webpage.

• Refactored xcube workflow to build docker images only on release and deleted the update xcube tag job.

• Used pyupgrade to automatically upgrade language syntax for Python versions >= 3.9.

• Migrated the xcube project setup from setup.py to the modern pyproject.toml format.

• The functions mask_dataset_by_geometry() and clip_dataset_by_geometry() of module xcube.core.geom have a new keyword argument update_attrs: bool = True as part of the fix for #995.

• Decreased number of warnings in the xcube workflow step unittest-xcube.

• Added new data store "https" that uses fsspec.implementations.http.HTTPFileSystem), so that the upcoming xcube STAC data store will be able to access files from URLs.

• The workflow .github/workflows/xcube_publish_pypi.yml changes the line in the pyproject.toml, where the package name is defined to name = "xcube-core". This allows to release xcube under the package name "xcube-core" on PyPI where the name "xcube" is already taken. #1010

• Updated the 'How do I ...' page in the xcube documentation.

Changes in 1.5.1

• Embedded xcube-viewer 1.1.1.

• Fixed xcube plugin auto-recognition in case a plugin project uses pyproject.toml file. (#963)

• Updated copyright notices in all source code files.

Changes in 1.5.0

• Enhanced spatial resampling in module xcube.core.resampling (#955):

  • Added optional keyword argument interpolation to function rectify_dataset() with values "nearest", "triangular", and "bilinear" where "triangular" interpolates between 3 and "bilinear" between 4 adjacent source pixels.
  • Function rectify_dataset() is now ~2 times faster by early detection of already transformed target pixels.
  • Added a documentation page that explains the algorithm used in rectify_dataset().
  • Added optional keyword argument rectify_kwargs to resample_in_space(). If given, it is spread into keyword arguments passed to the internal rectify_dataset() delegation, if any.
  • Deprecated unused keyword argument xy_var_names of function rectify_dataset().

• Replace use of deprecated method in testing module. (#961)

• Update dependencies to better match imports; remove the defaults channel; turn adlfs into a soft dependency. (#945)

• Reformatted xcube code base using black default settings. It implies a line length of 88 characters and double quotes for string literals. Also added .editorconfig for IDEs not recognising black's defaults.

• Renamed xcube's main branch from master to main on GitHub.

• xcube's code base changed its docstring format from reST style to the much better readable Google style. Existing docstrings have been converted using the awesome docconvert tool.

• Add a data_vars_only parameter to chunk_dataset and update_dataset_chunk_encoding (#958).

• Update some unit tests to make them compatible with xarray 2024.3.0 (#958).

• Added documentation page "How do I ..." that points users to applicable xcube Python API.

Changes in 1.4.1

Enhancements

• Data stores can now return data iterators from their open_data() method. For example, a data store implementation can now return a data cube either with a time dimension of size 100, or could be asked to return 100 cube time slices with dimension size 1 in form of an iterator. This feature has been added to effectively support the new zappend tool. (#919)

Fixes

• Fix two OGC Collections unit tests that were failing under Windows. (#937)

Other changes

• Minor updates to make xcube compatible with pandas 2 and Python 3.12. (#933)

• Minor updates to make xcube compatible with xarray >=2023.9.0. (#897, #939)

Changes in 1.4.0

Enhancements

• Added new reference filesystem data store to support "kerchunked" NetCDF files in object storage. (#928)

  See also

  • ReferenceFileSystem
  • kerchunk

• Improved xcube Server's STAC API:

  • Provide links for multiple coverages data formats
  • Add crs and crs_storage properties to STAC data
  • Add spatial and temporal grid data to collection descriptions
  • Add a schema endpoint returning a JSON schema of a dataset's data variables
  • Add links to domain set, range type, and range schema to collection descriptions

• Improved xcube Server's Coverages API:

  • Support scaling parameters scale-factor, scale-axes, and scale-size
  • Improve handling of bbox parameters
  • Handle half-open datetime intervals
  • More robust and standard-compliant parameter parsing and checking
  • More informative responses for incorrect or unsupported parameters
  • Omit unnecessary dimensions in TIFF and PNG coverages
  • Use crs_wkt when determining CRS, if present and needed
  • Change default subsetting and bbox CRS from EPSG:4326 to OGC:CRS84
  • Implement reprojection for bbox
  • Ensure datetime parameters match dataset’s timezone awareness
  • Reimplement subsetting (better standards conformance, cleaner code)
  • Set Content-Bbox and Content-Crs headers in the HTTP response
  • Support safe CURIE syntax for CRS specification

Fixes

• Fixed KeyError: 'lon_bnds' raised occasionally when opening (mostly NetCDF) datasets. (#930)
• Make S3 unit tests compatible with moto 5 server. (#922)
• Make some CLI unit tests compatible with pytest 8. (#922)
• Rename some test classes to avoid spurious warnings. (#924)

Other changes

• Require Python >=3.9 (previously >=3.8)

Changes in 1.3.1

• Updated Dockerfile and GitHub workflows; no changes to the xcube codebase itself

Changes in 1.3.0

Enhancements

• Added a basic implementation of the draft version of OGC API - Coverages. (#879, #889, #900)
• Adapted the STAC implementation to additionally offer datasets as individual collections for better integration with OGC API - Coverages. (#889)
• Various minor improvements to STAC implementation. (#900)

Fixes

• Resolved the issue for CRS84 error due to latest version of gdal (#869)
• Fixed incorrect additional variable data in STAC datacube properties. (#889)
• Fixed access of geotiff datasets from public s3 buckets (#893)

Other changes

• update_dataset_attrs can now also handle datasets with CRS other than WGS84 and update the metadata according to the ESIP Attribute Convention for Data Discovery.
• removed deprecated module xcube edit, which has been deprecated since version 0.13.0
• Update "Development process" section of developer guide.
• Updated GitHub workflow to build docker image for GitHub releases only and not on each commit to main.

Changes in 1.2.0

Enhancements

• Added a new, experimental /compute API to xcube server. It comprises the following endpoints:

  • GET compute/operations - List available operations.
  • GET compute/operations/{opId} - Get details of a given operation.
  • PUT compute/jobs - Start a new job that executes an operation.
  • GET compute/jobs - Get all jobs.
  • GET compute/jobs/{jobId} - Get details of a given job.
  • DELETE compute/jobs/{jobId} - Cancel a given job.

  The available operations are currently taken from module xcube.webapi.compute.operations.

  To disable the new API use the following server configuration:

  api_spec:
    excludes: ["compute"] 
  ...

Other changes

• Added shutdown_on_close=True parameter to coiled params to ensure that the clusters are shut down on close. (#881)
• Introduced new parameter region for utility function new_cluster in xcube.util.dask which will ensure coiled creates the dask cluster in the prefered default region: eu-central-1. (#882)
• Server offers the function add_place_group in places/context.py, which allows plugins to add place groups from external sources.

Changes in 1.1.3

Fixes

• Fixed Windows-only bug in xcube serve --config <path>: If config path is provided with back-slashes, a missing base_dir config parameter is now correctly set to the parent directory of path. Before, the current working directory was used.

Other changes

• Updated AppVeyor and GitHub workflow configurations to use micromamba rather than mamba (#785)

Changes in 1.1.2

Fixes

• Fixed issue where geotiff access from a protected s3 bucket was denied (#863)

Changes in 1.1.1

• Bundled new build of xcube-viewer 1.1.0.1 that will correctly respect a given xcube server from loaded from the viewer configuration.

Changes in 1.1.0

Enhancements

• Bundled xcube-viewer 1.1.0.

• Updated installation instructions (#859)

• Included support for FTP filesystem by adding a new data store ftp.

  These changes will enable access to data cubes (.zarr or .levels) in FTP storage as shown here:

  store = new_data_store(
      "ftp",                     # FTP filesystem protocol
      root="path/to/files",      # Path on FTP server
      storage_options= {'host':  'ftp.xxx',  # The url to the ftp server
                        'port': 21           # Port, defaults to 21  
                        # Optionally, use 
                        # 'username': 'xxx'
                        # 'password': 'xxx'}  
  )
  store.list_data_ids()

  Note that there is no anon parameter, as the store will assume no anonymity if no username and password are set.

  Same configuration for xcube Server:

  DataStores:
  - Identifier: siec
    StoreId: ftp
    StoreParams:
      root: my_path_on_the_host
      max_depth: 1
      storage_options:
        host: "ftp.xxx"
        port: xxx
        username: "xxx"
        password': "xxx"

• Updated xcube Dataset Specification. (addressing #844)

• Added xcube Data Access documentation.

Fixes

• Fixed various issues with the auto-generated Python API documentation.

• Fixed a problem where time series requests may have missed outer values of a requested time range. (#860)

  • Introduced query parameter tolerance for endpoint /timeseries/{datasetId}/{varName} which is the number of seconds by which the given time range is expanded. Its default value is one second to overcome rounding problems with microsecond fractions. (#860)
  • We now round the time dimension labels for a dataset as follows (rounding frequency is 1 second by default):
    • First times stamp: floor(time[0])
    • Last times stamp: ceil(time[-1])
    • In-between time stamps: round(time[1: -1])

Other changes

• Pinned gdal dependency to >=3.0, <3.6.3 due to incompatibilities.

Changes in 1.0.5

• When running xcube in a JupyterLab, the class xcube.webapi.viewer.Viewer can be used to programmatically launch a xcube Viewer UI. The class now recognizes an environment variable XCUBE_JUPYTER_LAB_URL that contains a JupyterLab's public base URL for a given user. To work properly, the jupyter-server-proxy extension must be installed and enabled.

• Bundled xcube-viewer 1.0.2.1.

Changes in 1.0.4

• Setting a dataset's BoundingBox in the server configuration is now recognised when requesting the dataset details. (#845)

• It is now possible to enforce the order of variables reported by xcube server. The new server configuration key Variables can be added to Datasets configurations. Is a list of wildcard patterns that determines the order of variables and the subset of variables to be reported. (#835)

• Pinned Pandas dependency to lower than 2.0 because of incompatibility with both xarray and xcube (see pydata/xarray#7716). Therefore, the following xcube deprecations have been introduced:

  • The optional --base/-b of the xcube resample CLI tool.
  • The keyword argument base of the xcube.core.resample.resample_in_time function.

• Bundled xcube-viewer 1.0.2.

Changes in 1.0.3

Same as 1.0.2, just fixed unit tests due to minor Python environment change.

Changes in 1.0.2

• Bundled latest xcube-viewer 1.0.1.

• xcube is now compatible with Python 3.10. (#583)

• The Viewer.add_dataset() method of the xcube JupyterLab integration has been enhanced by two optional keyword arguments style and color_mappings to allow for customized, initial color mapping of dataset variables. The example notebook xcube-viewer-in-jl.ipynb has been updated to reflect the enhancement.

• Fixed an issue with new xcube data store abfs for the Azure Blob filesystem. (#798)

Changes in 1.0.1

Fixes

• Fixed recurring issue where xcube server was unable to locate Python code downloaded from S3 when configuring dynamically computed datasets (configuration FileSystem: memory) or augmenting existing datasets by dynamically computed variables (configuration Augmentation). (#828)

Changes in 1.0.0

Enhancements

• Added a catalog API compliant to STAC to xcube server. (#455)

  • It serves a single collection named "datacubes" whose items are the datasets published by the service.
  • The collection items make use the STAC datacube extension.

• Simplified the cloud deployment of xcube server/viewer applications (#815). This has been achieved by the following new xcube server features:

  • Configuration files can now also be URLs which allows provisioning from S3-compatible object storage. For example, it is now possible to invoke xcube server as follows:

    $ xcube serve --config s3://cyanoalert/xcube/demo.yaml ...

  • A new endpoint /viewer/config/{*path} allows for configuring the viewer accessible via endpoint /viewer. The actual source for the configuration items is configured by xcube server configuration using the new entry Viewer/Configuration/Path, for example:

    Viewer:
      Configuration:
        Path: s3://cyanoalert/xcube/viewer-config

  • A typical xcube server configuration comprises many paths, and relative paths of known configuration parameters are resolved against the base_dir configuration parameter. However, for values of parameters passed to user functions that represent paths in user code, this cannot be done automatically. For such situations, expressions can be used. An expression is any string between "${" and "}" in a configuration value. An expression can contain the variables base_dir (a string), ctx the current server context (type xcube.webapi.datasets.DatasetsContext), as well as the function resolve_config_path(path) that is used to make a path absolut with respect to base_dir and to normalize it. For example

    Augmentation:
      Path: augmentation/metadata.py
      Function: metadata:update_metadata
      InputParameters:
        bands_config: ${resolve_config_path("../common/bands.yaml")}

• xcube's spatial resampling functions resample_in_space(), affine_transform_dataset(), and rectify_dataset() exported from module xcube.core.resampling now encode the target grid mapping into the resampled datasets. (#822)

  This new default behaviour can be switched off by keyword argument encode_cf=False. The grid mapping name can be set by keyword argument gm_name. If gm_name is not given a grid mapping will not be encoded if all the following conditions are true:

  • The target CRS is geographic;
  • The spatial dimension names are "lon" and "lat";
  • The spatial 1-D coordinate variables are named "lon" and "lat" and are evenly spaced.

  The encoding of the grid mapping is done according to CF conventions:

  • The CRS is encoded as attributes of a 0-D data variable named by gm_name
  • All spatial data variables receive an attribute grid_mapping that is set to the value of gm_name.

• Added Notebook xcube-viewer-in-jl.ipynb that explains how xcube Viewer can now be utilised in JupyterLab using the new (still experimental) xcube JupyterLab extension xcube-jl-ext. The xcube-jl-ext package is also available on PyPI.

• Updated example Notebook for CMEMS data store to reflect changes of parameter names that provide CMEMS API credentials.

• Included support for Azure Blob Storage filesystem by adding a new data store abfs. Many thanks to Ed! (#752)

  These changes will enable access to data cubes (.zarr or .levels) in Azure blob storage as shown here:

  store = new_data_store(
      "abfs",                    # Azure filesystem protocol
      root="my_blob_container",  # Azure blob container name
      storage_options= {'anon': True, 
                        # Alternatively, use 'connection_string': 'xxx'
                        'account_name': 'xxx', 
                        'account_key':'xxx'}  
  )
  store.list_data_ids()

  Same configuration for xcube Server:

  DataStores:
  - Identifier: siec
    StoreId: abfs
    StoreParams:
      root: my_blob_container
      max_depth: 1
      storage_options:
        anon: true
        account_name: "xxx"
        account_key': "xxx"
        # or
        # connection_string: "xxx"
    Datasets:
      - Path: "*.levels"
        Style: default

• Added Notebook 8_azure_blob_filesystem.ipynb. This notebook shows how a new data store instance can connect and list Zarr files from Azure bolb storage using the new abfs data store.

• xcube's Dockerfile no longer creates a conda environment xcube. All dependencies are now installed into the base environment making it easier to use the container as an executable for xcube applications. We are now also using a micromamba base image instead of miniconda. The result is a much faster build and smaller image size.

• Added a new_cluster function to xcube.util.dask, which can create Dask clusters with various configuration options.

• The xcube multi-level dataset specification has been enhanced. (#802)

  • When writing multi-level datasets (*.levels/) we now create a new JSON file .zlevels that contains the parameters used to create the dataset.
  • A new class xcube.core.mldataset.FsMultiLevelDataset that represents a multi-level dataset persisted to some filesystem, like "file", "s3", "memory". It can also write datasets to the filesystem.

• Changed the behaviour of the class xcube.core.mldataset.CombinedMultiLevelDataset to do what we actually expect: If the keyword argument combiner_func is not given or None is passed, a copy of the first dataset is made, which is then subsequently updated by the remaining datasets using xarray.Dataset.update(). The former default was using the xarray.merge(), which for some reason can eagerly load Dask array chunks into memory that won't be released.

Fixes

• Tiles of datasets with forward slashes in their identifiers (originated from nested directories) now display again correctly in xcube Viewer. Tile URLs have not been URL-encoded in such cases. (#817)

• The xcube server configuration parameters url_prefix and reverse_url_prefix can now be absolute URLs. This fixes a problem for relative prefixes such as "proxy/8000" used for xcube server running inside JupyterLab. Here, the expected returned self-referencing URL was https://{host}/users/{user}/proxy/8000/{path} but we got http://{host}/proxy/8000/{path}. (#806)

Changes in 0.13.0

Enhancements

• xcube Server has been rewritten almost from scratch.

  • Introduced a new endpoint ${server_url}/s3 that emulates and AWS S3 object storage for the published datasets. (#717) The bucket name can be either:

    • s3://datasets - publishes all datasets in Zarr format.
    • s3://pyramids - publishes all datasets in a multi-level levels format (multi-resolution N-D images) that comprises level datasets in Zarr format.

    Datasets published through the S3 API are slightly renamed for clarity. For bucket s3://pyramids:

    • if a dataset identifier has suffix .levels, the identifier remains;
    • if a dataset identifier has suffix .zarr, it will be replaced by .levels only if such a dataset doesn't exist;
    • otherwise, the suffix .levels is appended to the identifier. For bucket s3://datasets the opposite is true:
    • if a dataset identifier has suffix .zarr, the identifier remains;
    • if a dataset identifier has suffix .levels, it will be replaced by .zarr only if such a dataset doesn't exist;
    • otherwise, the suffix .zarr is appended to the identifier.

    With the new S3 endpoints in place, xcube Server instances can be used as xcube data stores as follows:

    store = new_data_store(
        "s3", 
        root="datasets",   # bucket "datasets", use also "pyramids"
        max_depth=2,       # optional, but we may have nested datasets
        storage_options=dict(
            anon=True,
            client_kwargs=dict(
                endpoint_url='http://localhost:8080/s3' 
            )
        )
    )

  • The limited s3bucket endpoints are no longer available and are replaced by s3 endpoints.

  • Added new endpoint /viewer that serves a self-contained, packaged build of xcube Viewer. The packaged viewer can be overridden by environment variable XCUBE_VIEWER_PATH that must point to a directory with a build of a compatible viewer.

  • The --show option of xcube serve has been renamed to --open-viewer. It now uses the self-contained, packaged build of xcube Viewer. (#750)

  • The --show option of xcube serve now outputs various aspects of the server configuration.

  • Added experimental endpoint /volumes. It is used by xcube Viewer to render 3-D volumes.

• xcube Server is now more tolerant with respect to datasets it can not open without errors. Implementation detail: It no longer fails if opening datasets raises any exception other than DatasetIsNotACubeError. (#789)

• xcube Server's colormap management has been improved in several ways:

  • Colormaps are no longer managed globally. E.g., on server configuration change, new custom colormaps are reloaded from files.
  • Colormaps are loaded dynamically from underlying matplotlib and cmocean registries, and custom SNAP color palette files. That means, latest matplotlib colormaps are now always available. (#687)
  • Colormaps can now be reversed (name suffix "_r"), can have alpha blending (name suffix "_alpha"), or both (name suffix "_r_alpha").
  • Loading of custom colormaps from SNAP *.cpd has been rewritten. Now also the isLogScaled property of the colormap is recognized. (#661)
  • The module xcube.util.cmaps has been redesigned and now offers three new classes for colormap management:
    • Colormap - a colormap
    • ColormapCategory - represents a colormap category
    • ColormapRegistry - manages colormaps and their categories

• The xcube filesystem data stores such as "file", "s3", "memory" can now filter the data identifiers reported by get_data_ids(). (#585) For this purpose, the data stores now accept two new optional keywords which both can take the form of a wildcard pattern or a sequence of wildcard patterns:

  1. excludes: if given and if any pattern matches the identifier, the identifier is not reported.
  2. includes: if not given or if any pattern matches the identifier, the identifier is reported.

• Added convenience method DataStore.list_data_ids() that works like get_data_ids(), but returns a list instead of an iterator. (#776)

• Replaced usages of deprecated numpy dtype numpy.bool by Python type bool.

Fixes

• xcube CLI tools no longer emit warnings when trying to import installed packages named xcube_* as xcube plugins.

• The xcube.util.timeindex module can now handle 0-dimensional ndarrays as indexers. This effectively avoids the warning Can't determine indexer timezone; leaving it unmodified. which was emitted in such cases.

• xcube serve will now also accept datasets with coordinate names longitude and latitude, even if the attribute long_name isn't set. (#763)

• Function xcube.core.resampling.affine.affine_transform_dataset() now assumes that geographic coordinate systems are equal by default and hence a resampling based on an affine transformation can be performed.

• Fixed a problem with xcube server's WMTS implementation. For multi-level resolution datasets with very coarse low resolution levels, the tile matrix sets WorldCRS84Quad and WorldWebMercatorQuad have reported a negative minimum z-level.

• Implementation of function xcube.core.geom.rasterize_features() has been changed to account for consistent use of a target variable's fill_value and dtype for a given feature. In-memory (decoded) variables now always use dtype float64 and use np.nan to represent missing values. Persisted (encoded) variable data will make use of the target fill_value and dtype. (#778)

• Relative local filesystem paths to datasets are now correctly resolved against the base directory of the xcube Server's configuration, i.e. configuration parameter base_dir. (#758)

• Fixed problem with xcube gen raising FileNotFoundError with Zarr >= 2.13.

• Provided backward compatibility with Python 3.8. (#760)

Other

• The CLI tool xcube edit has been deprecated in favour of the xcube patch. (#748)

• Deprecated CLI xcube tile has been removed.

• Deprecated modules, classes, methods, and functions have finally been removed:

  • xcube.core.geom.get_geometry_mask()
  • xcube.core.mldataset.FileStorageMultiLevelDataset
  • xcube.core.mldataset.open_ml_dataset()
  • xcube.core.mldataset.open_ml_dataset_from_local_fs()
  • xcube.core.mldataset.open_ml_dataset_from_object_storage()
  • xcube.core.subsampling.get_dataset_subsampling_slices()
  • xcube.core.tiledimage
  • xcube.core.tilegrid

• The following classes, methods, and functions have been deprecated:

  • xcube.core.xarray.DatasetAccessor.levels()
  • xcube.util.cmaps.get_cmap()
  • xcube.util.cmaps.get_cmaps()

• A new function compute_tiles() has been refactored out from function xcube.core.tile.compute_rgba_tile().

• Added method get_level_for_resolution(xy_res) to abstract base class xcube.core.mldataset.MultiLevelDataset.

• Removed outdated example resources from examples/serve/demo.

• Account for different spatial resolutions in x and y in xcube.core.geom.get_dataset_bounds().

• Make code robust against 0-size coordinates in xcube.core.update._update_dataset_attrs().

• xcube Server has been enhanced to load multi-module Python code for dynamic cubes both from both directories and zip archives. For example, the following dataset definition computes a dynamic cube from dataset "local" using function "compute_dataset" in Python module "resample_in_time.py":

    Path: resample_in_time.py
    Function: compute_dataset
    InputDatasets: ["local"]

  Users can now pack "resample_in_time.py" among any other modules and packages into a zip archive. Note that the original module name is now a prefix to the function name:

    Path: modules.zip
    Function: resample_in_time:compute_dataset
    InputDatasets: ["local"]

  Implementation note: this has been achieved by using xcube.core.byoa.CodeConfig in xcube.core.mldataset.ComputedMultiLevelDataset.

• Instead of the Function keyword it is now possible to use the Class keyword. While Function references a function that receives one or more datasets (type xarray.Dataset) and returns a new one, Class references a callable that receives one or more multi-level datasets and returns a new one. The callable is either a class derived from
  or a function that returns an instance of xcube.core.mldataset.MultiLevelDataset.

• Module xcube.core.mldataset has been refactored into a sub-package for clarity and maintainability.

• Removed deprecated example examples/tile.

Other Changes

• The utility function xcube.util.dask.create_cluster() now also generates the tag user for the current user's name.

Changes in 0.12.1

Enhancements

• Added a new package xcube.core.zarrstore that exports a number of useful Zarr store implementations and Zarr store utilities:

  • xcube.core.zarrstore.GenericZarrStore comprises user-defined, generic array definitions. Arrays will compute their chunks either from a function or a static data array.

  • xcube.core.zarrstore.LoggingZarrStore is used to log Zarr store access performance and therefore useful for runtime optimisation and debugging.

  • xcube.core.zarrstore.DiagnosticZarrStore is used for testing Zarr store implementations.

  • Added a xarray dataset accessor xcube.core.zarrstore.ZarrStoreHolder that enhances instances of xarray.Dataset by a new property zarr_store. It holds a Zarr store instance that represents the datasets as a key-value mapping. This will prepare later versions of xcube Server for publishing all datasets via an emulated S3 API.

    In turn, the classes of module xcube.core.chunkstore have been deprecated.

• Added a new function xcube.core.select.select_label_subset() that is used to select dataset labels along a given dimension using user-defined predicate functions.

• The xcube Python environment is now requiring xarray >= 2022.6 and zarr >= 2.11 to ensure sparse Zarr datasets can be written using dataset.to_zarr(store). (#688)

• Added new module xcube.util.jsonencoder that offers the class NumpyJSONEncoder used to serialize numpy-like scalar values to JSON. It also offers the function to_json_value() to convert Python objects into JSON-serializable versions. The new functionality is required to ensure dataset attributes that are JSON-serializable. For example, the latest version of the rioxarray package generates a _FillValue attribute with datatype np.uint8.

Fixes

• The filesystem-based data stores for the "s3", "file", and "memory" protocols can now provide xr.Dataset instances from image pyramids formats, i.e. the levels and geotiff formats.

Changes in 0.12.0

Enhancements

• Allow xcube Server to work with any OIDC-compliant auth service such as Auth0, Keycloak, or Google. Permissions of the form "read:dataset:\<dataset\>" and "read:variable:\<dataset\>" can now be passed by two id token claims:

  • permissions must be a JSON list of permissions;
  • scope must be a space-separated character string of permissions.

  It is now also possible to include id token claim values into the permissions as template variables. For example, if the currently authenticated user is demo_user, the permission "read:dataset:$username/*" will effectively be "read:dataset:demo_user/*" and only allow access to datasets with resource identifiers having the prefix demo_user/.

  With this change, server configuration has changed:

  Example of OIDC configuration for auth0

  Please note, there must be a trailing slash in the "Authority" URL.

  Authentication:
    Authority: https://some-demo-service.eu.auth0.com/
    Audience: https://some-demo-service/api/

  Example of OIDC configuration for Keycloak

  Please note, no trailing slash in the "Authority" URL.

  Authentication: 
    Authority: https://kc.some-demo-service.de/auth/realms/some-kc-realm
    Audience: some-kc-realm-xc-api

• Filesystem-based data stores like "file" and "s3" support reading GeoTIFF and Cloud Optimized GeoTIFF (COG). (#489)

• xcube server now also allows publishing also 2D datasets such as opened from GeoTIFF / COG files.

• Removed all upper version bounds of package dependencies. This increases compatibility with existing Python environments.

• A new CLI tool xcube patch has been added. It allows for in-place metadata patches of Zarr data cubes stored in almost any filesystem supported by fsspec including the protocols "s3" and "file". It also allows patching xcube multi-level datasets (*.levels format).

• In the configuration for xcube server, datasets defined in DataStores may now have user-defined identifiers. In case the path does not unambiguously define a dataset (because it contains wildcards), providing a user-defined identifier will raise an error.

Fixes

• xcube Server did not find any grid mapping if a grid mapping variable (e.g. spatial_ref or crs) encodes a geographic CRS (CF grid mapping name "latitude_longitude") and the related geographical 1-D coordinates were named "x" and "y". (#706)
• Fixed typo in metadata of demo cubes in examples/serve/demo. Demo cubes now all have consolidated metadata.
• When writing multi-level datasets with file data stores, i.e.,

  store.write_data(dataset, data_id="test.levels", use_saved_levels=True)

  and where dataset has different spatial resolutions in x and y, an exception was raised. This is no longer the case.
• xcube Server can now also compute spatial 2D datasets from users' Python code. In former versions, spatio-temporal 3D cubes were enforced.

Other important changes

• Deprecated all functions and classes defined in xcube.core.dsio in favor of the xcube data store API defined by xcube.core.store.

Changes in 0.11.2

Enhancements

• xcube serve now provides new metadata details of a dataset:

  • The spatial reference is now given by property spatialRef and provides a textual representation of the spatial CRS.
  • The dataset boundary is now given as property geometry and provides a GeoJSON Polygon in geographic coordinates.

• xcube serve now publishes the chunk size of a variable's time dimension for either for an associated time-chunked dataset or the dataset itself (new variable integer property timeChunkSize). This helps clients (e.g. xcube Viewer) to improve the server performance for time-series requests.

• The functions

  • mask_dataset_by_geometry()
  • rasterize_features() of module xcube.core.geom have been reimplemented to generate lazy dask arrays. Both should now be applicable to datasets that have arbitrarily large spatial dimensions. The spatial chunk sizes to be used can be specified using keyword argument tile_size. (#666)

Fixes

• Fixed ESA CCI example notebook. (#680)

• xcube serve now provides datasets after changes of the service configuration while the server is running. Previously, it was necessary to restart the server to load the changes. (#678)

Other changes

• xcube.core.resampling.affine_transform_dataset() has a new keyword argument reuse_coords: bool = False. If set to True the returned dataset will reuse the same spatial coordinates as the target. This is a workaround for xarray issue pydata/xarray#6573.

• Deprecated following functions of module xcube.core.geom:

  • is_dataset_y_axis_inverted() is no longer used;
  • get_geometry_mask() is no longer used;
  • convert_geometry() has been renamed to normalize_geometry().

Changes in 0.11.1

• Fixed broken generation of composite RGBA tiles. (#668)
• Fixing broken URLs in xcube viewer documentation, more revision still needed.

Changes in 0.11.0

Enhancements

• xcube serve can now serve datasets with arbitrary spatial coordinate reference systems. Before xcube 0.11, datasets where forced to have a geographical CRS such as EPSG:4326 or CRS84.

• xcube serve can now provide image tiles for two popular tile grids:

  1. global geographic grid, with 2 x 1 tiles at level zero (the default);
  2. global web mercator grid, with 1 x 1 tiles at level zero ("Google projection", OSM tile grid).

  The general form of the new xcube tile URL is (currently)

  /datasets/{ds_id}/vars/{var_name}/tile2/{z}/{y}/{x}
  

  The following query parameters can be used

  • crs: set to CRS84 to use the geographical grid (the default), or EPSG:3857 to use the web mercator grid.
  • cbar: color bar name such as viridis or plasma, see color bar names of matplotlib. Defaults to bone.
  • vmin: minimum value to be used for color mapping. Defaults to 0.
  • vmax: maximum value to be used for color mapping. Defaults to 1.
  • retina: if set to 1, tile size will be 512 instead of 256.

• The WMTS provided by xcube serve has been reimplemented from scratch. It now provides two common tile matrix sets:

  1. WorldCRS84Quad global geographic grid, with 2 x 1 tiles at level zero;
  2. WorldWebMercatorQuad global web mercator grid, with 1 x 1 tiles at level zero.

  New RESTful endpoints have been added to reflect this:

  /wmts/1.0.0/{TileMatrixSet}/WMTSCapabilities.xml
  /wmts/1.0.0/tile/{Dataset}/{Variable}/{TileMatrixSet}/{TileMatrix}/{TileRow}/{TileCol}.png
  

  The existing RESTful endpoints now use tile matrix set WorldCRS84Quad by default:

  /wmts/1.0.0/WMTSCapabilities.xml
  /wmts/1.0.0/tile/{Dataset}/{Variable}/{TileMatrix}/{TileRow}/{TileCol}.png
  

  The key-value pair (KVP) endpoint /wmts/kvp now recognises the TileMatrixSet key for the two values described above.

• Support for multi-level datasets aka ND image pyramids has been further improved (#655):

  • Introduced new parameter agg_methods for writing multi-level datasets with the "file", "s3", and "memory" data stores. The value of agg_methods is either a string "first", "min", "max", "mean", "median" or a dictionary that maps a variable name to an aggregation method. Variable names can be patterns that may contain wildcard characters '*' and '?'. The special aggregation method "auto" can be used to select "first" for integer variables and "mean" for floating point variables.
  • The xcube level CLI tool now has a new option --agg-methods (or -A) for the same purpose.

• The xcube package now consistently makes use of logging. We distinguish general logging and specific xcube logging. General logging refers to the log messages emitted by any Python module while xcube logging only refers to log messages emitted by xcube modules.

  • The output of general logging from xcube CLI tools can now be configured with two new CLI options:

    • --loglevel LEVEL: Can be one of CRITICAL, ERROR, WARNING, INFO, DETAIL, DEBUG, TRACE, or OFF (the default).
    • --logfile PATH: Effective only if log level is not OFF. If given, log messages will be written into the file given by PATH. If omitted, log messages will be redirected to standard error (sys.stderr).

    The output of general logging from xcube CLI is disabled by default. If enabled, the log message format includes the level, date-time, logger name, and message.

  • All xcube modules use the logger named xcube (i.e. LOG = logging.getLogger("xcube")) to emit messages regarding progress, debugging, errors. Packages that extend the xcube package should use a dot suffix for their logger names, e.g. xcube.cci for the xcube plugin package xcube-cci.

  • All xcube CLI tools will output log messages, if any, on standard error (sys.stderr). Only the actual result, if any, is written to standard out (sys.stdout).

  • Some xcube CLI tools have a --quiet/-q option to disable output of log messages on the console and a --verbose/-v option to enable it and control the log level. For this purpose the option -v can be given multiple times and even be combined: -v = INFO, -vv = DETAIL, -vvv = DEBUG, -vvvv = TRACE. The quiet and verbose settings only affect the logger named xcube and its children. If enabled, a simple message format will be used, unless the general logging is redirected to stdout.

Fixes

• Fixed a problem where the DataStores configuration of xcube serve did not recognize multi-level datasets. (#653)

• Opening of multi-level datasets with filesystem data stores now recognizes the cache_size open parameter.

• It is possible again to build and run docker containers from the docker file in the Github Repository. (#651) For more information, see https://xcube.readthedocs.io/en/latest/installation.html#docker

Other changes

• The xcube tile CLI tool has been deprecated. A new tool is planned that can work concurrently on dask clusters and also supports common tile grids such as global geographic and web mercator.

• The xcube.util.tiledimage module has been deprecated and is no longer used in xcube. It has no replacement.

• The xcube.util.tilegrid module has been deprecated and is no longer used in xcube. A new implementation is provided by xcube.core.tilingscheme which is used instead.

• All existing functions of the xcube.core.tile module have been deprecated and are no longer used in xcube. A newly exported function is xcube.core.tile.compute_rgba_tile() which is used in place of other tile generating functions.

Changes in 0.10.2

Enhancements

• Added new module xcube.core.subsampling for function subsample_dataset(dataset, step) that is now used by default to generate the datasets level of multi-level datasets.

• Added new setting Authentication.IsRequired to the xcube serve configuration. If set to true, xcube Server will reject unauthorized dataset requests by returning HTTP code 401.

• For authorized clients, the xcube Web API provided by xcube serve now allows granted scopes to contain wildcard characters *, **, and ?. This is useful to give access to groups of datasets, e.g. the scope read:dataset:*/S2-*.zarr permits access to any Zarr dataset in a subdirectory of the configured data stores and whose name starts with "S2-". (#632)

• xcube serve used to shut down with an error message if it encountered datasets it could not open. New behaviour is to emit a warning and ignore such datasets. (#630)

• Introduced helper function add_spatial_ref() of package xcube.core.gridmapping.cfconv that allows adding a spatial coordinate reference system to an existing
  Zarr dataset. (#629)

• Support for multi-level datasets has been improved:

  • Introduced new parameters for writing multi-level datasets with the "file", "s3", and "memory" data stores (#617). They are
    • base_dataset_id: If given, the base dataset will be linked only with the value of base_dataset_id, instead of being copied as-is. This can save large amounts of storage space.
    • tile_size: If given, it forces the spatial dimensions to be chunked accordingly. tile_size can be a positive integer or a pair of positive integers.
    • num_levels: If given, restricts the number of resolution levels to the given value. Must be a positive integer to be effective.
  • Added a new example notebook 5_multi_level_datasets.ipynb that demonstrates writing and opening multi-level datasets with the xcube filesystem data stores.
  • Specified xcube Multi-Resolution Datasets definition and format.

• xcube gen2 returns more expressive error messages.

Fixes

• Fixed problem where the dataset levels of multi-level datasets were written without spatial coordinate reference system. In fact, only spatial variables were written. (#646)

• Fixed problem where xcube Server instances that required user authentication published datasets and variables for unauthorised users.

• Fixed FsDataAccessor.write_data() implementations, which now always return the passed in data_id. (#623)

• Fixes an issue where some datasets seemed to be shifted in the y-(latitude-) direction and were misplaced on maps whose tiles are served by xcube serve. Images with ascending y-values are now tiled correctly. (#626)

Other

• The xcube level CLI tool has been rewritten from scratch to make use of xcube filesystem data stores. (#617)

• Deprecated numerous classes and functions around multi-level datasets. The non-deprecated functions and classes of xcube.core.mldataset should be used instead along with the xcube filesystem data stores for multi-level dataset i/o. (#516)

  • Deprecated all functions of the xcube.core.level module
    • compute_levels()
    • read_levels()
    • write_levels()
  • Deprecated numerous classes and functions of the xcube.core.mldataset module
    • FileStorageMultiLevelDataset
    • ObjectStorageMultiLevelDataset
    • open_ml_dataset()
    • open_ml_dataset_from_object_storage()
    • open_ml_dataset_from_local_fs()
    • write_levels()

• Added packages python-blosc and lz4 to the xcube Python environment for better support of Dask distributed and the Dask service Coiled.

• Replace the dependency on the rfc3339-validator PyPI package with a dependency on its recently created conda-forge package.

• Remove unneeded dependency on the no longer used strict-rfc3339 package.

Changes in 0.10.1

Fixes

• Deprecated argument xy_var_names in function GridMapping.from_dataset, thereby preventing a NotImplementedError. (#551)

Other Changes

• For compatibility, now also xcube.__version__ contains the xcube version number.

Changes in 0.10.0

Incompatible Changes

• The configuration DataStores for xcube serve changed in an incompatible way with xcube 0.9.x: The value of former Identifier must now be assigned to Path, which is a mandatory parameter. Path may contain wildcard characters **, *, ?. Identifier is now optional, the default is "${store_id}~${data_id}". If given, it should only be used to uniquely identify single datasets within a data store pointed to by Path. (#516)

Enhancements

• It is now possible to use environment variables in most
  xcube configuration files. Unix bash syntax is used, i.e. ${ENV_VAR_NAME} or $ENV_VAR_NAME. (#580)

  Supported tools include

  • xcube gen --config CONFIG
  • xcube gen2 --stores STORES_CONFIG --service SERVICE_CONFIG
  • xcube serve -c CONFIG

• Changed the xcube gen tool to extract metadata for pre-sorting inputs from other than NetCDF inputs, e.g. GeoTIFF.

• Optimized function xcube.core.geom.rasterize_features(). It is now twice as fast while its memory usage dropped to the half. (#593)

Fixes

• xcube serve now also serves datasets that are located in subdirectories of filesystem-based data stores such as "file", "s3", "memory". (#579)

• xcube serve now accepts datasets whose spatial resolutions differ up to 1%. (#590) It also no longer rejects datasets with large dimension sizes. (Formerly, an integer-overflow occurred in size computation.)

• DatasetChunkCacheSize is now optional in xcube serve configuration. (Formerly, when omitted, the server crashed.)

• Fixed bug that would cause that requesting data ids on some s3 stores would fail with a confusing ValueError.

• Fixed that only last dataset of a directory listing was published via xcube serve when using the DataStores configuration with filesystem-based datastores such as "s3" or "file". (#576)

Other

• Pinned Python version to < 3.10 to avoid import errors caused by a third-party library.

• Values obs and local for the FileSystem parameter in xcube configuration files have been replaced by s3 and file, but are kept temporarily for the sake of backwards compatibility.

Changes in 0.9.2

Fixes

• A xcube.core.store.fs.impl.FSDataStore no longer raises exceptions when root directories in data store configurations do not exist. Instead, they are created when data is written.

Changes in 0.9.1

New features

• The xcube.core.maskset.MaskSet class no longer allocates static numpy arrays for masks. Instead, it uses lazy dask arrays. (#556)

• Function xcube.core.geom.mask_dataset_by_geometry has a new parameter all_touched: If True, all pixels intersected by geometry outlines will be included in the mask. If False, only pixels whose center is within the polygon or that are selected by Bresenham’s line algorithm will be included
  in the mask. The default value is set to False.

Other

• Updated Dockerfile: Removed the usage of a no-longer-maintained base image. Ensured that the version tag 'latest' can be used with installation mode 'release' for xcube plugins.

• The xcube package now requires xarray >= 0.19, zarr >= 2.8, pandas >= 1.3.

Changes in 0.9.0

New features

• The implementations of the default data stores s3, directory, and memory have been replaced entirely by a new implementation that utilize the fsspec Python package. The preliminary filesystem-based data stores are now s3, file, and memory. All share a common implementations and tests. Others filesystem-based data stores can be added easily and will follow soon, for example hdfs. All filesystem-based data stores now support xarray datasets (type xarray.Dataset) in Zarr and NetCDF format as well as image pyramids (typexcube.core.multilevel.MultiLevelDataset) using a Zarr-based multi-level format. (#446)

• Several changes became necessary on the xcube Generator package xcube.core.gen2 and CLI xcube gen2. They are mostly not backward compatible:

  • The only supported way to instantiate cube generators is the CubeGenerator.new() factory method.
  • CubeGenerator.generate_cube() and CubeGenerator.get_cube_info() both now receive the request object that has formerly been passed to the generator constructors.
  • The CubeGenerator.generate_cube() method now returns a CubeGeneratorResult object rather than a simple string (the written data_id).
  • Empty cubes are no longer written, a warning status is generated instead.
  • The xcube gen2 CLI xcube gen2 has a new option --output RESULT to write the result to a JSON file. If it is omitted, the CLI will dump the result as JSON to stdout.

• Numerous breaking changes have been applied to this version in order to address generic resampling (#391), to support other CRS than WGS-84 (#112), and to move from the struct data cube specification to a more relaxed cube convention (#488):

  • The following components have been removed entirely
    • module xcube.core.imgeom with class ImageGeom
    • module xcube.core.geocoding with class GeoCoding
    • module xcube.core.reproject and all its functions
  • The following components have been added
    • module xcube.core.gridmapping with new class GridMapping is a CF compliant replacement for classes ImageGeom and GeoCoding
  • The following components have changed in an incompatible way:
    • Functionxcube.core.rectify.rectify_dataset() now uses source_gm: GridMapping and target_gm: GridMapping instead of geo_coding: GeoCoding and output_geom: ImageGeom.
    • Functionxcube.core.gen.iproc.InputProcessor.process() now uses source_gm: GridMapping and target_gm: GridMapping instead of geo_coding: GeoCoding and output_geom: ImageGeom.
  • xcube no longer depends on GDAL (at least not directly).

• Added a new feature to xcube called "BYOA" - Bring your own Algorithm. It is a generic utility that allows for execution of user-supplied Python code in both local and remote contexts. (#467) The new xcube.core.byoa package hosts the BYOA implementation and API. The entry point to the functionality is the xcube.core.byoa.CodeConfig class. It is currently utilized by the xcube Cube Generator that can now deal with an optional code_config request parameter. If given, the generated data cube will be post-processed by the configured user-code. The xcube Cube Generator with the BYOA feature is made available through the

  1. Generator API xcube.core.gen2.LocalCubeGenerator and xcube.core.gen2.service.RemoteCubeGenerator;
  2. Generator CLI xcube gen2.

• A dataset's cube subset and its grid mapping can now be accessed through the xcube property of xarray.Dataset instances. This feature requires importing the xcube.core.xarraypackage. Let dataset be an instance of xarray.Dataset, then

  • dataset.xcube.cube is a xarray.Dataset that contains all cube variables of dataset, namely the ones with dimensions ("time", [...,], y_dim_name, x_dim_name), where y_dim_name, x_dim_name are determined by the dataset's grid mapping. May be empty, if dataset has no cube variables.
  • dataset.xcube.gm is a xcube.core.gridmapping.GridMapping that describes the CF-compliant grid mapping of dataset. May be None, if dataset does not define a grid mapping.
  • dataset.xcube.non_cube is a xarray.Dataset that contains all variables of dataset that are not in dataset.xcube.cube. May be same as dataset, if dataset.xcube.cube is empty.

• Added a new utility module xcube.util.temp that allows for creating temporary files and directories that will be deleted when the current process ends.

• Added function xcube.util.versions.get_xcube_versions()
  that outputs the versions of packages relevant for xcube. Also added a new CLI xcube versions that outputs the result of the
  new function in JSON or YAML. (#522)

Other

• The xcube cube generator (API xcube.core.gen2, CLI xcube gen2) will now write consolidated Zarrs by default. (#500)

• xcube now issues a warning, if a data cube is opened from object storage, and credentials have neither been passed nor can be found, and the object storage has been opened with the default anon=False. (#412)

• xcube no longer internally caches directory listings, which prevents the situation where a data cube that has recently been written into object storage cannot be found.

• Removed example notebooks that used hard-coded local file paths. (#400)

• Added a GitHub action that will run xcube unit tests, and build and push Docker images. The version tag of the image is either latest when the main branch changed, or the same as the release tag.

• Removed warning module 'xcube_xyz' looks like an xcube-plugin but lacks a callable named 'init_plugin.

• Fixed an issue where xcube serve provided wrong layer source options for OpenLayers XYZ when latitude coordinates where increasing with the coordinate index. (#251)

• Function xcube.core.normalize.adjust_spatial_attrs() no longer removes existing global attributes of the form geospatial_vertical_<property>.

• Numerous classes and functions became obsolete in the xcube 0.9 code base and have been removed, also because we believe there is quite rare outside use, if at all.

  Removed from xcube.util.tiledimage:

  • class DownsamplingImage
  • class PilDownsamplingImage
  • class NdarrayDownsamplingImage
  • class FastNdarrayDownsamplingImage
  • class ImagePyramid
  • function create_pil_downsampling_image()
  • function create_ndarray_downsampling_image()
  • function downsample_ndarray()
  • functions aggregate_ndarray_xxx()

  Removed from xcube.util.tilegrid:

  • functions pow2_2d_subdivision()
  • functions pow2_1d_subdivision()

Changes in 0.8.2

• Fixed the issue that xcube gen2 would not print tracebacks to stderr when raising errors of type CubeGeneratorError (#448).
• Enhanced xcube.core.normalize.normalize_dataset() function to also normalize datasets with latitudes given as latitude_centers and to invert decreasing latitude coordinate values.
• Introduced xcube.core.normalize.cubify_dataset() function to normalize a dataset and finally assert the result complies to the xcube dataset conventions.
• Fixed that data stores directory and s3 were not able to handle data identifiers that they had assigned themselves during write_data().
  (#450)
• The xcube prune tool is no longer restricted to data cube datasets and should now be able to deal with datasets that comprise very many chunks. (#469)
• The xcube.core.extract.get_cube_values_for_points() function has been enhanced to also accept lists or tuples in the item values of the points arguments. (#431)
• Fixed exception raised in xcube extract CLI tool when called with the --ref option. This issue occurred with xarray 0.18.2+.

Changes in 0.8.1

• Improved support of datasets with time given as cftime.DatetimeGregorian or cftime.DatetimeJulian.

• Fixed out-of-memory error raised if spatial subsets were created from cubes with large spatial dimensions. (#442)

• Fixed example Notebook compute_dask_array and renamed it into compute_array_from_func. (#385)

• Fixed a problem with the S3 data store that occurred if the store was configured without bucket_name and the (Zarr) data was opened with consolidated=True.

• The functions xcube.core.compute.compute_cube() and xcube.core.compute.compute_dataset() can now alter the shape of input datasets. (#289)

Changes in 0.8.0

• Harmonized retrieval of spatial and temporal bounds of a dataset: To determine spatial bounds, use xcube.core.geom.get_dataset_bounds(), to determine temporal bounds, use xcube.core.timecoord.get_time_range_from_data(). Both methods will attempt to get the values from associated bounds arrays first.
• Fixed broken JSON object serialisation of objects returned by DataStore.describe_object(). (#432)
• Changed behaviour and signature of xcube.core.store.DataStore.get_dataset_ids(). The keyword argument include_titles: str = True has been replaced by include_attrs: Sequence[str] = None and the return value changes accordingly:
  • If include_attrs is None (the default), the method returns an iterator of dataset identifiers data_id of type str.
  • If include_attrs is a sequence of attribute names, the method returns an iterator of tuples (data_id, attrs) of type Tuple[str, Dict]. Hence include_attrs can be used to obtain a minimum set of dataset metadata attributes for each returned data_id. However, include_attrs is not yet implemented so far in the "s3", "memory", and "directory" data stores. (#420)
• Directory and S3 Data Store consider format of data denoted by data id when using get_opener_ids().
• S3 Data Store will only recognise a consolidated = True parameter setting, if the file {bucket}/{data_id}/.zmetadata exists.
• xcube gen2 will now ensure that temporal subsets can be created. (#430)
• Enhance xcube serve for use in containers: (#437)
  • In addition to option --config or -c, dataset configurations can now be passed via environment variable XCUBE_SERVE_CONFIG_FILE.
  • Added new option --base-dir or -b to pass the base directory to resolve relative paths in dataset configurations. In addition, the value can be passed via environment variable XCUBE_SERVE_BASE_DIR.

Changes in 0.7.2

• xcube gen2 now allows for specifying the final data cube's chunk sizes. The new cube_config parameter is named chunks, is optional and if given, must be a dictionary that maps a dimension name to a chunk size or to None (= no chunking). The chunk sizes only apply to data variables. Coordinate variables will not be affected, e.g. "time", "lat", "lon" will not be chunked. (#426)

• xcube gen2 now creates subsets from datasets returned by data stores that do not recognize cube subset parameters variable_names, bbox, and time_range. (#423)

• Fixed a problem where S3 data store returned outdated bucket items. (#422)

Changes in 0.7.1

• Dataset normalisation no longer includes reordering increasing latitude coordinates, as this creates datasets that are no longer writable to Zarr. (#347)
• Updated package requirements
  • Added s3fs requirement that has been removed by accident.
  • Added missing requirements requests and urllib3.

Changes in 0.7.0

• Introduced abstract base class xcube.util.jsonschema.JsonObject which is now the super class of many classes that have JSON object representations. In Jupyter notebooks, instances of such classes are automatically rendered as JSON trees.
• xcube gen2 CLI tool can now have multiple -v options, e.g. -vvv will now output detailed requests and responses.
• Added new Jupyter notebooks in examples/notebooks/gen2 for the data cube generators in the package xcube.core.gen2.
• Fixed a problem in JsonArraySchema that occurred if a valid instance was None. A TypeError TypeError: 'NoneType' object is not iterable was raised in this case.
• The S3 data store xcube.core.store.stores.s3.S3DataStore now implements the describe_data() method. It therefore can also be used as a data store from which data is queried and read.
• The xcube gen2 data cube generator tool has been hidden from the set of "official" xcube tools. It is considered as an internal tool that is subject to change at any time until its interface has stabilized. Please refer to xcube gen2 --help for more information.
• Added coords property to DatasetDescriptor class. The data_vars property of the DatasetDescriptor class is now a dictionary.
• Added chunks property to VariableDescriptor class.
• Removed function reproject_crs_to_wgs84() and tests (#375) because
  • it seemed to be no longer be working with GDAL 3.1+;
  • there was no direct use in xcube itself;
  • xcube plans to get rid of GDAL dependencies.
• CLI tool xcube gen2 may now also ingest non-cube datasets.
• Fixed unit tests broken by accident. (#396)
• Added new context manager xcube.util.observe_dask_progress() that can be used to observe tasks that known to be dominated by Dask computations:

  with observe_dask_progress('Writing dataset', 100):
      dataset.to_zarr(store)  

• The xcube normalisation process, which ensures that a dataset meets the requirements of a cube, internally requested a lot of data, causing the process to be slow and expensive in terms of memory consumption. This problem was resolved by avoiding to read in these large amounts of data. (#392)

Changes in 0.6.1

• Updated developer guide (#382)

Changes relating to maintenance of xcube's Python environment requirements in envrionment.yml:

• Removed explicit blas dependency (which required MKL as of blas =*.*=mkl) for better interoperability with existing environments.
• Removed restrictions of fsspec <=0.6.2 which was required due to Zarr #650. As #650 has been fixed, zarr=2.6.1 has been added as new requirement. (#360)

Changes in 0.6.0

Enhancements

• Added four new Jupyter Notebooks about xcube's new Data Store Framework in examples/notebooks/datastores.

• CLI tool xcube io dump now has new --config and --type options. (#370)

• New function xcube.core.store.get_data_store() and new class xcube.core.store.DataStorePool allow for maintaining a set of pre-configured data store instances. This will be used in future xcube tools that utilise multiple data stores, e.g. "xcube gen", "xcube serve". (#364)

• Replaced the concept of type_id used by several xcube.core.store.DataStore methods by a more flexible type_specifier. Documentation is provided in docs/source/storeconv.md.

  The DataStore interface changed as follows:

  • class method get_type_id() replaced by get_type_specifiers() replaces get_type_id();
  • new instance method get_type_specifiers_for_data();
  • replaced keyword-argument in get_data_ids();
  • replaced keyword-argument in has_data();
  • replaced keyword-argument in describe_data();
  • replaced keyword-argument in get_search_params_schema();
  • replaced keyword-argument in search_data();
  • replaced keyword-argument in get_data_opener_ids().

  The WritableDataStore interface changed as follows:

  • replaced keyword-argument in get_data_writer_ids().

• The JSON Schema classes in xcube.util.jsonschema have been extended:

  • date and date-time formats are now validated along with the rest of the schema
  • the JsonDateSchema and JsonDatetimeSchema subclasses of JsonStringSchema have been introduced, including a non-standard extension to specify date and time limits

• Extended xcube.core.store.DataStore docstring to include a basic convention for store open parameters. (#330)

• Added documentation for the use of the open parameters passed to xcube.core.store.DataOpener.open_data().

Fixes

• xcube serve no longer crashes, if configuration is lacking a Styles entry.

• xcube gen can now interpret start_date and stop_date from NetCDF dataset attributes. This is relevant for using xcube gen for Sentinel-2 Level 2 data products generated and provided by Brockmann Consult. (#352)

• Fixed both xcube.core.dsio.open_cube() and open_dataset() which failed with message "ValueError: group not found at path ''" if called with a bucket URL but no credentials given in case the bucket is not publicly readable. (#337) The fix for that issue now requires an additional s3_kwargs parameter when accessing datasets in public buckets:

  from xcube.core.dsio import open_cube 
    
  public_url = "https://s3.eu-central-1.amazonaws.com/xcube-examples/OLCI-SNS-RAW-CUBE-2.zarr"
  public_cube = open_cube(public_url, s3_kwargs=dict(anon=True))

• xcube now requires s3fs >= 0.5 which implies using faster async I/O when accessing object storage.

• xcube now requires gdal >= 3.0. (#348)

• xcube now only requires matplotlib-base package rather than matplotlib. (#361)

Other

• Restricted s3fs version in envrionment.yml in order to use a version which can handle pruned xcube datasets. This restriction will be removed once changes in zarr PR zarr-developers/zarr-python#650 are merged and released. (#360)
• Added a note in the xcube chunk CLI help, saying that there is a possibly more efficient way to (re-)chunk datasets through the dedicated tool "rechunker", see https://rechunker.readthedocs.io (thanks to Ryan Abernathey for the hint). (#335)
• For xcube serve dataset configurations where FileSystem: obs, users must now also specify Anonymous: True for datasets in public object storage buckets. For example:

  - Identifier: "OLCI-SNS-RAW-CUBE-2"
    FileSystem: "obs"
    Endpoint: "https://s3.eu-central-1.amazonaws.com"
    Path: "xcube-examples/OLCI-SNS-RAW-CUBE-2.zarr"
    Anyonymous: true
    ...
  - ...

• In environment.yml, removed unnecessary explicit dependencies on proj4 and pyproj and restricted gdal version to >=3.0,<3.1.

Changes in 0.5.1

• normalize_dataset now ensures that latitudes are decreasing.

Changes in 0.5.0

New

• xcube gen2 CONFIG will generate a cube from a data input store and a user given cube configuration. It will write the resulting cube in a user defined output store.

  • Input Stores: CCIODP, CDS, SentinelHub
  • Output stores: memory, directory, S3

• xcube serve CUBE will now use the last path component of CUBE as dataset title.

• xcube serve can now be run with AWS credentials (#296).

  • In the form xcube serve --config CONFIG, a Datasets entry in CONFIG may now contain the two new keys AccessKeyId: ... and SecretAccessKey: ... given that FileSystem: obs.
  • In the form xcube serve --aws-prof PROFILE CUBE the cube stored in bucket with URL CUBE will be accessed using the credentials found in section [PROFILE] of your ~/.aws/credentials file.
  • In the form xcube serve --aws-env CUBE the cube stored in bucket with URL CUBE will be accessed using the credentials found in environment variables AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY.

• xcube has been extended by a new Data Store Framework (#307). It is provided by the xcube.core.store package. It's usage is currently documented only in the form of Jupyter Notebook examples, see examples/store/*.ipynb.

• During the development of the new Data Store Framework, some
  utility packages have been added:

  • xcube.util.jsonschema - classes that represent JSON Schemas for types null, boolean, number, string, object, and array. Schema instances are used for JSON validation, and object marshalling.
  • xcube.util.assertions - numerous assert_* functions that are used for function parameter validation. All functions raise ValueError in case an assertion is not met.
  • xcube.util.ipython - functions that can be called for better integration of objects with Jupyter Notebooks.

Enhancements

• Added possibility to specify packing of variables within the configuration of xcube gen (#269). The user now may specify a different packing variables, which might be useful for reducing the storage size of the datacubes. Currently it is only implemented for zarr format. This may be done by passing the parameters for packing as the following:

  output_writer_params: 
  
    packing: 
      analysed_sst: 
        scale_factor: 0.07324442274239326
        add_offset: -300.0
        dtype: 'uint16'
        _FillValue: 0.65535

• Example configurations for xcube gen2 were added.

Fixes

• From 0.4.1: Fixed time-series performance drop (#299).

• Fixed xcube gen CLI tool to correctly insert time slices into an existing cube stored as Zarr (#317).

• When creating an ImageGeom from a dataset, correct the height if it would otherwise give a maximum latitude >90°.

• Disable the display of warnings in the CLI by default, only showing them if a --warnings flag is given.

• Fixed a regression when running "xcube serve" with cube path as parameter (#314)

• From 0.4.3: Extended xcube serve by reverse URL prefix option.

• From 0.4.1: Fixed time-series performance drop (#299).

Changes in 0.4.3

• Extended xcube serve by reverse URL prefix option --revprefix REFPREFIX. This can be used in cases where only URLs returned by the service need to be prefixed, e.g. by a web server's proxy pass.

Changes in 0.4.2

• Fixed a problem during release process. No code changes.

Changes in 0.4.1

• Fixed time-series performance drop (#299).

Changes in 0.4.0

New

• Added new /timeseries/{dataset}/{variable} POST operation to xcube web API. It extracts time-series for a given GeoJSON object provided as body. It replaces all of the /ts/{dataset}/{variable}/{geom-type} operations. The latter are still maintained for compatibility with the "VITO viewer".

• The xcube web API provided through xcube serve can now serve RGBA tiles using the dataset/{dataset}/rgb/tiles/{z}/{y}/{x} operation. The red, green, blue channels are computed from three configurable variables and normalisation ranges, the alpha channel provides transparency for missing values. To specify a default RGB schema for a dataset, a colour mapping for the "pseudo-variable" named rbg is provided in the configuration of xcube serve:

  Datasets:
    - Identifyer: my_dataset
      Style: my_style
      ...
    ...
  Styles:
    - Identifier: my_style
      ColorMappings:
        rgb:
          Red:
            Variable: rtoa_8
            ValueRange: [0., 0.25]
          Green:
            Variable: rtoa_6
            ValueRange: [0., 0.25]
          Blue:
            Variable: rtoa_4
            ValueRange: [0., 0.25]
        ...

  Note that this concept works nicely in conjunction with the new Augmentation feature (#272) used to compute new variables that could be input to the RGB generation.

• Introduced new (ortho-)rectification algorithm allowing reprojection of satellite images that come with (terrain-corrected) geo-locations for every pixel.

  • new CLI tool xcube rectify
  • new API function xcube.core.rectify.rectify_dataset()

• Utilizing the new rectification in xcube gen tool. It is now the default reprojection method in xcube.core.gen.iproc.XYInputProcessor and xcube.core.gen.iproc.DefaultInputProcessor, if ground control points are not specified, i.e. the input processor is configured with xy_gcp_step=None. (#206)

• Tile sizes for rectification in xcube gen are now derived from output_writer_params if given in configuration and if it contains a chunksizes parameter for 'lat' or 'lon'. This will force the generation of a chunked xcube dataset and will utilize Dask arrays for out-of-core computations. This is very useful for large data cubes whose time slices would otherwise not fit into memory.

• Introduced new function xcube.core.select.select_spatial_subset().

• Renamed function xcube.core.select.select_vars() into xcube.core.select.select_variables_subset().

• Now supporting xarray and numpy functions in expressions used by the xcube.core.evaluate.evaluate_dataset() function and in the configuration of the xcube gen tool. You can now use xr and np contexts in expressions, e.g. xr.where(CHL >= 0.0, CHL). (#257)

• The performance of the xcube gen tool for the case that expressions or expression parts are reused across multiple variables can now be improved. Such as expressions can now be assigned to intermediate variables and loaded into memory, so they are not recomputed again. For example, let the expression quality_flags.cloudy and CHL > 25.0 occur often in the configuration, then this is how recomputation can be avoided:

    processed_variables:
      no_cloud_risk:
        expression: not (quality_flags.cloudy and CHL_raw > 25.0)
        load: True
      CHL:
        expression: CHL_raw
        valid_pixel_expression: no_cloud_risk
      ...        
  

• Added ability to write xcube datasets in Zarr format into object storage bucket using the xcube python api xcube.core.dsio.write_cube(). (#224) The user needs to pass provide user credentials via

  client_kwargs = {'provider_access_key_id': 'user_id', 'provider_secret_access_key': 'user_secret'}
  

  and write to existing bucket by executing

  write_cube(ds1, 'https://s3.amazonaws.com/upload_bucket/cube-1-250-250.zarr', 'zarr',
                       client_kwargs=client_kwargs)
  

• Added new CLI tool xcube tile which is used to generate a tiled RGB image pyramid from any xcube dataset. The format and file organisation of the generated tile sets conforms to the TMS 1.0 Specification (#209).

• The configuration of xcube serve has been enhanced to support augmentation of data cubes by new variables computed on-the-fly (#272). You can now add a section Augmentation into a dataset descriptor, e.g.:

    Datasets:
      - Identifier: abc
        ...
        Augmentation:
          Path: compute_new_vars.py
          Function: compute_variables
          InputParameters:
            ...
      - ...

  where compute_variables is a function that receives the parent xcube dataset and is expected to return a new dataset with new variables.

• The xcube serve tool now provides basic access control via OAuth2 bearer tokens (#263). To configure a service instance with access control, add the following to the xcube serve configuration file:

    Authentication:
      Domain: "<your oauth2 domain>"
      Audience: "<your audience or API identifier>"
  

  Individual datasets can now be protected using the new AccessControl entry by configuring the RequiredScopes entry whose value is a list of required scopes, e.g. "read:datasets":

    Datasets:
      ...
      - Identifier: <some dataset id>
        ...
        AccessControl:
          RequiredScopes:
            - "read:datasets"
  

  If you want a dataset to disappear for authorized requests, set the IsSubstitute flag:

    Datasets:
      ...
      - Identifier: <some dataset id>
        ...
        AccessControl:
          IsSubstitute: true
  

Enhancements

• The xcube serve tool now also allows for per-dataset configuration of chunk caches for datasets read from remote object storage locations. Chunk caching avoids recurring fetching of remote data chunks for same region of interest. It can be configured as default for all remote datasets at top-level of the configuration file:

  DatasetChunkCacheSize: 100M
  

  or in individual dataset definitions:

  Datasets: 
     - Identifier: ...
       ChunkCacheSize: 2G
       ...
  

• Retrieval of time series in Python API function xcube.core.timeseries.get_time_series() has been optimized and is now much faster for point geometries. This enhances time-series performance of xcube serve.

  • The log-output of xcube serve now contains some more details time-series request so performance bottlenecks can be identified more easily from xcube-serve.log, if the server is started together with the flag --traceperf.

• CLI command xcube resample has been enhanced by a new value for the frequency option --frequency all With this value it will be possible to create mean, max , std, ... of the whole dataset, in other words, create an overview of a cube. By Alberto S. Rabaneda.

• The xcube serve tool now also serves dataset attribution information which will be displayed in the xcube-viewer's map. To add attribution information, use the DatasetAttribution in to your xcube serve configuration. It can be used on top-level (for all dataset), or on individual datasets. Its value may be a single text entry or a list of texts: For example:

  DatasetAttribution: 
    - "© by Brockmann Consult GmbH 2020, contains modified Copernicus Data 2019, processed by ESA."
    - "Funded by EU H2020 DCS4COP project."

• The xcube gen tool now always produces consolidated xcube datasets when the output format is zarr. Furthermore when appending to an existing zarr xcube dataset, the output now will be consolidated as well. In addition, xcube gen can now append input time slices to existing optimized (consolidated) zarr xcube datasets.

• The unchunk_coords keyword argument of Python API function xcube.core.optimize.optimize_dataset() can now be a name, or list of names
  of the coordinate variable(s) to be consolidated. If boolean True is used all variables will be consolidated.

• The xcube serve API operations datasets/ and datasets/{ds_id} now also return the metadata attributes of a given dataset and it variables in a property named attrs. For variables we added a new metadata property htmlRepr that is a string returned by a variable's var.data._repr_html_() method, if any.

• Renamed default log file for xcube serve command to xcube-serve.log.

• xcube gen now immediately flushes logging output to standard out

Changes in 0.3.1

Fixes

• Removing false user warning about custom SNAP colormaps when starting xcube serve command.

Changes in 0.3.0

New

• Added new parameter in xcube gen called --no_sort. Using --no_sort, the input file list wont be sorted before creating the xcube dataset. If --no_sort parameter is passed, order the input list will be kept. The parameter --sort is deprecated and the input files will be sorted by default.
• xcube now discovers plugin modules by module naming convention and by Setuptools entry points. See new chapter Plugins in xcube's documentation for details. (#211)
• Added new xcube compute CLI command and xcube.core.compute.compute_cube() API function that can be used to generate an output cube computed from a Python function that is applied to one or more input cubes. Replaces the formerly hidden xcube apply command. (#167)
• Added new function xcube.core.geom.rasterize_features() to rasterize vector-data features into a dataset. (#222)
• Extended CLI command xcube verify and API function xcube.core.verify.verify_cube to check whether spatial coordinate variables and their associated bounds variables are equidistant. (#231)
• Made xarray version 0.14.1 minimum requirement due to deprecation of xarray's Dataset.drop method and replaced it with drop_sel and drop_vars accordingly.

Enhancements

• CLI commands execute much faster now when invoked with the --help and --info options.
• Added serverPID property to response of web API info handler.
• Functions and classes exported by following modules no longer require data cubes to use the lon and lat coordinate variables, i.e. using WGS84 CRS coordinates. Instead, the coordinates' CRS may be a projected coordinate system and coordinate variables may be called x and y (#112):
  • xcube.core.new
  • xcube.core.geom
  • xcube.core.schema
  • xcube.core.verify
• Sometimes the cell bounds coordinate variables of a given coordinate variables are not in a proper, CF compliant order, e.g. for decreasing latitudes lat the respective bounds coordinate lat_bnds is decreasing for lat_bnds[:, 0] and lat_bnds[:, 1], but lat_bnds[i, 0] < lat_bnds[i, 1] for all i. xcube is now more tolerant w.r.t. to such wrong ordering of cell boundaries and will compute the correct spatial extent. (#233)
• For xcube serve, any undefined color bar name will default to "viridis". (#238)

Fixes

• xcube resample now correctly re-chunks its output. By default, chunking of the time dimension is set to one. (#212)

Incompatible changes

The following changes introduce incompatibilities with former xcube 0.2.x versions.

• The function specified by xcube_plugins entry points now receives an single argument of type xcube.api.ExtensionRegistry. Plugins are asked to add their extensions to this registry. As an example, have a look at the default xcube_plugins entry points in ./setup.py.

• xcube.api.compute_dataset() function has been renamed to xcube.api.evaluate_dataset(). This has been done in order avoid confusion with new API function xcube.api.compute_cube().

• xcube's package structure has been drastically changed:

  • all of xcube's __init__.py files are now empty and no longer have side effects such as sub-module aggregations. Therefore, components need to be imported from individual modules.
  • renamed xcube.api into xcube.core
  • moved several modules from xcube.util into xcube.core
  • the new xcube.constants module contains package level constants
  • the new xcube.plugin module now registers all standard extensions
  • moved contents of module xcube.api.readwrite into xcube.core.dsio.
  • removed functions read_cube and read_dataset as open_cube and open_dataset are sufficient
  • all internal module imports are now absolute, rather than relative

Changes in 0.2.1

Enhancements

• Added new CLI tool xcube edit and API function xcube.api.edit_metadata which allows editing the metadata of an existing xcube dataset. (#170)
• xcube serve now recognises xcube datasets with metadata consolidated by the xcube opmimize command. (#141)

Fixes

• xcube gen now parses time stamps correcly from input data. (#207)
• Dataset multi-resolution pyramids (*.levels directories) can be stored in cloud object storage and are now usable with xcube serve (#179)
• xcube optimize now consolidates metadata only after consolidating coordinate variables. (#194)
• Removed broken links from ./README.md (#197)
• Removed obsolete entry points from ./setup.py.

Changes in 0.2.0

New

• Added first version of the xcube documentation generated from ./docs folder.

Enhancements

• Reorganisation of the Documentation and Examples Section (partly addressing #106)

• Loosened python conda environment to satisfy conda-forge requirements

• xcube is now available as a conda package on the conda-forge channel. To install latest xcube package, you can now type: conda install -c conda-forge xcube

• Changed the unittesting code to minimize warnings reported by 3rd-party packages

• Making CLI parameters consistent and removing or changing parameter abbreviations in case they were used twice for different params. (partly addressing #91) For every CLI command which is generating an output a path must be provided by the option -o, --output. If not provided by the user, a default output_path is generated. The following CLI parameter have changed and their abbreviation is not enabled anymore :

  • xcube gen -v is now only xcube gen --vars or xcube gen --variables
  • xcube gen -p is now xcube gen -P
  • xcube gen -i is now xcube gen -I
  • xcube gen -r is now xcube gen -R
  • xcube gen -s is now xcube gen -S
  • xcube chunk -c is now xcube chunk -C
  • xcube level -l is now xcube level -L
  • xcube dump -v is now xcube dump --variable or xcube dump --var
  • xcube dump -e is now xcube dump -E
  • xcube vars2dim -v is now xcube vars2dim --variable or xcube vars2dim --var
  • xcube vars2dim --var_name is now xcube vars2dim --variable or xcube vars2dim --var
  • xcube vars2dim -d is now xcube vars2dim -D
  • xcube grid res -d is now xcube grid res -D
  • xcube grid res -c is now xcube grid res --cov or xcube grid res --coverage
  • xcube grid res -n is now xcube grid res -N or xcube grid res --num_results
  • xcube serve -p is now xcube serve -P
  • xcube serve -a is now xcube serve -A

• Added option inclStDev and inclCount query parameters to ts/{dataset}/{variable}/geometry and derivates. If used with inclStDev=1, Xcube Viewer will show error bars for each time series point.

• xcube.api.new_cube function now accepts callables as values for variables. This allows to compute variable values depending on the (t, y, x) position in the cube. Useful for testing.

• xcube.api now exports the MaskSet class which is useful for decoding flag values encoding following the CF conventions.

• Added new CLI tool xcube optimize and API function xcube.api.optimize_dataset optimizes data cubes for cloud object storage deployment. (#141)

• Added two new spatial dataset operations to Python API xcube.api (#148):

  • mask_dataset_by_geometry(dataset, geometry) clip and mask a dataset by geometry
  • clip_dataset_by_geometry(dataset, geometry) just clip a dataset by geometry

• Changed the dev version tag from 0.2.0.dev3 to 0.2.0.dev

• The behavior of web API /datasets?details=1 has changed. The call no longer includes associated vector data as GeoJSON. Instead new API has beed added to fetch new vector data on demand: /datasets/{dataset}/places and /datasets/{dataset}/places/{place} (#130)

• xcube serve accepts custom SNAP colormaps. The path to a SAP .cpd file can be passed via the server configuration file with the paramter [ColorFile] instead of [ColorBar]. (#84)

• xcube serve can now be configured to serve cubes that are associated with another cube with same data but different chunking (#115). E.g. using chunks such as time=512,lat=1,lon=1 can drastically improve time-series extractions. Have a look at the demo config in xube/webapi/res/demo/config.yml.

• xcube serve does now offer a AWS S3 compatible data access API (#97):

  • List bucket objects: /s3bucket, see AWS docs GET
  • Get bucket object: /s3bucket/{ds_id}/{path}, see AWS docs HEAD and GET

• xcube serve now verifies that a configured cube is valid once it is opened. (#107)

• Added new CLI command xcube verify performing xcube dataset verification. (#19)

• Reworked xcube extract to be finally useful and effective for point data extraction. (#102)

• xcube servercan now filter datasets by point coordinate, e.g. /datasets?point=12.5,52.8. (#50)

• xcube servercan now limit time series to a maximum number of valid (not NaN) values. To activate, pass optional query parameter maxValids to the various /ts functions. The special value -1 will restrict the result to contain only valid values. (#113)

• Reworked xcube gen to be more user-friendly and more consistent with other tools. The changes are

  • Removed --dir and --name options and replaced it by single --output option, whose default value is out.zarr. (#45)
  • The --format option no longer has a default value. If not given, format is guessed from --output option.
  • Renamed following parameters in the configuration file:
    • input_files into input_paths, also because paths may point into object storage locations (buckets);
    • output_file into output_path, to be consistent with input_paths.

• Added new CLI command xcube prune. The tool deletes all block files associated with empty (NaN- only) chunks in given INPUT cube, which must have ZARR format. This can drastically reduce files in sparse cubes and improve cube reading performance. (#92)

• xcube serve has a new prefix option which is a path appended to the server's host. The prefix option replaces the name option which is now deprecated but kept for backward compatibility. (#79)

• Added new CLI command xcube resample that is used to generate temporarily up- or downsampled data cubes from other data cubes.

• xcube serve can now be run with xcube dataset paths and styling information given via the CLI rather than a configuration file. For example xcube serve --styles conc_chl=(0,20,"viridis") /path/to/my/chl-cube.zarr. This allows for quick inspection of newly generated cubes via xcube gen. Also added option --show that starts the Xcube viewer on desktop environments in a browser.

• Added new xcube apply command that can be used to generate an output cube computed from a Python function that is applied to one or more input cubes. The command is still in development and therefore hidden.

• Added new xcube timeit command that can be used to measure the time required for parameterized command invocations. The command is still in development and therefore hidden.

• Added global xcube --scheduler SCHEDULER option for Dask distributed computing (#58)

• Added global xcube --traceback option, removed local xcube gen --traceback option

• Completed version 1 of an xcube developer guide.

• Added xcube serve command (#43)

• xcube serve: Time-series web API now also returns "uncertainty" (#48)

• Added xcube level command to allow for creating spatial pyramid levels (#38)

• xcube gen accepts multiple configuration files that will be merged in order (#21)

• Added xcube gen option --sort when input data list should be sorted (#33)

• Added xcube vars2dim command to make variables a cube dimension (#31)

• Added xcube serve option --traceperf that allows switching on performance diagnostics.

• Included possibility to read the input file paths from a text file. (#47)

• Restructured and clarified code base (#27)

• Moved to Python 3.7 (#25)

• Excluding all input processors except for the default one. They are now plugins and have own repositories within the xcube's organisation. (#49)

Fixes

• xcube gen CLI now updates metadata correctly. (#181)
• It was no longer possible to use the xcube gen CLI with --proc option. (#120)
• totalCount attribute of time series returned by Web API ts/{dataset}/{variable}/{geom-type} now contains the correct number of possible observations. Was always 1 before.
• Renamed Web API function ts/{dataset}/{variable}/places into ts/{dataset}/{variable}/features.
• xcube gen is now taking care that when new time slices are added to an existing cube, this is done by maintaining the chronological order. New time slices are either appended, inserted, or replaced. (#64) (#139)
• Fixed xcube serve issue with WMTS KVP method GetTile with query parameter time whose value can now also have the two forms <start-date>/<end-date> and just <date>. (#132)
• Fixed xcube extract regression that stopped working after Pandas update (#95)
• Fixed problem where CTRL+C didn't function anymore with xcube serve. (#87)
• Fixed error indexes along dimension 'y' are not equal occurred when using xcube gen with processed variables that used flag values (#86)
• Fixed xcube serve WMTS KVP API to allow for case-insensitive query parameters. (#77)
• Fixed error in plugins when importing xcube.api.gen (#62)
• Fixed import of plugins only when executing xcube.cli (#66)

Changes in 0.1.0

• Respecting chunk sizes when computing tile sizes #44
• The RESTful tile operations now have a query parameter debug=1 which toggles tile computation performance diagnostics.
• Can now associate place groups with datasets.
• Major revision of API. URLs are now more consistent.
• Request for obtaining a legend for a layer of given by a variable of a data set was added.
• Added a Dockerfile to build an xcube docker image and to run the demo
• The RESTful time-series API now returns ISO-formatted UTC dates #26