CHANGES.md

Changes in 0.3.5 (in development)

• Fixed a bug in rectify_dataset where the conversion from block-space indices to global-space indices was incorrect when output_indices_names was provided.

Changes in 0.3.4

• Added utils.transform_resolution to convert spatial resolution between coordinate reference systems (CRS).

Changes in 0.3.3

• Added function utils.resolution_degrees_to_meters to convert spatial resolution from degrees to meters at a given geographic latitude.

Changes in 0.3.2

• Change inconsistent license classifier in pyproject.toml
• Removed the dependency on zarr. As part of this change, xcube_resampling.grid_mapping.cfconv.add_spatial_ref has been removed, as it only loosely fit the scope of the xcube-resampling repository.
• Added xcube_resampling.utils.get_utm_crs to return the UTM CRS for a given longitude and latitude pair.
• In xcube_resampling.reproject_dataset, the dataset is now clipped when the target grid-mapping covers less than 80% of the source grid-mapping, improving performance for reprojected cutouts.
• In xcube_resampling.affine_transform_dataset, resampling is now performed per spatial slice, ensuring that interpolation is applied only along the spatial axes and no longer across non-spatial dimensions (e.g., time).
• When prevent_nan_propagation=True, computation is now fully lazy. Previously, the implementation checked for NaN values in the source array, which triggered loading of the entire dataset into memory.
• Enhanced GridMapping API documentation.

Changes in 0.3.1

• Fixed a bug in xcube_resampling.rectify_dataset that occurred when processing
  datasets with decreasing x-coordinates.
• Boolean (bool) data arrays now default to nearest-neighbor interpolation.
  For spatial aggregation, values are centered, and 0 is used as the default fill value—matching the behavior of integer arrays to ensure consistent and valid results.

Changes in 0.3.0

• Fixed an issue when retrieving coordinate names from regular grid mappings.
• Improved documentation for the frequency and tolerance arguments of xcube_resampling.resample_in_time.
• Fixed incorrect time coverage metadata in dataset attributes after calling xcube_resampling.resample_in_time.
• xcube_resampling.rectify_dataset is now fully lazy and scalable.
• Removed methods test_ij_bbox_from_xy_bbox and test_ij_bboxes_from_xy_bboxes from xcube_resampling.gridmapping.GridMapping. These methods were only used by xcube_resampling.rectify_dataset; their updated implementations now live in xcube_resampling.rectify._compute_source_tile_indexing.
• Added support for rectifying datasets with decreasing x-coordinate in xcube_resampling.rectify_dataset.

Changes in 0.2.4

• Improved performance when creating GridMapping from 2D coordinates. Parameter estimation now uses only corner points instead of the full set of coordinates, significantly reducing runtime for large datasets.
• Bug fix in affine transformations. Fixed a half-pixel shift that occurred in xcube_resampling.affine_transform_dataset and xcube_resampling.resample_in_space when the source and target grid mappings shared the same CRS and were regular.

Changes in 0.2.3

• In xcube_resampling.resample_in_time, the method suffix in the output data variable names is now added only when the agg_methods or interp_methods parameters are provided as mappings, or when multiple resampling methods are applied to the data variable.

Changes in 0.2.2

• Added support for spatial rectification of datasets in xcube_resampling.rectify_dataset even when spatial coordinates are not stored in the dataset’s coordinate variables.

Changes in 0.2.1

• Added the output_indices_names parameter to xcube_resampling.rectify_dataset to allow storing the source pixel indices.

Changes in 0.2.0

• Enhanced the function bbox_overlap so that it can handle bounding boxes crossing the antimeridian
• Added new class method GridMapping.regular_from_bbox, which allows creating a regular grid mapping directly from a bounding box, spatial resolution, and CRS.
• Added new function utils.resolution_meters_to_degrees which converts spatial resolution from meters to degrees in latitude and longitude at a given geographic latitude.
  • 1 degree of latitude ≈ 111,320 meters (constant approximation).
  • 1 degree of longitude ≈ 111,320 * cos(latitude) meters.
• Bug fix: fixed grid mapping creation for irregular grids with decreasing longitude along axis 1.
• Added xcube_resampling.resample_in_time which allows to resample a dataset along the time axis. It supports up- and down-sampling.
• Renamed the parameter recover_nans to prevent_nan_propagations, which can be a boolean or a mapping (from variable name or dtype to boolean) that, when True, prevents NaN propagation during upsampling or interpolation.

Changes in 0.1.1

• Improved xcube_resampling.utils.clip_dataset_by_bbox to support datasets with 2D coordinates. This function is also used internally by xcube_resampling.rectify.rectify_dataset.
• Added reproject_bbox, a utility to reproject a bounding box from one CRS to another.
• Added bbox_overlap, a utility that computes the fraction of the source bounding box overlapped by the target bounding box.

Changes in 0.1.0

• Added algorithm for affine transformation. (#4)
• Added algorithm for rectification of non-regular grids. (#4)
• Added algorithm for reprojection to a different coordinate reference system (CRS). (#4)
• Introduced main function resample_in_space, which dynamically selects the appropriate resampling algorithm based on the input dataset. (#4)
• Added initial unit tests to verify core functionality. (#4)
• Introduced a new unified keyword argument interp_method that supports values 0, 1, "nearest", "triangular", and "bilinear". This argument applies consistently across all three resampling algorithms, simplifying usage and improving API consistency. (#8)
• Documentation added, which is available at https://xcube-dev.github.io/xcube-resampling/. (#10)