def rectify_dataset(source_ds: xr.Dataset,
*,
var_names: Union[str, Sequence[str]] = None,
source_gm: GridMapping = None,
xy_var_names: Tuple[str, str] = None,
target_gm: GridMapping = None,
tile_size: Union[int, Tuple[int, int]] = None,
is_j_axis_up: bool = None,
output_ij_names: Tuple[str, str] = None,
compute_subset: bool = True,
uv_delta: float = 1e-3) -> Optional[xr.Dataset]:
"""
Reproject dataset *source_ds* using its per-pixel
x,y coordinates or the given *source_gm*.
The function expects *source_ds* or the given
*source_gm* to have either one- or two-dimensional
coordinate variables that provide spatial x,y coordinates
for every data variable with the same spatial dimensions.
For example, a dataset may comprise variables with
spatial dimensions ``var(..., y_dim, x_dim)``,
then one the function expects coordinates to be provided
in two forms:
1. One-dimensional ``x_var(x_dim)``
and ``y_var(y_dim)`` (coordinate) variables.
2. Two-dimensional ``x_var(y_dim, x_dim)``
and ``y_var(y_dim, x_dim)`` (coordinate) variables.
If *target_gm* is given and it defines a tile size
or *tile_size* is given, and the number of tiles is
greater than one in the output's x- or y-direction, then the
returned dataset will be composed of lazy, chunked dask
arrays. Otherwise the returned dataset will be composed
of ordinary numpy arrays.
:param source_ds: Source dataset grid mapping.
:param var_names: Optional variable name or sequence of
variable names.
:param source_gm: Target dataset grid mapping.
:param xy_var_names: Optional tuple of the x- and y-coordinate
variables in *source_ds*. Ignored if *source_gm* is given.
:param target_gm: Optional output geometry. If not given,
output geometry will be computed to spatially fit *dataset*
and to retain its spatial resolution.
:param tile_size: Optional tile size for the output.
:param is_j_axis_up: Whether y coordinates are increasing with
positive image j axis.
:param output_ij_names: If given, a tuple of variable names in
which to store the computed source pixel coordinates in
the returned output.
:param compute_subset: Whether to compute a spatial subset
from *dataset* using *output_geom*. If set, the function
may return ``None`` in case there is no overlap.
:param uv_delta: A normalized value that is used to determine
whether x,y coordinates in the output are contained
in the triangles defined by the input x,y coordinates.
The higher this value, the more inaccurate the rectification
will be.
:return: a reprojected dataset, or None if the requested output
does not intersect with *dataset*.
"""
if source_gm is None:
source_gm = GridMapping.from_dataset(source_ds,
xy_var_names=xy_var_names)
src_attrs = dict(source_ds.attrs)
if target_gm is None:
target_gm = source_gm.to_regular(tile_size=tile_size)
elif compute_subset:
source_ds_subset = select_spatial_subset(
source_ds,
xy_bbox=target_gm.xy_bbox,
ij_border=1,
xy_border=0.5 * (target_gm.x_res + target_gm.y_res),
grid_mapping=source_gm)
if source_ds_subset is None:
return None
if source_ds_subset is not source_ds:
# TODO: GridMapping.from_dataset() may be expensive.
# Find a more effective way.
source_gm = GridMapping.from_dataset(source_ds_subset)
source_ds = source_ds_subset
# if src_geo_coding.xy_var_names != output_geom.xy_var_names:
# output_geom = output_geom.derive(
# xy_var_names=src_geo_coding.xy_var_names
# )
# if src_geo_coding.xy_dim_names != output_geom.xy_dim_names:
# output_geom = output_geom.derive(
# xy_dim_names=src_geo_coding.xy_dim_names
# )
if tile_size is not None or is_j_axis_up is not None:
target_gm = target_gm.derive(tile_size=tile_size,
is_j_axis_up=is_j_axis_up)
src_vars = _select_variables(source_ds, source_gm, var_names)
if target_gm.is_tiled:
compute_dst_src_ij_images = _compute_ij_images_xarray_dask
compute_dst_var_image = _compute_var_image_xarray_dask
else:
compute_dst_src_ij_images = _compute_ij_images_xarray_numpy
compute_dst_var_image = _compute_var_image_xarray_numpy
dst_src_ij_array = compute_dst_src_ij_images(source_gm, target_gm,
uv_delta)
dst_x_dim, dst_y_dim = target_gm.xy_dim_names
dst_dims = dst_y_dim, dst_x_dim
dst_ds_coords = target_gm.to_coords()
dst_vars = dict()
for src_var_name, src_var in src_vars.items():
dst_var_dims = src_var.dims[0:-2] + dst_dims
dst_var_coords = {
d: src_var.coords[d]
for d in dst_var_dims if d in src_var.coords
}
dst_var_coords.update(
{d: dst_ds_coords[d]
for d in dst_var_dims if d in dst_ds_coords})
dst_var_array = compute_dst_var_image(src_var,
dst_src_ij_array,
fill_value=np.nan)
dst_var = xr.DataArray(dst_var_array,
dims=dst_var_dims,
coords=dst_var_coords,
attrs=src_var.attrs)
dst_vars[src_var_name] = dst_var
if output_ij_names:
output_i_name, output_j_name = output_ij_names
dst_ij_coords = {
d: dst_ds_coords[d]
for d in dst_dims if d in dst_ds_coords
}
dst_vars[output_i_name] = xr.DataArray(dst_src_ij_array[0],
dims=dst_dims,
coords=dst_ij_coords)
dst_vars[output_j_name] = xr.DataArray(dst_src_ij_array[1],
dims=dst_dims,
coords=dst_ij_coords)
return xr.Dataset(dst_vars, coords=dst_ds_coords, attrs=src_attrs)
def rechunk_cube(cube: xr.Dataset,
gm: GridMapping,
chunks: Optional[Dict[str, int]] = None,
tile_size: Optional[Tuple[int, int]] = None) \
-> Tuple[xr.Dataset, GridMapping]:
"""
Re-chunk data variables of *cube* so they all share the same chunk
sizes for their dimensions.
This functions rechunks *cube* for maximum compatibility with
the Zarr format. Therefore it removes the "chunks" encoding
from all variables.
:param cube: A data cube
:param gm: The cube's grid mapping
:param chunks: Optional mapping of dimension names to chunk sizes
:param tile_size: Optional tile sizes, i.e. chunk size of
spatial dimensions, given as (width, height)
:return: A potentially rechunked *cube* and adjusted grid mapping.
"""
# get initial, common cube chunk sizes from given cube
cube_chunks = get_dataset_chunks(cube)
# Given chunks will overwrite initial values
if chunks:
for dim_name, size in chunks.items():
cube_chunks[dim_name] = size
# Given tile size will overwrite spatial dims
x_dim_name, y_dim_name = gm.xy_dim_names
if tile_size is not None:
cube_chunks[x_dim_name] = tile_size[0]
cube_chunks[y_dim_name] = tile_size[1]
# Given grid mapping's tile size will overwrite
# spatial dims only if missing still
if gm.tile_size is not None:
if x_dim_name not in cube_chunks:
cube_chunks[x_dim_name] = gm.tile_size[0]
if y_dim_name not in cube_chunks:
cube_chunks[y_dim_name] = gm.tile_size[1]
# If there is no chunking required, return identities
if not cube_chunks:
return cube, gm
chunked_cube = xr.Dataset(attrs=cube.attrs)
# Coordinate variables are always
# chunked automatically
chunked_cube = chunked_cube.assign_coords(
coords={
var_name: var.chunk({dim_name: 'auto'
for dim_name in var.dims})
for var_name, var in cube.coords.items()
})
# Data variables are chunked according to cube_chunks,
# or if not specified, by the dimension size.
chunked_cube = chunked_cube.assign(
variables={
var_name: var.chunk({
dim_name: cube_chunks.get(dim_name, cube.dims[dim_name])
for dim_name in var.dims
})
for var_name, var in cube.data_vars.items()
})
# Update chunks encoding for Zarr
for var_name, var in chunked_cube.variables.items():
if 'chunks' in var.encoding:
del var.encoding['chunks']
# if var.chunks is not None:
# # sizes[0] is the first of
# # e.g. sizes = (512, 512, 71)
# var.encoding.update(chunks=[
# sizes[0] for sizes in var.chunks
# ])
# elif 'chunks' in var.encoding:
# del var.encoding['chunks']
# print(f"--> {var_name}: encoding={var.encoding.get('chunks')!r}, chunks={var.chunks!r}")
# Test whether tile size has changed after re-chunking.
# If so, we will change the grid mapping too.
tile_width = cube_chunks.get(x_dim_name)
tile_height = cube_chunks.get(y_dim_name)
assert tile_width is not None
assert tile_height is not None
tile_size = (tile_width, tile_height)
if tile_size != gm.tile_size:
# Note, changing grid mapping tile size may
# rechunk (2D) coordinates in chunked_cube too
gm = gm.derive(tile_size=tile_size)
return chunked_cube, gm
def process(self,
dataset: xr.Dataset,
geo_coding: GridMapping,
output_geom: GridMapping,
output_resampling: str,
include_non_spatial_vars=False) -> xr.Dataset:
"""
Perform reprojection using tie-points / ground control points.
"""
reprojection_info = self.get_reprojection_info(dataset)
warn_prefix = 'unsupported argument in np-GCP rectification mode'
if reprojection_info.xy_crs is not None:
warnings.warn(
f'{warn_prefix}: ignoring '
f'reprojection_info.xy_crs = {reprojection_info.xy_crs!r}')
if reprojection_info.xy_tp_names is not None:
warnings.warn(
f'{warn_prefix}: ignoring '
f'reprojection_info.xy_tp_names = {reprojection_info.xy_tp_names!r}'
)
if reprojection_info.xy_gcp_step is not None:
warnings.warn(
f'{warn_prefix}: ignoring '
f'reprojection_info.xy_gcp_step = {reprojection_info.xy_gcp_step!r}'
)
if reprojection_info.xy_tp_gcp_step is not None:
warnings.warn(
f'{warn_prefix}: ignoring '
f'reprojection_info.xy_tp_gcp_step = {reprojection_info.xy_tp_gcp_step!r}'
)
if output_resampling != 'Nearest':
warnings.warn(f'{warn_prefix}: ignoring '
f'dst_resampling = {output_resampling!r}')
if include_non_spatial_vars:
warnings.warn(
f'{warn_prefix}: ignoring '
f'include_non_spatial_vars = {include_non_spatial_vars!r}')
geo_coding = geo_coding.derive(
xy_var_names=(reprojection_info.xy_names[0],
reprojection_info.xy_names[1]))
dataset = rectify_dataset(dataset,
compute_subset=False,
source_gm=geo_coding,
target_gm=output_geom)
if output_geom.is_tiled:
# The following condition may become true,
# if we have used rectified_dataset(input, ..., is_y_reverse=True)
# In this case y-chunksizes will also be reversed. So that the first chunk is smaller than any other.
# Zarr will reject such datasets, when written.
if dataset.chunks.get('lat')[0] < dataset.chunks.get('lat')[-1]:
dataset = dataset.chunk({
'lat': output_geom.tile_height,
'lon': output_geom.tile_width
})
if dataset is not None \
and geo_coding.crs.is_geographic \
and geo_coding.xy_var_names != ('lon', 'lat'):
dataset = dataset.rename({
geo_coding.xy_var_names[0]: 'lon',
geo_coding.xy_var_names[1]: 'lat'
})
return dataset