def _fill_domain_axes(self):
for axis in self.axes:
if axis.is_easting():
east_axis = RegularAxis(axis.label, axis.uom,
self.gdal_dataset.get_extents_x()[0],
self.gdal_dataset.get_extents_x()[1],
self.gdal_dataset.get_origin_x(), axis)
self.subsets.append(
AxisSubset(
CoverageAxis(east_axis, None, True),
Interval(self.gdal_dataset.get_extents_x()[0],
self.gdal_dataset.get_extents_x()[1])))
elif axis.is_northing():
north_axis = RegularAxis(axis.label, axis.uom,
self.gdal_dataset.get_extents_y()[0],
self.gdal_dataset.get_extents_y()[1],
self.gdal_dataset.get_origin_y(),
axis)
self.subsets.append(
AxisSubset(
CoverageAxis(north_axis, None, True),
Interval(self.gdal_dataset.get_extents_y()[0],
self.gdal_dataset.get_extents_y()[1])))
else:
unknown_axis = Axis(axis.label, axis.uom, 0, 0, 0, axis)
self.subsets.append(
AxisSubset(CoverageAxis(unknown_axis, None, False),
Interval(0)))
def _axis_subset(self, crs_axis, nc_file):
"""
Returns an axis subset using the given crs axis in the context of the nc file
:param CRSAxis crs_axis: the crs definition of the axis
:param File nc_file: the netcdf file
:rtype AxisSubset
"""
user_axis = self._user_axis(self._get_user_axis_by_crs_axis_name(crs_axis.label), NetcdfEvaluatorSlice(nc_file))
# Normally, without pixelIsPoint:true, in the ingredient needs to +/- 0.5 * resolution for each regular axis
# e.g: resolution for axis E is 10000, then
# "min": "${netcdf:variable:E:min} - 10000 / 2",
# "max": "${netcdf:variable:E:max} + 10000 / 2",
# with pixelIsPoint: true, no need to add these values as the service will do it automatically
if self.pixel_is_point:
PointPixelAdjuster.adjust_axis_bounds_to_continuous_space(user_axis, crs_axis)
else:
# No adjustment for all regular axes but still need to translate time in datetime to decimal to calculate
if user_axis.type == UserAxisType.DATE:
user_axis.interval.low = decimal.Decimal(str(arrow.get(user_axis.interval.low).float_timestamp))
if user_axis.interval.high:
user_axis.interval.high = decimal.Decimal(str(arrow.get(user_axis.interval.high).float_timestamp))
# if low < high, adjust it
if user_axis.interval.high is not None and user_axis.interval.low > user_axis.interval.high:
user_axis.interval.low, user_axis.interval.high = user_axis.interval.high, user_axis.interval.low
high = user_axis.interval.high if user_axis.interval.high else user_axis.interval.low
origin = PointPixelAdjuster.get_origin(user_axis, crs_axis)
if isinstance(user_axis, RegularUserAxis):
geo_axis = RegularAxis(crs_axis.label, crs_axis.uom, user_axis.interval.low, high, origin, crs_axis)
else:
if user_axis.type == UserAxisType.DATE:
if crs_axis.is_uom_day():
coefficients = self._translate_day_date_direct_position_to_coefficients(user_axis.interval.low,
user_axis.directPositions)
else:
coefficients = self._translate_seconds_date_direct_position_to_coefficients(user_axis.interval.low,
user_axis.directPositions)
else:
coefficients = self._translate_number_direct_position_to_coefficients(user_axis.interval.low,
user_axis.directPositions)
geo_axis = IrregularAxis(crs_axis.label, crs_axis.uom, user_axis.interval.low, high, origin, coefficients, crs_axis)
grid_low = 0
grid_high = PointPixelAdjuster.get_grid_points(user_axis, crs_axis)
# NOTE: Grid Coverage uses the direct intervals as in Rasdaman
if self.grid_coverage is False and grid_high > grid_low:
grid_high -= 1
grid_axis = GridAxis(user_axis.order, crs_axis.label, user_axis.resolution, grid_low, grid_high)
if user_axis.type == UserAxisType.DATE:
self._translate_decimal_to_datetime(user_axis, geo_axis)
return AxisSubset(CoverageAxis(geo_axis, grid_axis, user_axis.dataBound),
Interval(user_axis.interval.low, user_axis.interval.high))
def _get_slices(self, coverage_axes, intervals):
"""
Returns the slices
:param list[CoverageAxis] coverage_axes: the coverage axes
:param list[list[Interval]] intervals: all the possible intervals defining the coverage space
:return:
"""
slices = []
for interval_list in intervals:
subsets = []
for index in range(0, len(coverage_axes)):
subsets.append(
AxisSubset(coverage_axes[index], interval_list[index]))
if len(coverage_axes) == 1:
# For 1D we have to parse the gml and create a tuple data provider
data_provider = TupleListDataProvider(
self._get_coverage_data_as_array(
self._get_coverage_url(subsets)))
else:
data_provider = UrlDataProvider(
self._get_coverage_url(subsets))
slices.append(Slice(subsets, data_provider))
return slices
def _axis_subset(self, crs_axis, evaluator_slice, resolution=None):
"""
Returns an axis subset using the given crs axis in the context of the gdal file
:param CRSAxis crs_axis: the crs definition of the axis
:param GDALEvaluatorSlice evaluator_slice: the evaluator for GDAL file
:param resolution: Known axis resolution, no need to evaluate sentence expression from ingredient file (e.g: Sentinel2 recipe)
:rtype AxisSubset
"""
user_axis = self._user_axis(
self._get_user_axis_by_crs_axis_name(crs_axis.label),
evaluator_slice)
if resolution is not None:
user_axis.resolution = resolution
high = user_axis.interval.high if user_axis.interval.high is not None else user_axis.interval.low
if user_axis.type == UserAxisType.DATE:
# it must translate datetime string to float by arrow for calculating later
user_axis.interval.low = arrow.get(
user_axis.interval.low).float_timestamp
if user_axis.interval.high is not None:
user_axis.interval.high = arrow.get(
user_axis.interval.high).float_timestamp
if isinstance(user_axis, RegularUserAxis):
geo_axis = RegularAxis(crs_axis.label, crs_axis.uom,
user_axis.interval.low, high,
user_axis.interval.low, crs_axis)
else:
# Irregular axis (coefficients must be number, not datetime string)
if user_axis.type == UserAxisType.DATE:
if crs_axis.is_time_day_axis():
coefficients = self._translate_day_date_direct_position_to_coefficients(
user_axis.interval.low, user_axis.directPositions)
else:
coefficients = self._translate_seconds_date_direct_position_to_coefficients(
user_axis.interval.low, user_axis.directPositions)
else:
coefficients = self._translate_number_direct_position_to_coefficients(
user_axis.interval.low, user_axis.directPositions)
self._update_for_slice_group_size(self.coverage_id, user_axis,
crs_axis, coefficients)
geo_axis = IrregularAxis(crs_axis.label, crs_axis.uom,
user_axis.interval.low, high,
user_axis.interval.low, coefficients,
crs_axis)
if not crs_axis.is_x_axis() and not crs_axis.is_y_axis():
# GDAL model is 2D so on any axis except x/y we expect to have only one value
grid_low = 0
grid_high = None
if user_axis.interval.high is not None:
grid_high = 0
else:
grid_low = 0
number_of_grid_points = decimal.Decimal(str(user_axis.interval.high)) \
- decimal.Decimal(str(user_axis.interval.low))
# number_of_grid_points = (geo_max - geo_min) / resolution
grid_high = grid_low + number_of_grid_points / decimal.Decimal(
user_axis.resolution)
grid_high = HighPixelAjuster.adjust_high(grid_high)
# Negative axis, e.g: Latitude (min <--- max)
if user_axis.resolution < 0:
grid_high = int(abs(math.floor(grid_high)))
else:
# Positive axis, e.g: Longitude (min --> max)
grid_high = int(abs(math.ceil(grid_high)))
# NOTE: Grid Coverage uses the direct intervals as in Rasdaman
if self.grid_coverage is False and grid_high is not None:
if grid_high > grid_low:
grid_high -= 1
grid_axis = GridAxis(user_axis.order, crs_axis.label,
user_axis.resolution, grid_low, grid_high)
geo_axis.origin = PointPixelAdjuster.get_origin(user_axis, crs_axis)
if user_axis.type == UserAxisType.DATE:
self._translate_decimal_to_datetime(user_axis, geo_axis)
# NOTE: current, gdal recipe supports only has 2 axes which are "bounded" (i.e: they exist as 2D axes in file)
# and 1 or more another axes gotten (i.e: from fileName) which are not "bounded" to create 3D+ coverage.
data_bound = crs_axis.is_y_axis() or crs_axis.is_x_axis()
return AxisSubset(
CoverageAxis(geo_axis, grid_axis, data_bound),
Interval(user_axis.interval.low, user_axis.interval.high))
def _axis_subset(self, grib_file, evaluated_messages, crs_axis):
"""
Returns an axis subset using the given crs axis in the context of the grib file
:param File grib_file: the current grib file (slice) is evaluated
:param List[GirbMessages] evaluated_messages: all Grib messages was evaluated
:param CRSAxis crs_axis: the crs definition of the axis
:rtype AxisSubset
"""
# first grib message from grib file, used to extract grib variables only
first_grib_message = self.dataset.message(1)
# As all the messages contain same axes (but different intervals), so first message is ok to get user_axis
first_user_axis = self._get_user_axis_in_evaluated_message(
evaluated_messages[0], crs_axis.label)
# NOTE: we don't want to change this user_axis belongs to messages, so clone it
user_axis = copy.deepcopy(first_user_axis)
# Then, we calculate the geo, grid bounds, origin, resolution of this axis for the slice
self._set_low_high(evaluated_messages, user_axis)
high = user_axis.interval.high if user_axis.interval.high is not None else user_axis.interval.low
origin = PointPixelAdjuster.get_origin(user_axis, crs_axis)
if isinstance(user_axis, RegularUserAxis):
geo_axis = RegularAxis(crs_axis.label, crs_axis.uom,
user_axis.interval.low, high, origin,
crs_axis)
else:
# after all messages was evaluated, we could get the direct_positions of the axis as in netcdf
# then, it can evaluate the grib sentence normally, e.g: ${grib:axis:level} + 5
evaluating_sentence = user_axis.directPositions
direct_positions = self._get_axis_values(evaluated_messages,
user_axis)
# convert all of values in the list to string then it can be evaluated
direct_positions = list_util.to_list_string(direct_positions)
evaluator_slice = GribMessageEvaluatorSlice(
first_grib_message, grib_file, direct_positions)
user_axis.directPositions = self.sentence_evaluator.evaluate(
evaluating_sentence, evaluator_slice, user_axis.statements)
# axis is datetime
if user_axis.type == UserAxisType.DATE:
if crs_axis.is_time_day_axis():
coefficients = self._translate_day_date_direct_position_to_coefficients(
user_axis.interval.low, user_axis.directPositions)
else:
coefficients = self._translate_seconds_date_direct_position_to_coefficients(
user_axis.interval.low, user_axis.directPositions)
else:
# number axis like Index1D
coefficients = self._translate_number_direct_position_to_coefficients(
user_axis.interval.low, user_axis.directPositions)
self._update_for_slice_group_size(self.coverage_id, user_axis,
crs_axis, coefficients)
geo_axis = IrregularAxis(crs_axis.label, crs_axis.uom,
user_axis.interval.low, high, origin,
coefficients, crs_axis)
grid_low = 0
grid_high = PointPixelAdjuster.get_grid_points(user_axis, crs_axis)
# NOTE: Grid Coverage uses the direct intervals as in Rasdaman
if self.grid_coverage is False and grid_high > grid_low:
grid_high -= 1
grid_axis = GridAxis(user_axis.order, crs_axis.label,
user_axis.resolution, grid_low, grid_high)
if user_axis.type == UserAxisType.DATE:
self._translate_decimal_to_datetime(user_axis, geo_axis)
return AxisSubset(
CoverageAxis(geo_axis, grid_axis, user_axis.dataBound),
Interval(user_axis.interval.low, user_axis.interval.high))
def _axis_subset(self, crs_axis, gdal_file):
"""
Returns an axis subset using the given crs axis in the context of the gdal file
:param CRSAxis crs_axis: the crs definition of the axis
:param File gdal_file: the gdal file
:rtype AxisSubset
"""
user_axis = self._user_axis(
self._get_user_axis_by_crs_axis_name(crs_axis.label),
GDALEvaluatorSlice(GDALGmlUtil(gdal_file.get_filepath())))
high = user_axis.interval.high if user_axis.interval.high else user_axis.interval.low
if isinstance(user_axis, RegularUserAxis):
geo_axis = RegularAxis(crs_axis.label, crs_axis.uom,
user_axis.interval.low, high,
user_axis.interval.low, crs_axis)
else:
# if irregular axis value is fetched from fileName so the coefficient is [0] as slicing
if user_axis.directPositions == AbstractToCoverageConverter.DIRECT_POSITIONS_SLICING:
user_axis.directPositions = AbstractToCoverageConverter.COEFFICIENT_SLICING
geo_axis = IrregularAxis(crs_axis.label, crs_axis.uom,
user_axis.interval.low, high,
user_axis.interval.low,
user_axis.directPositions, crs_axis)
if not crs_axis.is_easting() and not crs_axis.is_northing():
# GDAL model is 2D so on any axis except x/y we expect to have only one value
grid_low = 0
grid_high = 0
else:
grid_low = 0
number_of_grid_points = decimal.Decimal(str(user_axis.interval.high)) \
- decimal.Decimal(str(user_axis.interval.low))
# number_of_grid_points = (geo_max - geo_min) / resolution
grid_high = grid_low + number_of_grid_points / decimal.Decimal(
user_axis.resolution)
grid_high = HighPixelAjuster.adjust_high(grid_high)
# Negative axis, e.g: Latitude (min <--- max)
if user_axis.resolution < 0:
grid_high = int(abs(math.floor(grid_high)))
else:
# Positive axis, e.g: Longitude (min --> max)
grid_high = int(abs(math.ceil(grid_high)))
# NOTE: Grid Coverage uses the direct intervals as in Rasdaman
if self.grid_coverage is False:
if grid_high > grid_low:
grid_high -= 1
grid_axis = GridAxis(user_axis.order, crs_axis.label,
user_axis.resolution, grid_low, grid_high)
geo_axis.origin = PointPixelAdjuster.get_origin(user_axis, crs_axis)
if user_axis.type == UserAxisType.DATE:
self._translate_decimal_to_datetime(user_axis, geo_axis)
# NOTE: current, gdal recipe supports only has 2 axes which are "bounded" (i.e: they exist as 2D axes in file)
# and 1 or more another axes gotten (i.e: from fileName) which are not "bounded" to create 3D+ coverage.
data_bound = crs_axis.is_northing() or crs_axis.is_easting()
return AxisSubset(
CoverageAxis(geo_axis, grid_axis, data_bound),
Interval(user_axis.interval.low, user_axis.interval.high))