NumpyRaster¶
-
class
buzzard.
ASource
(<implementation detail>)[source] Base abstract class defining the common behavior of all sources opened in the Dataset.
Features Defined
Has a stored spatial reference
Has a virtual spatial reference that is influenced by the Dataset’s opening mode
Can be closed
-
property
wkt_stored
The spatial reference that can be found in the metadata of a source, in wkt format.
string or None
-
property
proj4_stored
The spatial reference that can be found in the metadata of a source, in proj4 format.
string or None
-
property
wkt_virtual
The spatial reference considered to be written in the metadata of a source, in wkt format.
string or None
-
property
proj4_virtual
The spatial reference considered to be written in the metadata of a source, in proj4 format.
string or None
-
get_keys
()[source] Get the list of keys under which this source is registered to in the Dataset
-
property
close
Close a source with a call or a context management. The close attribute returns an object that can be both called and used in a with statement
Examples
>>> ds.dem.close() >>> with ds.dem.close: # code... >>> with ds.acreate_raster('result.tif', fp, float, 1).close as result: # code... >>> with ds.acreate_vector('results.shp', 'linestring').close as roofs: # code...
-
__del__
()[source]
-
class
buzzard.
ASourceRaster
(<implementation detail>)[source] Base abstract class defining the common behavior of all rasters.
Features Defined
Has a stored Footprint that defines the location of the raster
Has a Footprint that is influenced by the Dataset’s opening mode
Has a length that defines how many channels are available
Has a channels_schema that defines per channel attributes (e.g. nodata)
Has a dtype (like np.float32)
Has a get_data method that allows to read pixels in their current state to numpy arrays
-
property
fp_stored
-
property
fp
-
property
channels_schema
-
property
dtype
-
property
nodata
Accessor for first channel’s nodata value
-
get_nodata
(channel=0)[source] Accessor for nodata value
-
__len__
()[source] Return the number of channels
-
get_data
(fp=None, channels=None, dst_nodata=None, interpolation='cv_area', **kwargs)[source] Read a rectangle of data on several channels from the source raster.
If fp is not fully within the source raster, the external pixels are set to nodata. If nodata is missing, 0 is used. If fp is not on the same grid as the source raster, remapping is performed using interpolation algorithm. (It fails if the allow_interpolation parameter is set to False in Dataset (default)). When remapping, the nodata values are not interpolated, they are correctly spread to the output.
If dst_nodata is provided, nodata pixels are set to dst_nodata.
Warning
The alpha channels are currently resampled like any other channels, this behavior may change in the future. To normalize an rgba array after a resampling operation, use this piece of code:
>>> arr = np.where(arr[..., -1] == 255, arr, 0)
Warning
Bands in GDAL are indexed from 1. Channels in buzzard are indexed from 0.
Parameters
- fp: Footprint of shape (Y, X) or None
If None: return the full source raster
If Footprint: return this window from the raster
- channels: None or int or slice or sequence of int (see Channels Parameter below)
The channels to be read
- dst_nodata: nbr or None
nodata value in output array If None and raster.nodata is not None: raster.nodata is used If None and raster.nodata is None: 0 is used
- interpolation: one of {‘cv_area’, ‘cv_nearest’, ‘cv_linear’, ‘cv_cubic’, ‘cv_lanczos4’} or None
OpenCV method used if intepolation is necessary
Returns
- array: numpy.ndarray of shape (Y, X) or (Y, X, C)
If the channels parameter is -1, the returned array is of shape (Y, X) when C=1, (Y, X, C) otherwise.
If the channels parameter is an integer >=0, the returned array is of shape (Y, X).
If the channels parameter is a sequence or a slice, the returned array is always of shape (Y, X, C), no matter the size of C.
(see Channels Parameter below)
Channels Parameter
type
value
meaning
output shape
NoneType
None (default)
All channels
(Y, X) or (Y, X, C)
slice
slice(None), slice(1), slice(0, 2), slice(2, 0, -1)
Those channels
(Y, X, C)
int
0, 1, 2, -1, -2, -3
Channel idx
(Y, X)
(int, …)
[0], [1], [2], [-1], [-2], [-3], [0, 1], [-1, 2, 1]
Those channels
(Y, X, C)
-
class
buzzard.
AStored
(<implementation detail>)[source] Base abstract class defining the common behavior of all sources that are stored somewhere (like RAM or disk).
Features Defined
Has an opening mode
-
property
mode
Open mode, one of {‘r’, ‘w’}
-
class
buzzard.
AStoredRaster
(<implementation detail>)[source] Base abstract class defining the common behavior of all rasters that are stored somewhere (like RAM or disk).
Features Defined
Has a set_data method that allows to write pixels to storage
-
set_data
(array, fp=None, channels=None, interpolation='cv_area', mask=None, **kwargs)[source] Write a rectangle of data to the destination raster. Each channel in array is written to one channel in raster in the same order as described by the channels parameter. An optional mask may be provided to only write certain pixels of array.
If fp is not fully within the destination raster, only the overlapping pixels are written. If fp is not on the same grid as the destination raster, remapping is automatically performed using the interpolation algorithm. (It fails if the allow_interpolation parameter is set to False in Dataset (default)). When interpolating:
The nodata values are not interpolated, they are correctly spread to the output.
At most one pixel may be lost at edges due to interpolation. Provide more context in array to compensate this loss.
The mask parameter is also interpolated.
The alpha bands are currently resampled like any other band, this behavior may change in the future.
This method is not thread-safe.
Parameters
- array: numpy.ndarray of shape (Y, X) or (Y, X, C)
The values to be written
- fp: Footprint of shape (Y, X) or None
If None: write the full source raster If Footprint: write this window to the raster
- channels: None or int or slice or sequence of int (see Channels Parameter below)
The channels to be written.
- interpolation: one of {‘cv_area’, ‘cv_nearest’, ‘cv_linear’, ‘cv_cubic’, ‘cv_lanczos4’} or None
OpenCV method used if intepolation is necessary
- mask: numpy array of shape (Y, X) and dtype bool OR inputs accepted by Footprint.burn_polygons
Channels Parameter
type
value
meaning
NoneType
None (default)
All channels
slice
slice(None), slice(1), slice(0, 2), slice(2, 0, -1)
Those channels
int
0, 1, 2, -1, -2, -3
Channel idx
(int, …)
[0], [1], [2], [-1], [-2], [-3], [0, 1], [-1, 2, 1]
Those channels
Caveat
When using a Raster backed by a driver (like a GDAL driver), the data might be flushed to disk only after the garbage collection of the driver object. To be absolutely sure that the driver cache is flushed to disk, call .close or .deactivate on this Raster.
-
fill
(value, channels=None, **kwargs)[source] Fill raster with value.
This method is not thread-safe.
Parameters
- value: nbr
- channels: int or sequence of int (see Channels Parameter below)
The channels to be written
Channels Parameter
type
value
meaning
NoneType
None (default)
All channels
slice
slice(None), slice(1), slice(0, 2), slice(2, 0, -1)
Those channels
int
0, 1, 2, -1, -2, -3
Channel idx
(int, …)
[0], [1], [2], [-1], [-2], [-3], [0, 1], [-1, 2, 1]
Those channels
Caveat
When using a Raster backed by a driver (like a GDAL driver), the data might be flushed to disk only after the garbage collection of the driver object. To be absolutely sure that the driver cache is flushed to disk, call .close or .deactivate on this Raster.
-
class
buzzard.
NumpyRaster
(<implementation detail>)[source] Concrete class defining the behavior of a wrapped numpy array
>>> help(Dataset.wrap_numpy_raster)
Features Defined
Has an array property that points to the numpy array provided at construction.
-
property
array
Returns the Raster’s full input data as a Numpy array