def add_band(self,
dtype,
band_idx,
color_interp,
nodata=None,
hidenodata=False):
vrtrasterband = ET.SubElement(self.vrtdataset, 'VRTRasterBand')
dtype = dtype if isinstance(dtype, str) else dtype.name
vrtrasterband.attrib['dataType'] = _gdal_typename(
dtype) if check_dtype(dtype) else dtype
vrtrasterband.attrib['band'] = str(band_idx)
if nodata is not None:
nodatavalue = ET.SubElement(vrtrasterband, 'NoDataValue')
nodatavalue.text = str(nodata)
if hidenodata:
hidenodatavalue = ET.SubElement(vrtrasterband,
'HideNoDataValue')
hidenodatavalue.text = "1" if hidenodata else "0"
colorinterp = ET.SubElement(vrtrasterband, 'ColorInterp')
colorinterp.text = color_interp.capitalize()
return vrtrasterband
def _setup_band_simplesource(self, simplesource, band_idx, dtype, relative_to_vrt, file_name,
rasterxsize, rasterysize, blockxsize, blockysize, nodata):
sourcefilename = ET.SubElement(simplesource, 'SourceFilename')
sourcefilename.attrib['relativeToVRT'] = "1" if relative_to_vrt else "0"
sourcefilename.text = vsi_path(parse_path(file_name))
sourceband = ET.SubElement(simplesource, 'SourceBand')
sourceband.text = str(band_idx)
sourceproperties = ET.SubElement(simplesource, 'SourceProperties')
sourceproperties.attrib['RasterXSize'] = str(rasterxsize)
sourceproperties.attrib['RasterYSize'] = str(rasterysize)
if blockxsize is not None and blockysize is not None:
sourceproperties.attrib['BlockXSize'] = str(blockxsize)
sourceproperties.attrib['BlockYSize'] = str(blockysize)
dtype = dtype if isinstance(dtype, str) else dtype.name
sourceproperties.attrib['DataType'] = _gdal_typename(dtype) if check_dtype(dtype) else dtype
def open(fp, mode='r', driver=None, width=None, height=None, count=None,
crs=None, transform=None, dtype=None, nodata=None, **kwargs):
"""Open a dataset for reading or writing.
The dataset may be located in a local file, in a resource located by
a URL, or contained within a stream of bytes.
In read ('r') or read/write ('r+') mode, no keyword arguments are
required: these attributes are supplied by the opened dataset.
In write ('w') mode, the driver, width, height, count, and dtype
keywords are strictly required.
Parameters
----------
fp : str or file object
A filename or URL, or file object opened in binary ('rb') mode
mode : str, optional
'r' (read, the default), 'r+' (read/write), or 'w' (write)
driver : str, optional
A short format driver name (e.g. "GTiff" or "JPEG") or a list of
such names (see GDAL docs at
http://www.gdal.org/formats_list.html). In 'w' mode a single
name is required. In 'r' or 'r+' mode the driver can usually be
omitted. Registered drivers will be tried sequentially until a
match is found. When multiple drivers are available for a format
such as JPEG2000, one of them can be selected by using this
keyword argument.
width, height : int, optional
The numbers of rows and columns of the raster dataset. Required
in 'w' mode, they are ignored in 'r' or 'r+' mode.
count : int, optional
The count of dataset bands. Required in 'w' mode, it is ignored
in 'r' or 'r+' mode.
dtype : str or numpy dtype
The data type for bands. For example: 'uint8' or
``rasterio.uint16``. Required in 'w' mode, it is ignored in
'r' or 'r+' mode.
crs : str, dict, or CRS; optional
The coordinate reference system. Required in 'w' mode, it is
ignored in 'r' or 'r+' mode.
transform : Affine instance, optional
Affine transformation mapping the pixel space to geographic
space. Required in 'w' mode, it is ignored in 'r' or 'r+' mode.
nodata : int, float, or nan; optional
Defines the pixel value to be interpreted as not valid data.
Required in 'w' mode, it is ignored in 'r' or 'r+' mode.
kwargs : optional
These are passed to format drivers as directives for creating
or interpreting datasets. For example: in 'w' a `tiled=True`
keyword argument will direct the GeoTIFF format driver to
create a tiled, rather than striped, TIFF.
Returns
-------
A ``DatasetReader`` or ``DatasetUpdater`` object.
Examples
--------
To open a GeoTIFF for reading using standard driver discovery and
no directives:
>>> import rasterio
>>> with rasterio.open('example.tif') as dataset:
... print(dataset.profile)
To open a JPEG2000 using only the JP2OpenJPEG driver:
>>> with rasterio.open(
... 'example.jp2', driver='JP2OpenJPEG') as dataset:
... print(dataset.profile)
To create a new 8-band, 16-bit unsigned, tiled, and LZW-compressed
GeoTIFF with a global extent and 0.5 degree resolution:
>>> from rasterio.transform import from_origin
>>> with rasterio.open(
... 'example.tif', 'w', driver='GTiff', dtype='uint16',
... width=720, height=360, count=8, crs='EPSG:4326',
... transform=from_origin(-180.0, 90.0, 0.5, 0.5),
... nodata=0, tiled=True, compress='lzw') as dataset:
... dataset.write(...)
"""
if not isinstance(fp, string_types):
if not (hasattr(fp, 'read') or hasattr(fp, 'write')):
raise TypeError("invalid path or file: {0!r}".format(fp))
if mode and not isinstance(mode, string_types):
raise TypeError("invalid mode: {0!r}".format(mode))
if driver and not isinstance(driver, string_types):
raise TypeError("invalid driver: {0!r}".format(driver))
if dtype and not check_dtype(dtype):
raise TypeError("invalid dtype: {0!r}".format(dtype))
if nodata is not None:
nodata = float(nodata)
if 'affine' in kwargs:
# DeprecationWarning's are ignored by default
with warnings.catch_warnings():
warnings.warn(
"The 'affine' kwarg in rasterio.open() is deprecated at 1.0 "
"and only remains to ease the transition. Please switch to "
"the 'transform' kwarg. See "
"https://github.com/mapbox/rasterio/issues/86 for details.",
DeprecationWarning,
stacklevel=2)
if transform:
warnings.warn(
"Found both 'affine' and 'transform' in rasterio.open() - "
"choosing 'transform'")
transform = transform
else:
transform = kwargs.pop('affine')
if transform:
transform = guard_transform(transform)
# Check driver/mode blacklist.
if driver and is_blacklisted(driver, mode):
raise RasterioIOError(
"Blacklisted: file cannot be opened by "
"driver '{0}' in '{1}' mode".format(driver, mode))
# Special case for file object argument.
if mode == 'r' and hasattr(fp, 'read'):
@contextmanager
def fp_reader(fp):
memfile = MemoryFile(fp.read())
dataset = memfile.open()
try:
yield dataset
finally:
dataset.close()
memfile.close()
return fp_reader(fp)
elif mode == 'w' and hasattr(fp, 'write'):
@contextmanager
def fp_writer(fp):
memfile = MemoryFile()
dataset = memfile.open(driver=driver, width=width, height=height,
count=count, crs=crs, transform=transform,
dtype=dtype, nodata=nodata, **kwargs)
try:
yield dataset
finally:
dataset.close()
memfile.seek(0)
fp.write(memfile.read())
memfile.close()
return fp_writer(fp)
else:
# The 'normal' filename or URL path.
_, _, scheme = parse_path(fp)
with Env() as env:
if scheme == 's3':
env.credentialize()
# Create dataset instances and pass the given env, which will
# be taken over by the dataset's context manager if it is not
# None.
if mode == 'r':
s = DatasetReader(fp, driver=driver, **kwargs)
elif mode == 'r-':
warnings.warn("'r-' mode is deprecated, use 'r'",
DeprecationWarning)
s = DatasetReader(fp)
elif mode == 'r+':
s = get_writer_for_path(fp)(fp, mode, driver=driver, **kwargs)
elif mode == 'w':
s = get_writer_for_driver(driver)(fp, mode, driver=driver,
width=width, height=height,
count=count, crs=crs,
transform=transform,
dtype=dtype, nodata=nodata,
**kwargs)
else:
raise ValueError(
"mode must be one of 'r', 'r+', or 'w', not %s" % mode)
return s
def open(fp, mode='r', driver=None, width=None, height=None, count=None,
crs=None, transform=None, dtype=None, nodata=None, sharing=True,
**kwargs):
"""Open a dataset for reading or writing.
The dataset may be located in a local file, in a resource located by
a URL, or contained within a stream of bytes.
In read ('r') or read/write ('r+') mode, no keyword arguments are
required: these attributes are supplied by the opened dataset.
In write ('w' or 'w+') mode, the driver, width, height, count, and dtype
keywords are strictly required.
Parameters
----------
fp : str, file object or pathlib.Path object
A filename or URL, a file object opened in binary ('rb') mode,
or a Path object.
mode : str, optional
'r' (read, the default), 'r+' (read/write), 'w' (write), or
'w+' (write/read).
driver : str, optional
A short format driver name (e.g. "GTiff" or "JPEG") or a list of
such names (see GDAL docs at
http://www.gdal.org/formats_list.html). In 'w' or 'w+' modes
a single name is required. In 'r' or 'r+' modes the driver can
usually be omitted. Registered drivers will be tried
sequentially until a match is found. When multiple drivers are
available for a format such as JPEG2000, one of them can be
selected by using this keyword argument.
width, height : int, optional
The numbers of rows and columns of the raster dataset. Required
in 'w' or 'w+' modes, they are ignored in 'r' or 'r+' modes.
count : int, optional
The count of dataset bands. Required in 'w' or 'w+' modes, it is
ignored in 'r' or 'r+' modes.
dtype : str or numpy dtype
The data type for bands. For example: 'uint8' or
``rasterio.uint16``. Required in 'w' or 'w+' modes, it is
ignored in 'r' or 'r+' modes.
crs : str, dict, or CRS; optional
The coordinate reference system. Required in 'w' or 'w+' modes,
it is ignored in 'r' or 'r+' modes.
transform : Affine instance, optional
Affine transformation mapping the pixel space to geographic
space. Required in 'w' or 'w+' modes, it is ignored in 'r' or
'r+' modes.
nodata : int, float, or nan; optional
Defines the pixel value to be interpreted as not valid data.
Required in 'w' or 'w+' modes, it is ignored in 'r' or 'r+'
modes.
sharing : bool
A flag that allows sharing of dataset handles. Default is
`True`. Should be set to `False` in a multithreaded:w program.
kwargs : optional
These are passed to format drivers as directives for creating or
interpreting datasets. For example: in 'w' or 'w+' modes
a `tiled=True` keyword argument will direct the GeoTIFF format
driver to create a tiled, rather than striped, TIFF.
Returns
-------
A ``DatasetReader`` or ``DatasetUpdater`` object.
Examples
--------
To open a GeoTIFF for reading using standard driver discovery and
no directives:
>>> import rasterio
>>> with rasterio.open('example.tif') as dataset:
... print(dataset.profile)
To open a JPEG2000 using only the JP2OpenJPEG driver:
>>> with rasterio.open(
... 'example.jp2', driver='JP2OpenJPEG') as dataset:
... print(dataset.profile)
To create a new 8-band, 16-bit unsigned, tiled, and LZW-compressed
GeoTIFF with a global extent and 0.5 degree resolution:
>>> from rasterio.transform import from_origin
>>> with rasterio.open(
... 'example.tif', 'w', driver='GTiff', dtype='uint16',
... width=720, height=360, count=8, crs='EPSG:4326',
... transform=from_origin(-180.0, 90.0, 0.5, 0.5),
... nodata=0, tiled=True, compress='lzw') as dataset:
... dataset.write(...)
"""
if not isinstance(fp, string_types):
if not (hasattr(fp, 'read') or hasattr(fp, 'write') or isinstance(fp, Path)):
raise TypeError("invalid path or file: {0!r}".format(fp))
if mode and not isinstance(mode, string_types):
raise TypeError("invalid mode: {0!r}".format(mode))
if driver and not isinstance(driver, string_types):
raise TypeError("invalid driver: {0!r}".format(driver))
if dtype and not check_dtype(dtype):
raise TypeError("invalid dtype: {0!r}".format(dtype))
if nodata is not None:
nodata = float(nodata)
if transform:
transform = guard_transform(transform)
# Check driver/mode blacklist.
if driver and is_blacklisted(driver, mode):
raise RasterioIOError(
"Blacklisted: file cannot be opened by "
"driver '{0}' in '{1}' mode".format(driver, mode))
# Special case for file object argument.
if mode == 'r' and hasattr(fp, 'read'):
@contextmanager
def fp_reader(fp):
memfile = MemoryFile(fp.read())
dataset = memfile.open()
try:
yield dataset
finally:
dataset.close()
memfile.close()
return fp_reader(fp)
elif mode in ('w', 'w+') and hasattr(fp, 'write'):
@contextmanager
def fp_writer(fp):
memfile = MemoryFile()
dataset = memfile.open(driver=driver, width=width, height=height,
count=count, crs=crs, transform=transform,
dtype=dtype, nodata=nodata, **kwargs)
try:
yield dataset
finally:
dataset.close()
memfile.seek(0)
fp.write(memfile.read())
memfile.close()
return fp_writer(fp)
else:
# If a pathlib.Path instance is given, convert it to a string path.
if isinstance(fp, Path):
fp = str(fp)
# The 'normal' filename or URL path.
path = parse_path(fp)
# Create dataset instances and pass the given env, which will
# be taken over by the dataset's context manager if it is not
# None.
if mode == 'r':
s = DatasetReader(path, driver=driver, **kwargs)
elif mode == 'r+':
s = get_writer_for_path(path)(path, mode, driver=driver, **kwargs)
elif mode.startswith("w"):
s = get_writer_for_driver(driver)(path, mode, driver=driver,
width=width, height=height,
count=count, crs=crs,
transform=transform,
dtype=dtype, nodata=nodata,
**kwargs)
else:
raise ValueError(
"mode must be one of 'r', 'r+', or 'w', not %s" % mode)
return s
def test_check_dtype_invalid():
assert check_dtype('foo') == False
def test_dt_ubyte():
assert check_dtype(ubyte)
def test_np_dt_uint8():
assert check_dtype(np.uint8)
def open(path, mode='r', driver=None, width=None, height=None,
count=None, crs=None, transform=None, dtype=None, nodata=None,
**kwargs):
"""Open file at ``path`` in ``mode`` 'r' (read), 'r+' (read and
write), or 'w' (write) and return a dataset Reader or Updater
object.
In write mode, a driver name such as "GTiff" or "JPEG" (see GDAL
docs or ``gdal_translate --help`` on the command line),
``width`` (number of pixels per line) and ``height`` (number of
lines), the ``count`` number of bands in the new file must be
specified. Additionally, the data type for bands such as
``rasterio.ubyte`` for 8-bit bands or ``rasterio.uint16`` for
16-bit bands must be specified using the ``dtype`` argument.
Parameters
----------
mode: string
"r" (read), "r+" (read/write), or "w" (write)
driver: string
driver code specifying the format name (e.g. "GTiff" or
"JPEG"). See GDAL docs at
http://www.gdal.org/formats_list.html (optional, required
for writing).
width: int
number of pixels per line
(optional, required for write)
height: int
number of lines
(optional, required for write)
count: int > 0
number of bands
(optional, required for write)
dtype: rasterio.dtype
the data type for bands such as ``rasterio.ubyte`` for
8-bit bands or ``rasterio.uint16`` for 16-bit bands
(optional, required for write)
crs: dict or string
Coordinate reference system
(optional, recommended for write)
transform: Affine instance
Affine transformation mapping the pixel space to geographic
space (optional, recommended for writing).
nodata: number
Defines pixel value to be interpreted as null/nodata
(optional, recommended for write)
Returns
-------
A ``DatasetReader`` or ``DatasetUpdater`` object.
Notes
-----
In write mode, you must specify at least ``width``, ``height``,
``count`` and ``dtype``.
A coordinate reference system for raster datasets in write mode
can be defined by the ``crs`` argument. It takes Proj4 style
mappings like
.. code::
{'proj': 'longlat', 'ellps': 'WGS84', 'datum': 'WGS84',
'no_defs': True}
An affine transformation that maps ``col,row`` pixel coordinates
to ``x,y`` coordinates in the coordinate reference system can be
specified using the ``transform`` argument. The value should be
an instance of ``affine.Affine``
.. code:: python
>>> from affine import Affine
>>> Affine(0.5, 0.0, -180.0, 0.0, -0.5, 90.0)
These coefficients are shown in the figure below.
.. code::
| x | | a b c | | c |
| y | = | d e f | | r |
| 1 | | 0 0 1 | | 1 |
a: rate of change of X with respect to increasing column,
i.e. pixel width
b: rotation, 0 if the raster is oriented "north up"
c: X coordinate of the top left corner of the top left pixel
d: rotation, 0 if the raster is oriented "north up"
e: rate of change of Y with respect to increasing row,
usually a negative number (i.e. -1 * pixel height) if
north-up.
f: Y coordinate of the top left corner of the top left pixel
A 6-element sequence of the affine transformation matrix
coefficients in ``c, a, b, f, d, e`` order, (i.e. GDAL
geotransform order) will be accepted until 1.0 (deprecated).
A virtual filesystem can be specified. The ``vfs`` parameter may
be an Apache Commons VFS style string beginning with "zip://" or
"tar://"". In this case, the ``path`` must be an absolute path
within that container.
"""
if not isinstance(path, string_types):
raise TypeError("invalid path: {0!r}".format(path))
if mode and not isinstance(mode, string_types):
raise TypeError("invalid mode: {0!r}".format(mode))
if driver and not isinstance(driver, string_types):
raise TypeError("invalid driver: {0!r}".format(driver))
if dtype and not check_dtype(dtype):
raise TypeError("invalid dtype: {0!r}".format(dtype))
if transform:
transform = guard_transform(transform)
elif 'affine' in kwargs:
affine = kwargs.pop('affine')
transform = guard_transform(affine)
# If there is no currently active GDAL/AWS environment, create one.
defenv()
# Get AWS credentials if we're attempting to access a raster
# on S3.
pth, archive, scheme = parse_path(path)
if scheme == 's3':
Env().get_aws_credentials()
log.debug("AWS credentials have been obtained")
# Create dataset instances and pass the given env, which will
# be taken over by the dataset's context manager if it is not
# None.
if mode == 'r':
from rasterio._io import RasterReader
s = RasterReader(path)
elif mode == 'r+':
from rasterio._io import writer
s = writer(path, mode)
elif mode == 'r-':
from rasterio._base import DatasetReader
s = DatasetReader(path)
elif mode == 'w':
from rasterio._io import writer
s = writer(path, mode, driver=driver,
width=width, height=height, count=count,
crs=crs, transform=transform, dtype=dtype,
nodata=nodata, **kwargs)
else:
raise ValueError(
"mode string must be one of 'r', 'r+', or 'w', not %s" % mode)
s.start()
return s
def test_dt_ubyte():
assert check_dtype(ubyte)
def test_np_dt_uint8():
assert check_dtype(np.uint8)
def open(fp, mode='r', driver=None, width=None, height=None, count=None,
crs=None, transform=None, dtype=None, nodata=None, sharing=False,
**kwargs):
"""Open a dataset for reading or writing.
The dataset may be located in a local file, in a resource located by
a URL, or contained within a stream of bytes.
In read ('r') or read/write ('r+') mode, no keyword arguments are
required: these attributes are supplied by the opened dataset.
In write ('w' or 'w+') mode, the driver, width, height, count, and dtype
keywords are strictly required.
Parameters
----------
fp : str, file object, PathLike object, FilePath, or MemoryFile
A filename or URL, a file object opened in binary ('rb') mode, a
Path object, or one of the rasterio classes that provides the
dataset-opening interface (has an open method that returns a
dataset).
mode : str, optional
'r' (read, the default), 'r+' (read/write), 'w' (write), or
'w+' (write/read).
driver : str, optional
A short format driver name (e.g. "GTiff" or "JPEG") or a list of
such names (see GDAL docs at
https://gdal.org/drivers/raster/index.html). In 'w' or 'w+' modes
a single name is required. In 'r' or 'r+' modes the driver can
usually be omitted. Registered drivers will be tried
sequentially until a match is found. When multiple drivers are
available for a format such as JPEG2000, one of them can be
selected by using this keyword argument.
width : int, optional
The number of columns of the raster dataset. Required in 'w' or
'w+' modes, it is ignored in 'r' or 'r+' modes.
height : int, optional
The number of rows of the raster dataset. Required in 'w' or
'w+' modes, it is ignored in 'r' or 'r+' modes.
count : int, optional
The count of dataset bands. Required in 'w' or 'w+' modes, it is
ignored in 'r' or 'r+' modes.
crs : str, dict, or CRS; optional
The coordinate reference system. Required in 'w' or 'w+' modes,
it is ignored in 'r' or 'r+' modes.
transform : Affine instance, optional
Affine transformation mapping the pixel space to geographic
space. Required in 'w' or 'w+' modes, it is ignored in 'r' or
'r+' modes.
dtype : str or numpy dtype
The data type for bands. For example: 'uint8' or
``rasterio.uint16``. Required in 'w' or 'w+' modes, it is
ignored in 'r' or 'r+' modes.
nodata : int, float, or nan; optional
Defines the pixel value to be interpreted as not valid data.
Required in 'w' or 'w+' modes, it is ignored in 'r' or 'r+'
modes.
sharing : bool; optional
To reduce overhead and prevent programs from running out of file
descriptors, rasterio maintains a pool of shared low level
dataset handles. When `True` this function will use a shared
handle if one is available. Multithreaded programs must avoid
sharing and should set *sharing* to `False`.
kwargs : optional
These are passed to format drivers as directives for creating or
interpreting datasets. For example: in 'w' or 'w+' modes
a `tiled=True` keyword argument will direct the GeoTIFF format
driver to create a tiled, rather than striped, TIFF.
Returns
-------
A ``DatasetReader`` or ``DatasetWriter`` object.
Examples
--------
To open a GeoTIFF for reading using standard driver discovery and
no directives:
>>> import rasterio
>>> with rasterio.open('example.tif') as dataset:
... print(dataset.profile)
To open a JPEG2000 using only the JP2OpenJPEG driver:
>>> with rasterio.open(
... 'example.jp2', driver='JP2OpenJPEG') as dataset:
... print(dataset.profile)
To create a new 8-band, 16-bit unsigned, tiled, and LZW-compressed
GeoTIFF with a global extent and 0.5 degree resolution:
>>> from rasterio.transform import from_origin
>>> with rasterio.open(
... 'example.tif', 'w', driver='GTiff', dtype='uint16',
... width=720, height=360, count=8, crs='EPSG:4326',
... transform=from_origin(-180.0, 90.0, 0.5, 0.5),
... nodata=0, tiled=True, compress='lzw') as dataset:
... dataset.write(...)
"""
if not isinstance(fp, str):
if not (
hasattr(fp, "read")
or hasattr(fp, "write")
or isinstance(fp, (os.PathLike, MemoryFile, FilePath))
):
raise TypeError("invalid path or file: {0!r}".format(fp))
if mode and not isinstance(mode, str):
raise TypeError("invalid mode: {0!r}".format(mode))
if driver and not isinstance(driver, str):
raise TypeError("invalid driver: {0!r}".format(driver))
if dtype and not check_dtype(dtype):
raise TypeError("invalid dtype: {0!r}".format(dtype))
if nodata is not None:
nodata = float(nodata)
if transform:
transform = guard_transform(transform)
# Check driver/mode blacklist.
if driver and is_blacklisted(driver, mode):
raise RasterioIOError(
"Blacklisted: file cannot be opened by "
"driver '{0}' in '{1}' mode".format(driver, mode))
# If the fp argument is a file-like object and can be adapted by
# rasterio's FilePath we do so. Otherwise, we use a MemoryFile to
# hold fp's contents and store that in an ExitStack attached to the
# dataset object that we will return. When a dataset's close method
# is called, this ExitStack will be unwound and the MemoryFile's
# storage will be cleaned up.
if mode == 'r' and hasattr(fp, 'read'):
if have_vsi_plugin:
return FilePath(fp).open(driver=driver, sharing=sharing, **kwargs)
else:
memfile = MemoryFile(fp.read())
dataset = memfile.open(driver=driver, sharing=sharing, **kwargs)
dataset._env.enter_context(memfile)
return dataset
elif mode in ('w', 'w+') and hasattr(fp, 'write'):
memfile = MemoryFile()
dataset = memfile.open(
driver=driver,
width=width,
height=height,
count=count,
crs=crs,
transform=transform,
dtype=dtype,
nodata=nodata,
sharing=sharing,
**kwargs
)
dataset._env.enter_context(memfile)
# For the writing case we push an extra callback onto the
# ExitStack. It ensures that the MemoryFile's contents are
# copied to the open file object.
def func(*args, **kwds):
memfile.seek(0)
fp.write(memfile.read())
dataset._env.callback(func)
return dataset
# TODO: test for a shared base class or abstract type.
elif isinstance(fp, (FilePath, MemoryFile)):
if mode.startswith("r"):
dataset = fp.open(driver=driver, sharing=sharing, **kwargs)
# Note: FilePath does not support writing and an exception will
# result from this.
elif mode.startswith("w"):
dataset = fp.open(
driver=driver,
width=width,
height=height,
count=count,
crs=crs,
transform=transform,
dtype=dtype,
nodata=nodata,
sharing=sharing,
**kwargs
)
return dataset
# At this point, the fp argument is a string or path-like object
# which can be converted to a string.
else:
raw_dataset_path = os.fspath(fp)
path = _parse_path(raw_dataset_path)
if mode == "r":
dataset = DatasetReader(path, driver=driver, sharing=sharing, **kwargs)
elif mode == "r+":
dataset = get_writer_for_path(path, driver=driver)(
path, mode, driver=driver, sharing=sharing, **kwargs
)
elif mode.startswith("w"):
if not driver:
driver = driver_from_extension(path)
writer = get_writer_for_driver(driver)
if writer is not None:
dataset = writer(
path,
mode,
driver=driver,
width=width,
height=height,
count=count,
crs=crs,
transform=transform,
dtype=dtype,
nodata=nodata,
sharing=sharing,
**kwargs
)
else:
raise DriverCapabilityError(
"Writer does not exist for driver: %s" % str(driver)
)
else:
raise DriverCapabilityError(
"mode must be one of 'r', 'r+', or 'w', not %s" % mode)
return dataset
def add_mask_band(self, dtype):
maskband = ET.SubElement(self.vrtdataset, 'MaskBand')
vrtrasterband = ET.SubElement(maskband, 'VRTRasterBand')
vrtrasterband.attrib['dataType'] = _gdal_typename(dtype) if check_dtype(dtype) else dtype
return vrtrasterband
def test_check_dtype_invalid():
assert check_dtype('foo') == False
def open(path,
mode='r',
driver=None,
width=None,
height=None,
count=None,
crs=None,
transform=None,
dtype=None,
nodata=None,
**kwargs):
"""Open file at ``path`` in ``mode`` 'r' (read), 'r+' (read and
write), or 'w' (write) and return a dataset Reader or Updater
object.
In write mode, a driver name such as "GTiff" or "JPEG" (see GDAL
docs or ``gdal_translate --help`` on the command line),
``width`` (number of pixels per line) and ``height`` (number of
lines), the ``count`` number of bands in the new file must be
specified. Additionally, the data type for bands such as
``rasterio.ubyte`` for 8-bit bands or ``rasterio.uint16`` for
16-bit bands must be specified using the ``dtype`` argument.
Parameters
----------
mode: string
"r" (read), "r+" (read/write), or "w" (write)
driver: string
driver code specifying the format name (e.g. "GTiff" or
"JPEG"). See GDAL docs at
http://www.gdal.org/formats_list.html (optional, required
for writing).
width: int
number of pixels per line
(optional, required for write)
height: int
number of lines
(optional, required for write)
count: int > 0
number of bands
(optional, required for write)
dtype: rasterio.dtype
the data type for bands such as ``rasterio.ubyte`` for
8-bit bands or ``rasterio.uint16`` for 16-bit bands
(optional, required for write)
crs: dict or string
Coordinate reference system
(optional, recommended for write)
transform: Affine instance
Affine transformation mapping the pixel space to geographic
space (optional, recommended for writing).
nodata: number
Defines pixel value to be interpreted as null/nodata
(optional, recommended for write)
Returns
-------
A ``DatasetReader`` or ``DatasetUpdater`` object.
Notes
-----
In write mode, you must specify at least ``width``, ``height``,
``count`` and ``dtype``.
A coordinate reference system for raster datasets in write mode
can be defined by the ``crs`` argument. It takes Proj4 style
mappings like
.. code::
{'proj': 'longlat', 'ellps': 'WGS84', 'datum': 'WGS84',
'no_defs': True}
An affine transformation that maps ``col,row`` pixel coordinates
to ``x,y`` coordinates in the coordinate reference system can be
specified using the ``transform`` argument. The value should be
an instance of ``affine.Affine``
.. code:: python
>>> from affine import Affine
>>> transform = Affine(0.5, 0.0, -180.0, 0.0, -0.5, 90.0)
These coefficients are shown in the figure below.
.. code::
| x | | a b c | | c |
| y | = | d e f | | r |
| 1 | | 0 0 1 | | 1 |
a: rate of change of X with respect to increasing column,
i.e. pixel width
b: rotation, 0 if the raster is oriented "north up"
c: X coordinate of the top left corner of the top left pixel
d: rotation, 0 if the raster is oriented "north up"
e: rate of change of Y with respect to increasing row,
usually a negative number (i.e. -1 * pixel height) if
north-up.
f: Y coordinate of the top left corner of the top left pixel
A 6-element sequence of the affine transformation matrix
coefficients in ``c, a, b, f, d, e`` order, (i.e. GDAL
geotransform order) will be accepted until 1.0 (deprecated).
A virtual filesystem can be specified. The ``vfs`` parameter may
be an Apache Commons VFS style string beginning with "zip://" or
"tar://"". In this case, the ``path`` must be an absolute path
within that container.
"""
if not isinstance(path, string_types):
raise TypeError("invalid path: {0!r}".format(path))
if mode and not isinstance(mode, string_types):
raise TypeError("invalid mode: {0!r}".format(mode))
if driver and not isinstance(driver, string_types):
raise TypeError("invalid driver: {0!r}".format(driver))
if dtype and not check_dtype(dtype):
raise TypeError("invalid dtype: {0!r}".format(dtype))
if transform:
transform = guard_transform(transform)
elif 'affine' in kwargs:
affine = kwargs.pop('affine')
transform = guard_transform(affine)
# Get AWS credentials if we're attempting to access a raster
# on S3.
pth, archive, scheme = parse_path(path)
if scheme == 's3':
Env().get_aws_credentials()
log.debug("AWS credentials have been obtained")
# Create dataset instances and pass the given env, which will
# be taken over by the dataset's context manager if it is not
# None.
if mode == 'r':
from rasterio._io import RasterReader
s = RasterReader(path)
elif mode == 'r+':
from rasterio._io import writer
s = writer(path, mode)
elif mode == 'r-':
from rasterio._base import DatasetReader
s = DatasetReader(path)
elif mode == 'w':
from rasterio._io import writer
s = writer(path,
mode,
driver=driver,
width=width,
height=height,
count=count,
crs=crs,
transform=transform,
dtype=dtype,
nodata=nodata,
**kwargs)
else:
raise ValueError(
"mode string must be one of 'r', 'r+', or 'w', not %s" % mode)
s.start()
return s
def open(fp,
mode='r',
driver=None,
width=None,
height=None,
count=None,
crs=None,
transform=None,
dtype=None,
nodata=None,
**kwargs):
"""Open a dataset for reading or writing.
The dataset may be located in a local file, in a resource located
by a URL, or contained within a stream of bytes.
To access a dataset within a zip file without unzipping the archive
use an Apache VFS style zip:// URL like
zip://path/to/archive.zip!path/to/example.tif
In read ('r') or read/write ('r+') mode, no other keyword arguments
are required: the attributes are supplied by the opened dataset.
In write mode, a driver name such as "GTiff" or "JPEG" (see GDAL
docs or ``gdal_translate --help`` on the command line), ``width``
(number of pixels per line) and ``height`` (number of lines), the
``count`` number of bands in the new file must be specified.
Additionally, the data type for bands such as ``rasterio.ubyte`` for
8-bit bands or ``rasterio.uint16`` for 16-bit bands must be
specified using the ``dtype`` argument.
Parameters
----------
fp: string or file
A filename or URL, or file object opened in binary mode.
mode: string
"r" (read), "r+" (read/write), or "w" (write)
driver: string
Driver code specifying the format name (e.g. "GTiff" or
"JPEG"). See GDAL docs at
http://www.gdal.org/formats_list.html (optional, required
for writing).
width: int
Number of pixels per line (optional, required for write).
height: int
Number of lines (optional, required for write).
count: int > 0
Count of bands (optional, required for write).
dtype: rasterio.dtype
the data type for bands such as ``rasterio.ubyte`` for
8-bit bands or ``rasterio.uint16`` for 16-bit bands
(optional, required for write)
crs: dict or string
Coordinate reference system (optional, recommended for write).
transform: Affine instance
Affine transformation mapping the pixel space to geographic
space (optional, recommended for writing).
nodata: number
Defines pixel value to be interpreted as null/nodata
(optional, recommended for write, will be broadcast to all
bands).
Returns
-------
A ``DatasetReader`` or ``DatasetUpdater`` object.
Notes
-----
In write mode, you must specify at least ``width``, ``height``,
``count`` and ``dtype``.
A coordinate reference system for raster datasets in write mode
can be defined by the ``crs`` argument. It takes Proj4 style
mappings like
.. code::
{'proj': 'longlat', 'ellps': 'WGS84', 'datum': 'WGS84',
'no_defs': True}
An affine transformation that maps ``col,row`` pixel coordinates
to ``x,y`` coordinates in the coordinate reference system can be
specified using the ``transform`` argument. The value should be
an instance of ``affine.Affine``
.. code:: python
>>> from affine import Affine
>>> transform = Affine(0.5, 0.0, -180.0, 0.0, -0.5, 90.0)
These coefficients are shown in the figure below.
.. code::
| x | | a b c | | c |
| y | = | d e f | | r |
| 1 | | 0 0 1 | | 1 |
a: rate of change of X with respect to increasing column,
i.e. pixel width
b: rotation, 0 if the raster is oriented "north up"
c: X coordinate of the top left corner of the top left pixel
d: rotation, 0 if the raster is oriented "north up"
e: rate of change of Y with respect to increasing row,
usually a negative number (i.e. -1 * pixel height) if
north-up.
f: Y coordinate of the top left corner of the top left pixel
A 6-element sequence of the affine transformation matrix
coefficients in ``c, a, b, f, d, e`` order, (i.e. GDAL
geotransform order) will be accepted until 1.0 (deprecated).
A virtual filesystem can be specified. The ``vfs`` parameter may
be an Apache Commons VFS style string beginning with "zip://" or
"tar://"". In this case, the ``path`` must be an absolute path
within that container.
"""
if not isinstance(fp, string_types):
if not (hasattr(fp, 'read') or hasattr(fp, 'write')):
raise TypeError("invalid path or file: {0!r}".format(fp))
if mode and not isinstance(mode, string_types):
raise TypeError("invalid mode: {0!r}".format(mode))
if driver and not isinstance(driver, string_types):
raise TypeError("invalid driver: {0!r}".format(driver))
if dtype and not check_dtype(dtype):
raise TypeError("invalid dtype: {0!r}".format(dtype))
if nodata is not None:
nodata = float(nodata)
if 'affine' in kwargs:
# DeprecationWarning's are ignored by default
with warnings.catch_warnings():
warnings.warn(
"The 'affine' kwarg in rasterio.open() is deprecated at 1.0 "
"and only remains to ease the transition. Please switch to "
"the 'transform' kwarg. See "
"https://github.com/mapbox/rasterio/issues/86 for details.",
DeprecationWarning,
stacklevel=2)
if transform:
warnings.warn(
"Found both 'affine' and 'transform' in rasterio.open() - "
"choosing 'transform'")
transform = transform
else:
transform = kwargs.pop('affine')
if transform:
transform = guard_transform(transform)
# Check driver/mode blacklist.
if driver and is_blacklisted(driver, mode):
raise RasterioIOError("Blacklisted: file cannot be opened by "
"driver '{0}' in '{1}' mode".format(
driver, mode))
# Special case for file object argument.
if mode == 'r' and hasattr(fp, 'read'):
@contextmanager
def fp_reader(fp):
memfile = MemoryFile(fp.read())
dataset = memfile.open()
try:
yield dataset
finally:
dataset.close()
memfile.close()
return fp_reader(fp)
elif mode == 'w' and hasattr(fp, 'write'):
@contextmanager
def fp_writer(fp):
memfile = MemoryFile()
dataset = memfile.open(driver=driver,
width=width,
height=height,
count=count,
crs=crs,
transform=transform,
dtype=dtype,
nodata=nodata,
**kwargs)
try:
yield dataset
finally:
dataset.close()
memfile.seek(0)
fp.write(memfile.read())
memfile.close()
return fp_writer(fp)
else:
# The 'normal' filename or URL path.
_, _, scheme = parse_path(fp)
with Env() as env:
# Get AWS credentials only if we're attempting to access a
# raster using the S3 scheme.
if scheme == 's3':
env.get_aws_credentials()
log.debug("AWS credentials have been obtained")
# Create dataset instances and pass the given env, which will
# be taken over by the dataset's context manager if it is not
# None.
if mode == 'r':
s = DatasetReader(fp)
elif mode == 'r-':
warnings.warn("'r-' mode is deprecated, use 'r'",
DeprecationWarning)
s = DatasetReader(fp)
elif mode == 'r+':
s = get_writer_for_path(fp)(fp, mode)
elif mode == 'w':
s = get_writer_for_driver(driver)(fp,
mode,
driver=driver,
width=width,
height=height,
count=count,
crs=crs,
transform=transform,
dtype=dtype,
nodata=nodata,
**kwargs)
else:
raise ValueError(
"mode must be one of 'r', 'r+', or 'w', not %s" % mode)
s.start()
return s
def test_check_dtype_invalid():
assert not check_dtype('foo')
def open(fp, mode='r', driver=None, width=None, height=None, count=None,
crs=None, transform=None, dtype=None, nodata=None, sharing=True,
**kwargs):
"""Open a dataset for reading or writing.
The dataset may be located in a local file, in a resource located by
a URL, or contained within a stream of bytes.
In read ('r') or read/write ('r+') mode, no keyword arguments are
required: these attributes are supplied by the opened dataset.
In write ('w' or 'w+') mode, the driver, width, height, count, and dtype
keywords are strictly required.
Parameters
----------
fp : str, file object or pathlib.Path object
A filename or URL, a file object opened in binary ('rb') mode,
or a Path object.
mode : str, optional
'r' (read, the default), 'r+' (read/write), 'w' (write), or
'w+' (write/read).
driver : str, optional
A short format driver name (e.g. "GTiff" or "JPEG") or a list of
such names (see GDAL docs at
http://www.gdal.org/formats_list.html). In 'w' or 'w+' modes
a single name is required. In 'r' or 'r+' modes the driver can
usually be omitted. Registered drivers will be tried
sequentially until a match is found. When multiple drivers are
available for a format such as JPEG2000, one of them can be
selected by using this keyword argument.
width, height : int, optional
The numbers of rows and columns of the raster dataset. Required
in 'w' or 'w+' modes, they are ignored in 'r' or 'r+' modes.
count : int, optional
The count of dataset bands. Required in 'w' or 'w+' modes, it is
ignored in 'r' or 'r+' modes.
dtype : str or numpy dtype
The data type for bands. For example: 'uint8' or
``rasterio.uint16``. Required in 'w' or 'w+' modes, it is
ignored in 'r' or 'r+' modes.
crs : str, dict, or CRS; optional
The coordinate reference system. Required in 'w' or 'w+' modes,
it is ignored in 'r' or 'r+' modes.
transform : Affine instance, optional
Affine transformation mapping the pixel space to geographic
space. Required in 'w' or 'w+' modes, it is ignored in 'r' or
'r+' modes.
nodata : int, float, or nan; optional
Defines the pixel value to be interpreted as not valid data.
Required in 'w' or 'w+' modes, it is ignored in 'r' or 'r+'
modes.
sharing : bool
A flag that allows sharing of dataset handles. Default is
`True`. Should be set to `False` in a multithreaded:w program.
kwargs : optional
These are passed to format drivers as directives for creating or
interpreting datasets. For example: in 'w' or 'w+' modes
a `tiled=True` keyword argument will direct the GeoTIFF format
driver to create a tiled, rather than striped, TIFF.
Returns
-------
A ``DatasetReader`` or ``DatasetWriter`` object.
Examples
--------
To open a GeoTIFF for reading using standard driver discovery and
no directives:
>>> import rasterio
>>> with rasterio.open('example.tif') as dataset:
... print(dataset.profile)
To open a JPEG2000 using only the JP2OpenJPEG driver:
>>> with rasterio.open(
... 'example.jp2', driver='JP2OpenJPEG') as dataset:
... print(dataset.profile)
To create a new 8-band, 16-bit unsigned, tiled, and LZW-compressed
GeoTIFF with a global extent and 0.5 degree resolution:
>>> from rasterio.transform import from_origin
>>> with rasterio.open(
... 'example.tif', 'w', driver='GTiff', dtype='uint16',
... width=720, height=360, count=8, crs='EPSG:4326',
... transform=from_origin(-180.0, 90.0, 0.5, 0.5),
... nodata=0, tiled=True, compress='lzw') as dataset:
... dataset.write(...)
"""
if not isinstance(fp, string_types):
if not (hasattr(fp, 'read') or hasattr(fp, 'write') or isinstance(fp, Path)):
raise TypeError("invalid path or file: {0!r}".format(fp))
if mode and not isinstance(mode, string_types):
raise TypeError("invalid mode: {0!r}".format(mode))
if driver and not isinstance(driver, string_types):
raise TypeError("invalid driver: {0!r}".format(driver))
if dtype and not check_dtype(dtype):
raise TypeError("invalid dtype: {0!r}".format(dtype))
if nodata is not None:
nodata = float(nodata)
if transform:
transform = guard_transform(transform)
# Check driver/mode blacklist.
if driver and is_blacklisted(driver, mode):
raise RasterioIOError(
"Blacklisted: file cannot be opened by "
"driver '{0}' in '{1}' mode".format(driver, mode))
# Special case for file object argument.
if mode == 'r' and hasattr(fp, 'read'):
@contextmanager
def fp_reader(fp):
memfile = MemoryFile(fp.read())
dataset = memfile.open(driver=driver, sharing=sharing)
try:
yield dataset
finally:
dataset.close()
memfile.close()
return fp_reader(fp)
elif mode in ('w', 'w+') and hasattr(fp, 'write'):
@contextmanager
def fp_writer(fp):
memfile = MemoryFile()
dataset = memfile.open(driver=driver, width=width, height=height,
count=count, crs=crs, transform=transform,
dtype=dtype, nodata=nodata, sharing=sharing, **kwargs)
try:
yield dataset
finally:
dataset.close()
memfile.seek(0)
fp.write(memfile.read())
memfile.close()
return fp_writer(fp)
else:
# If a pathlib.Path instance is given, convert it to a string path.
if isinstance(fp, Path):
fp = str(fp)
# The 'normal' filename or URL path.
path = parse_path(fp)
# Create dataset instances and pass the given env, which will
# be taken over by the dataset's context manager if it is not
# None.
if mode == 'r':
s = DatasetReader(path, driver=driver, sharing=sharing, **kwargs)
elif mode == 'r+':
s = get_writer_for_path(path)(path, mode, driver=driver, sharing=sharing, **kwargs)
elif mode.startswith("w"):
s = get_writer_for_driver(driver)(path, mode, driver=driver,
width=width, height=height,
count=count, crs=crs,
transform=transform,
dtype=dtype, nodata=nodata,
sharing=sharing,
**kwargs)
else:
raise ValueError(
"mode must be one of 'r', 'r+', or 'w', not %s" % mode)
return s
def write_geotiff(dataset,
filename,
time_index=None,
profile_override=None,
overwrite=False):
"""
Write an xarray dataset to a geotiff
Modified from datacube.helpers.write_geotiff to support:
- dask lazy arrays,
- arrays with no time dimension
- Nodata values
- Small rasters (row or cols < blocksize)
- Nodata values
- dtype checks and upcasting
- existing output checks
https://github.com/opendatacube/datacube-core/blob/develop/datacube/helpers.py
Original code licensed under the Apache License, Version 2.0 (the "License");
:param Union(str, Path) filename: Output filename
:param xarray.Dataset dataset: xarray dataset containing multiple bands to write to file
:param int time_index: time index to write to file
:param dict profile_override: option dict, overrides rasterio file creation options.
:param bool overwrite: Allow overwriting existing files.
"""
filepath = Path(filename)
if filepath.exists() and not overwrite:
raise RuntimeError('Output file exists "{}"'.format(filename))
profile_override = profile_override or {}
profile_override = lcase_dict(
profile_override) # Sanitise user modifiable values
dtype = get_dtype(dataset)
if not check_dtype(dtype): # Check for invalid dtypes
dataset, dtype = upcast(dataset, dtype)
if time_index is not None:
dataset = dataset.isel(time=time_index)
profile = lcase_dict(
GTIFF_DEFAULTS.copy()) # Sanitise user modifiable values
geobox = getattr(dataset, 'geobox', None)
if geobox is None:
raise ValueError(
'Can only write datasets with specified `crs` attribute')
height, width = geobox.shape
nodatavals = get_nodatavals(dataset)
try:
nodata = nodatavals[0]
except TypeError:
nodata = None
profile.update({
'width': width,
'height': height,
'transform': geobox.affine,
'crs': geobox.crs.crs_str,
'count': len(dataset.data_vars),
'nodata': nodata,
'dtype': str(dtype)
})
profile.update(profile_override)
# Block size must be smaller than the image size, and for geotiffs must be divisible by 16
if profile['blockxsize'] > profile['width']:
if profile['width'] < 16:
profile['tiled'] = False
else:
profile['blockxsize'] = profile['width'] // 16 * 16
if profile['blockysize'] > profile['height']:
if profile['height'] < 16:
profile['tiled'] = False
else:
profile['blockysize'] = profile['height'] // 16 * 16
if not profile.get('tiled', False):
profile.pop('blockxsize', None)
profile.pop('blockysize', None)
with rio.Env():
with rio.open(str(filename), 'w', sharing=False, **profile) as dest:
if hasattr(dataset, 'data_vars'):
for bandnum, data in enumerate(dataset.data_vars.values(),
start=1):
if hasattr(data, 'compute'):
data = data.compute()
dest.write(data.values, bandnum)
def open(fp, mode='r', driver=None, width=None, height=None, count=None,
crs=None, transform=None, dtype=None, nodata=None, **kwargs):
"""Open a dataset for reading or writing.
The dataset may be located in a local file, in a resource located
by a URL, or contained within a stream of bytes.
To access a dataset within a zip file without unzipping the archive
use an Apache VFS style zip:// URL like
zip://path/to/archive.zip!path/to/example.tif
In read ('r') or read/write ('r+') mode, no other keyword arguments
are required: the attributes are supplied by the opened dataset.
In write mode, a driver name such as "GTiff" or "JPEG" (see GDAL
docs or ``gdal_translate --help`` on the command line), ``width``
(number of pixels per line) and ``height`` (number of lines), the
``count`` number of bands in the new file must be specified.
Additionally, the data type for bands such as ``rasterio.ubyte`` for
8-bit bands or ``rasterio.uint16`` for 16-bit bands must be
specified using the ``dtype`` argument.
Parameters
----------
fp: string or file
A filename or URL, or file object opened in binary mode.
mode: string
"r" (read), "r+" (read/write), or "w" (write)
driver: string
Driver code specifying the format name (e.g. "GTiff" or
"JPEG"). See GDAL docs at
http://www.gdal.org/formats_list.html (optional, required
for writing).
width: int
Number of pixels per line (optional, required for write).
height: int
Number of lines (optional, required for write).
count: int > 0
Count of bands (optional, required for write).
dtype: rasterio.dtype
the data type for bands such as ``rasterio.ubyte`` for
8-bit bands or ``rasterio.uint16`` for 16-bit bands
(optional, required for write)
crs: dict or string
Coordinate reference system (optional, recommended for write).
transform: Affine instance
Affine transformation mapping the pixel space to geographic
space (optional, recommended for writing).
nodata: number
Defines pixel value to be interpreted as null/nodata
(optional, recommended for write, will be broadcast to all
bands).
Returns
-------
A ``DatasetReader`` or ``DatasetUpdater`` object.
Notes
-----
In write mode, you must specify at least ``width``, ``height``,
``count`` and ``dtype``.
A coordinate reference system for raster datasets in write mode
can be defined by the ``crs`` argument. It takes Proj4 style
mappings like
.. code::
{'proj': 'longlat', 'ellps': 'WGS84', 'datum': 'WGS84',
'no_defs': True}
An affine transformation that maps ``col,row`` pixel coordinates
to ``x,y`` coordinates in the coordinate reference system can be
specified using the ``transform`` argument. The value should be
an instance of ``affine.Affine``
.. code:: python
>>> from affine import Affine
>>> transform = Affine(0.5, 0.0, -180.0, 0.0, -0.5, 90.0)
These coefficients are shown in the figure below.
.. code::
| x | | a b c | | c |
| y | = | d e f | | r |
| 1 | | 0 0 1 | | 1 |
a: rate of change of X with respect to increasing column,
i.e. pixel width
b: rotation, 0 if the raster is oriented "north up"
c: X coordinate of the top left corner of the top left pixel
d: rotation, 0 if the raster is oriented "north up"
e: rate of change of Y with respect to increasing row,
usually a negative number (i.e. -1 * pixel height) if
north-up.
f: Y coordinate of the top left corner of the top left pixel
A 6-element sequence of the affine transformation matrix
coefficients in ``c, a, b, f, d, e`` order, (i.e. GDAL
geotransform order) will be accepted until 1.0 (deprecated).
A virtual filesystem can be specified. The ``vfs`` parameter may
be an Apache Commons VFS style string beginning with "zip://" or
"tar://"". In this case, the ``path`` must be an absolute path
within that container.
"""
if not isinstance(fp, string_types):
if not (hasattr(fp, 'read') or hasattr(fp, 'write')):
raise TypeError("invalid path or file: {0!r}".format(fp))
if mode and not isinstance(mode, string_types):
raise TypeError("invalid mode: {0!r}".format(mode))
if driver and not isinstance(driver, string_types):
raise TypeError("invalid driver: {0!r}".format(driver))
if dtype and not check_dtype(dtype):
raise TypeError("invalid dtype: {0!r}".format(dtype))
if nodata is not None:
nodata = float(nodata)
if 'affine' in kwargs:
# DeprecationWarning's are ignored by default
with warnings.catch_warnings():
warnings.warn(
"The 'affine' kwarg in rasterio.open() is deprecated at 1.0 "
"and only remains to ease the transition. Please switch to "
"the 'transform' kwarg. See "
"https://github.com/mapbox/rasterio/issues/86 for details.",
DeprecationWarning,
stacklevel=2)
if transform:
warnings.warn(
"Found both 'affine' and 'transform' in rasterio.open() - "
"choosing 'transform'")
transform = transform
else:
transform = kwargs.pop('affine')
if transform:
transform = guard_transform(transform)
# Check driver/mode blacklist.
if driver and is_blacklisted(driver, mode):
raise RasterioIOError(
"Blacklisted: file cannot be opened by "
"driver '{0}' in '{1}' mode".format(driver, mode))
# Special case for file object argument.
if mode == 'r' and hasattr(fp, 'read'):
@contextmanager
def fp_reader(fp):
memfile = MemoryFile(fp.read())
dataset = memfile.open()
try:
yield dataset
finally:
dataset.close()
memfile.close()
return fp_reader(fp)
elif mode == 'w' and hasattr(fp, 'write'):
@contextmanager
def fp_writer(fp):
memfile = MemoryFile()
dataset = memfile.open(driver=driver, width=width, height=height,
count=count, crs=crs, transform=transform,
dtype=dtype, nodata=nodata, **kwargs)
try:
yield dataset
finally:
dataset.close()
memfile.seek(0)
fp.write(memfile.read())
memfile.close()
return fp_writer(fp)
else:
# The 'normal' filename or URL path.
_, _, scheme = parse_path(fp)
with Env() as env:
# Get AWS credentials only if we're attempting to access a
# raster using the S3 scheme.
if scheme == 's3':
env.get_aws_credentials()
log.debug("AWS credentials have been obtained")
# Create dataset instances and pass the given env, which will
# be taken over by the dataset's context manager if it is not
# None.
if mode == 'r':
s = DatasetReader(fp)
elif mode == 'r-':
warnings.warn("'r-' mode is deprecated, use 'r'",
DeprecationWarning)
s = DatasetReader(fp)
elif mode == 'r+':
s = get_writer_for_path(fp)(fp, mode)
elif mode == 'w':
s = get_writer_for_driver(driver)(fp, mode, driver=driver,
width=width, height=height,
count=count, crs=crs,
transform=transform,
dtype=dtype, nodata=nodata,
**kwargs)
else:
raise ValueError(
"mode must be one of 'r', 'r+', or 'w', not %s" % mode)
return s
def write_geotiff(dataset,
filename,
time_index=None,
profile_override=None,
overwrite=False):
"""
Write an xarray dataset to a geotiff
Modified from datacube.helpers.write_geotiff to support:
- dask lazy arrays,
- arrays with no time dimension
- Nodata values
- Small rasters (row or cols < blocksize)
- Nodata values
- dtype checks and upcasting
- existing output checks
https://github.com/opendatacube/datacube-core/blob/develop/datacube/helpers.py
Original code licensed under the Apache License, Version 2.0 (the "License");
:param Union(str, Path) filename: Output filename
:param xarray.Dataset dataset: xarray dataset containing multiple bands to write to file
:param int time_index: time index to write to file
:param dict profile_override: option dict, overrides rasterio file creation options.
:param bool overwrite: Allow overwriting existing files.
"""
filepath = Path(filename)
if filepath.exists() and not overwrite:
raise RuntimeError('Output file exists "{}"'.format(filename))
profile_override = profile_override or {}
try:
dtypes = {val.dtype for val in dataset.data_vars.values()}
assert len(dtypes) == 1 # Check for multiple dtypes
except AttributeError:
dtypes = [dataset.dtype]
dtype = dtypes.pop()
if not check_dtype(dtype): # Check for invalid dtypes
dataset, dtype = upcast(dataset, dtype)
dimx = dataset.dims[dataset.crs.dimensions[1]]
dimy = dataset.dims[dataset.crs.dimensions[0]]
profile = GTIFF_DEFAULTS.copy()
profile.update({
'width': dimx,
'height': dimy,
'transform': dataset.affine,
'crs': dataset.crs.crs_str,
'count': len(dataset.data_vars),
'dtype': str(dtype)
})
profile.update(profile_override)
profile = lcase_dict(profile)
blockx = profile.get('blockxsize')
blocky = profile.get('blockysize')
if (blockx and blockx > dimx) or (blocky and blocky > dimy):
del profile['blockxsize']
del profile['blockysize']
profile['tiled'] = False
with rio.open(str(filename), 'w', **profile) as dest:
for bandnum, data in enumerate(dataset.data_vars.values(), start=1):
try:
nodata = data.nodata
profile.update({'nodata': nodata})
except AttributeError:
pass
if time_index is None:
data = data.data
else:
data = data.isel(time=time_index).data
if isinstance(data, dask.array.Array):
data = data.compute()
dest.write(data, bandnum)