baseclasses.py 68.7 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
# -*- coding: utf-8 -*-

import os
import warnings
from collections import OrderedDict

import numpy as np
from matplotlib import pyplot as plt
from osgeo import gdal_array
# custom
from shapely.geometry import Polygon
from shapely.wkt import loads as shply_loads
from six import PY3

# mpl_toolkits.basemap -> imported when GeoArray.show_map() is used
# dill -> imported when dumping GeoArray

try:
    from osgeo import gdal
    from osgeo import gdalnumeric
except ImportError:
    import gdal
    import gdalnumeric
from geopandas import GeoDataFrame, GeoSeries
from pandas import DataFrame
26
27
28
29
30
31
32
from py_tools_ds.convenience.object_oriented import alias_property
from py_tools_ds.geo.coord_calc import get_corner_coordinates
from py_tools_ds.geo.coord_grid import snap_bounds_to_pixGrid
from py_tools_ds.geo.coord_trafo import mapXY2imXY, imXY2mapXY, transform_any_prj, reproject_shapelyGeometry
from py_tools_ds.geo.projection import prj_equal, WKT2EPSG, EPSG2WKT
from py_tools_ds.geo.raster.conversion import raster2polygon
from py_tools_ds.geo.vector.topology \
33
    import get_footprint_polygon, polyVertices_outside_poly, fill_holes_within_poly
34
35
from py_tools_ds.geo.vector.geometry import boxObj
from py_tools_ds.io.raster.gdal import get_GDAL_ds_inmem
36
37
38
39
40
from py_tools_ds.compatibility.gdal import get_gdal_func

#  internal imports
from .subsetting import get_array_at_mapPos

41
if PY3:
42
    # noinspection PyCompatibility
43
44
45
    from builtins import TimeoutError, FileNotFoundError
else:
    from py_tools_ds.compatibility.python.exceptions import TimeoutError, FileNotFoundError
46

47
__author__ = 'Daniel Scheffler'
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72


class GeoArray(object):
    def __init__(self, path_or_array, geotransform=None, projection=None, bandnames=None, nodata=None, progress=True,
                 q=False):
        # type: (any, tuple, str, list, float, bool, bool) -> None
        """This class creates a fast Python interface for geodata - either on disk or in memory. It can be instanced with
        a file path or with a numpy array and the corresponding geoinformation. Instances can always be indexed like
        normal numpy arrays, no matter if GeoArray has been instanced from file or from an in-memory array. GeoArray
        provides a wide range of geo-related attributes belonging to the dataset as well as some functions for quickly
        visualizing the data as a map, a simple image or an interactive image.

        :param path_or_array:   a numpy.ndarray or a valid file path
        :param geotransform:    GDAL geotransform of the given array or file on disk
        :param projection:      projection of the given array or file on disk as WKT string
                                (only needed if GeoArray is instanced with an array)
        :param bandnames:       names of the bands within the input array, e.g. ['mask_1bit', 'mask_clouds'],
                                (default: ['B1', 'B2', 'B3', ...])
        :param nodata:          nodata value
        :param progress:        show progress bars (default: True)
        :param q:               quiet mode (default: False)
        """

        # TODO implement compatibility to GDAL VRTs
        if not (isinstance(path_or_array, (str, np.ndarray, GeoArray)) or
73
           issubclass(getattr(path_or_array, '__class__'), GeoArray)):
74
            raise ValueError("%s parameter 'arg' takes only string, np.ndarray or GeoArray(and subclass) instances. "
75
                             "Got %s." % (self.__class__.__name__, type(path_or_array)))
76
77

        if path_or_array is None:
78
            raise ValueError("The %s parameter 'path_or_array' must not be None!" % self.__class__.__name__)
79
80
81
82
83

        if isinstance(path_or_array, str):
            assert ' ' not in path_or_array, "The given path contains whitespaces. This is not supported by GDAL."

            if not os.path.exists(path_or_array):
84
                raise FileNotFoundError(path_or_array)
85

86
87
        if isinstance(path_or_array, GeoArray) or issubclass(getattr(path_or_array, '__class__'), GeoArray):
            self.__dict__ = path_or_array.__dict__.copy()
88
            self._initParams = dict([x for x in locals().items() if x[0] != "self"])
89
90
91
92
93
94
            self.geotransform = geotransform or self.geotransform
            self.projection = projection or self.projection
            self.bandnames = bandnames or list(self.bandnames.values())
            self._nodata = nodata if nodata is not None else self._nodata
            self.progress = False if progress is False else self.progress
            self.q = q if q is not None else self.q
95
96

        else:
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
            self._initParams = dict([x for x in locals().items() if x[0] != "self"])
            self.arg = path_or_array
            self._arr = path_or_array if isinstance(path_or_array, np.ndarray) else None
            self.filePath = path_or_array if isinstance(path_or_array, str) and path_or_array else None
            self.basename = os.path.splitext(os.path.basename(self.filePath))[0] if not self.is_inmem else 'IN_MEM'
            self.progress = progress
            self.q = q
            self._arr_cache = None  # dict containing key 'pos' and 'arr_cached'
            self._geotransform = None
            self._projection = None
            self._shape = None
            self._dtype = None
            self._nodata = nodata
            self._mask_nodata = None
            self._mask_baddata = None
112
113
            self._footprint_poly = None
            self._gdalDataset_meta_already_set = False
114
115
            self._metadata = None
            self._bandnames = None
116
117

            if bandnames:
118
                self.bandnames = bandnames  # use property in order to validate given value
119
            if geotransform:
120
                self.geotransform = geotransform  # use property in order to validate given value
121
            if projection:
122
                self.projection = projection  # use property in order to validate given value
123
124
125
126
127
128
129
130
131
132

            if self.filePath:
                self.set_gdalDataset_meta()

    @property
    def arr(self):
        return self._arr

    @arr.setter
    def arr(self, ndarray):
133
134
        assert isinstance(ndarray, np.ndarray), "'arr' can only be set to a numpy array! Got %s." % type(ndarray)
        # assert ndarray.shape == self.shape, "'arr' can only be set to a numpy array with shape %s. Received %s. " \
135
136
137
138
139
        #                                    "If you need to change the dimensions, create a new instance of %s." \
        #                                    %(self.shape, ndarray.shape, self.__class__.__name__)
        #  THIS would avoid warping like this: geoArr.arr, geoArr.gt, geoArr.prj = warp(...)

        if ndarray.shape != self.shape:
140
            self.flush_cache()  # the cached array is not useful anymore
141
142
143
144
145

        self._arr = ndarray

    @property
    def bandnames(self):
146
        if self._bandnames and len(self._bandnames) == self.bands:
147
148
149
150
151
152
153
154
155
156
157
            return self._bandnames
        else:
            self._bandnames = OrderedDict(('B%s' % band, i) for i, band in enumerate(range(1, self.bands + 1)))
            return self._bandnames

    @bandnames.setter
    def bandnames(self, list_bandnames):
        # type: (list) -> None

        if list_bandnames:
            assert isinstance(list_bandnames, list), "A list must be given when setting the 'bandnames' attribute. " \
158
                                                     "Received %s." % type(list_bandnames)
159
160
161
162
163
            assert len(list_bandnames) == self.bands, \
                'Number of given bandnames does not match number of bands in array.'
            assert len(list(set([type(b) for b in list_bandnames]))) == 1 and type(list_bandnames[0] == 'str'), \
                "'bandnames must be a set of strings. Got other datetypes in there.'"
            bN_dict = OrderedDict((band, i) for i, band in enumerate(list_bandnames))
164
165
            assert len(
                bN_dict) == self.bands, 'Bands must not have the same name. Received band list: %s' % list_bandnames
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210

            self._bandnames = bN_dict

    @property
    def is_inmem(self):
        """Check if associated image array is completely loaded into memory."""

        return isinstance(self.arr, np.ndarray)

    @property
    def shape(self):
        """Get the array shape of the associated image array."""

        if self.is_inmem:
            return self.arr.shape
        else:
            if self._shape:
                return self._shape
            else:
                self.set_gdalDataset_meta()
                return self._shape

    @property
    def ndim(self):
        """Get the number dimensions of the associated image array."""
        return len(self.shape)

    @property
    def rows(self):
        """Get the number of rows of the associated image array."""

        return self.shape[0]

    @property
    def columns(self):
        """Get the number of columns of the associated image array."""

        return self.shape[1]

    cols = alias_property('columns')

    @property
    def bands(self):
        """Get the number of bands of the associated image array."""

211
        return self.shape[2] if len(self.shape) > 2 else 1
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226

    @property
    def dtype(self):
        """Get the numpy data type of the associated image array."""

        if self._dtype:
            return self._dtype
        elif self.is_inmem:
            return self.arr.dtype
        else:
            self.set_gdalDataset_meta()
            return self._dtype

    @property
    def geotransform(self):
227
        """Get the GDAL GeoTransform of the associated image, e.g., (283500.0, 5.0, 0.0, 4464500.0, 0.0, -5.0)"""
228
229
230
231
232
233
234

        if self._geotransform:
            return self._geotransform
        elif not self.is_inmem:
            self.set_gdalDataset_meta()
            return self._geotransform
        else:
235
            return [0, 1, 0, 0, 0, -1]
236
237
238

    @geotransform.setter
    def geotransform(self, gt):
239
240
        assert isinstance(gt, (list, tuple)) and len(gt) == 6,\
            'geotransform must be a list with 6 numbers. Got %s.' % str(gt)
241

242
243
        for i in gt:
            assert isinstance(i, (int, float)), "geotransform must contain only numbers. Got '%s'." % i
244

245
        self._geotransform = gt
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267

    gt = alias_property('geotransform')

    @property
    def xgsd(self):
        """Get the X resolution in units of the given or detected projection."""

        return self.geotransform[1]

    @property
    def ygsd(self):
        """Get the Y resolution in units of the given or detected projection."""

        return abs(self.geotransform[5])

    @property
    def xygrid_specs(self):
        """
        Get the specifications for the X/Y coordinate grid, e.g. [[15,30], [0,30]] for a coordinate with its origin
        at X/Y[15,0] and a GSD of X/Y[15,30].
        """

268
        def get_grid(gt, xgsd, ygsd): return [[gt[0], gt[0] + xgsd], [gt[3], gt[3] - ygsd]]
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
        return get_grid(self.geotransform, self.xgsd, self.ygsd)

    @property
    def projection(self):
        """
        Get the projection of the associated image. Setting the projection is only allowed if GeoArray has been
        instanced from memory or the associated file on disk has no projection.
        """

        if self._projection:
            return self._projection
        elif not self.is_inmem:
            self.set_gdalDataset_meta()
            return self._projection
        else:
            return ''

    @projection.setter
    def projection(self, prj):
        if self.filePath:
289
            assert self.projection is None or prj_equal(self.projection, prj), \
290
                "Cannot set %s.projection to the given value because it does not match the projection from the file " \
291
                "on disk." % self.__class__.__name__
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
        else:
            self._projection = prj

    prj = alias_property('projection')

    @property
    def epsg(self):
        """Get the EPSG code of the projection of the GeoArray."""

        return WKT2EPSG(self.projection)

    @epsg.setter
    def epsg(self, epsg_code):
        self.projection = EPSG2WKT(epsg_code)

    @property
    def box(self):
        mapPoly = get_footprint_polygon(get_corner_coordinates(gt=self.geotransform, cols=self.columns, rows=self.rows))
        return boxObj(gt=self.geotransform, prj=self.projection, mapPoly=mapPoly)

    @property
    def nodata(self):
        """
        Get the nodata value of the GeoArray. If GeoArray has been instanced with a file path the file is checked
        for an existing nodata value. Otherwise (if no value is exlicitly given during object instanciation) the nodata
        value is tried to be automatically detected.
        """

        if self._nodata is not None:
            return self._nodata
        else:
            # try to get nodata value from file
            if not self.is_inmem:
                self.set_gdalDataset_meta()
            if self._nodata is None:
Daniel Scheffler's avatar
Bugfix    
Daniel Scheffler committed
327
                self._nodata = self.find_noDataVal()
328
329
330
331
332
333
                if self._nodata == 'ambiguous':
                    warnings.warn('Nodata value could not be clearly identified. It has been set to None.')
                    self._nodata = None
                else:
                    if self._nodata is not None and not self.q:
                        print("Automatically detected nodata value for %s '%s': %s"
334
                              % (self.__class__.__name__, self.basename, self._nodata))
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
            return self._nodata

    @nodata.setter
    def nodata(self, value):
        self._nodata = value

    @property
    def mask_nodata(self):
        """
        Get the nodata mask of the associated image array. It is calculated using all image bands.
        """

        if self._mask_nodata is not None:
            return self._mask_nodata
        else:
350
            self.calc_mask_nodata()  # sets self._mask_nodata
351
352
353
354
355
356
357
358
359
360
361
            return self._mask_nodata

    @mask_nodata.setter
    def mask_nodata(self, mask):
        """Set bad data mask.

        :param mask:    Can be a file path, a numpy array or an instance o GeoArray.
        """

        if mask is not None:
            from .masks import NoDataMask
362
363
            geoArr_mask = NoDataMask(mask, progress=self.progress, q=self.q)
            geoArr_mask.gt = geoArr_mask.gt if geoArr_mask.gt not in [None, [0, 1, 0, 0, 0, -1]] else self.gt
364
            geoArr_mask.prj = geoArr_mask.prj if geoArr_mask.prj else self.prj
365
            imName = "the %s '%s'" % (self.__class__.__name__, self.basename)
366
367
368
369

            assert geoArr_mask.bands == 1, \
                'Expected one single band as nodata mask for %s. Got %s bands.' % (self.basename, geoArr_mask.bands)
            assert geoArr_mask.shape[:2] == self.shape[:2], 'The provided nodata mask must have the same number of ' \
370
                                                            'rows and columns as the %s itself.' % imName
371
372
            assert geoArr_mask.gt == self.gt, \
                'The geotransform of the given nodata mask for %s must match the geotransform of the %s itself. ' \
373
                'Got %s.' % (imName, self.__class__.__name__, geoArr_mask.gt)
374
375
            assert not geoArr_mask.prj or prj_equal(geoArr_mask.prj, self.prj), \
                'The projection of the given nodata mask for the %s must match the projection of the %s itself.' \
376
                % (imName, self.__class__.__name__)
377
378

            self._mask_nodata = geoArr_mask
379
380
381
382
383
384
        else:
            del self.mask_nodata

    @mask_nodata.deleter
    def mask_nodata(self):
        self._mask_nodata = None
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403

    @property
    def mask_baddata(self):
        """
        Returns the bad data mask for the associated image array if it has been explicitly previously. It can be set
         by passing a file path, a numpy array or an instance of GeoArray to the setter of this property.
        """

        return self._mask_baddata

    @mask_baddata.setter
    def mask_baddata(self, mask):
        """Set bad data mask.

        :param mask:    Can be a file path, a numpy array or an instance o GeoArray.
        """

        if mask is not None:
            from .masks import BadDataMask
404
405
            geoArr_mask = BadDataMask(mask, progress=self.progress, q=self.q)
            geoArr_mask.gt = geoArr_mask.gt if geoArr_mask.gt not in [None, [0, 1, 0, 0, 0, -1]] else self.gt
406
            geoArr_mask.prj = geoArr_mask.prj if geoArr_mask.prj else self.prj
407
            imName = "the %s '%s'" % (self.__class__.__name__, self.basename)
408
409
410
411

            assert geoArr_mask.bands == 1, \
                'Expected one single band as bad data mask for %s. Got %s bands.' % (self.basename, geoArr_mask.bands)
            assert geoArr_mask.shape[:2] == self.shape[:2], 'The provided bad data mask must have the same number of ' \
412
                                                            'rows and columns as the %s itself.' % imName
413
414
            assert geoArr_mask.gt == self.gt, \
                'The geotransform of the given bad data mask for %s must match the geotransform of the %s itself. ' \
415
                'Got %s.' % (imName, self.__class__.__name__, geoArr_mask.gt)
416
417
            assert prj_equal(geoArr_mask.prj, self.prj), \
                'The projection of the given bad data mask for the %s must match the projection of the %s itself.' \
418
                % (imName, self.__class__.__name__)
419
420

            self._mask_baddata = geoArr_mask
421
422
423
424
425
426
        else:
            del self.mask_baddata

    @mask_baddata.deleter
    def mask_baddata(self):
        self._mask_baddata = None
427
428
429
430
431
432
433
434
435
436

    @property
    def footprint_poly(self):
        # FIXME should return polygon in image coordinates if no projection is available
        """
        Get the footprint polygon of the associated image array (returns an instance of shapely.geometry.Polygon.
        """

        if self._footprint_poly is None:
            assert self.mask_nodata is not None, 'A nodata mask is needed for calculating the footprint polygon. '
437
            if np.std(self.mask_nodata[:]) == 0:
438
439
440
441
                # do not run raster2polygon if whole image is filled with data
                self._footprint_poly = self.box.mapPoly
            else:
                try:
442
443
                    multipolygon = raster2polygon(self.mask_nodata.astype(np.uint8), self.gt, self.prj, exact=False,
                                                  progress=self.progress, q=self.q, maxfeatCount=10, timeout=3)
444
                    self._footprint_poly = fill_holes_within_poly(multipolygon)
445
                except (RuntimeError, TimeoutError):
446
447
448
449
                    if not self.q:
                        warnings.warn("\nCalculation of footprint polygon failed for %s '%s'. Using outer bounds. One "
                                      "reason could be that the nodata value appears within the actual image (not only "
                                      "as fill value). To avoid this use another nodata value. Current nodata value is "
450
                                      "%s." % (self.__class__.__name__, self.basename, self.nodata))
451
452
453
                    self._footprint_poly = self.box.mapPoly

            # validation
454
455
456
457
            assert not polyVertices_outside_poly(self._footprint_poly, self.box.mapPoly), \
                "Computing footprint polygon for %s '%s' failed. The resulting polygon is partly or completely " \
                "outside of the image bounds." % (self.__class__.__name__, self.basename)
            # assert self._footprint_poly
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
            # for XY in self.corner_coord:
            #    assert self.GeoArray.box.mapPoly.contains(Point(XY)) or self.GeoArray.box.mapPoly.touches(Point(XY)), \
            #        "The corner position '%s' is outside of the %s." % (XY, self.imName)

        return self._footprint_poly

    @footprint_poly.setter
    def footprint_poly(self, poly):
        if isinstance(poly, Polygon):
            self._footprint_poly = poly
        elif isinstance(poly, str):
            self._footprint_poly = shply_loads(poly)
        else:
            raise ValueError("'footprint_poly' can only be set from a shapely polygon or a WKT string.")

    @property
    def metadata(self):
        """
        Returns a GeoDataFrame containing all available metadata (read from file if available).
        Use 'metadata[band_index].to_dict()' to get a metadata dictionary for a specific band.
        Use 'metadata.loc[row_name].to_dict()' to get all metadata values of the same key for all bands as dictionary.
        Use 'metadata.loc[row_name, band_index] = value' to set a new value.

        :return:  geopandas.GeoDataFrame
        """

        if self._metadata is not None:
            return self._metadata
        else:
            default = GeoDataFrame(columns=range(self.bands))
488
            # for bn,idx in self.bandnames.items():
489
490
491
492
493
494
495
496
497
498
            #    default.loc['band_index',bn] = idx
            self._metadata = default
            if not self.is_inmem:
                self.set_gdalDataset_meta()
                return self._metadata
            else:
                return self._metadata

    @metadata.setter
    def metadata(self, GDF):
499
        assert isinstance(GDF, (GeoDataFrame, DataFrame)) and len(GDF.columns) == self.bands, \
500
            "%s.metadata can only be set with an instance of geopandas.GeoDataFrame of which the column number " \
501
            "corresponds to the band number of %s." % (self.__class__.__name__, self.__class__.__name__)
502
503
504
505
506
        self._metadata = GDF

    meta = alias_property('metadata')

    def __getitem__(self, given):
507
        if isinstance(given, (int, float, slice)) and self.ndim == 3:
508
509
510
511
512
513
514
515
516
517
518
519
520
            # handle 'given' as index for 3rd (bands) dimension
            if self.is_inmem:
                return self.arr[:, :, given]
            else:
                return self.from_path(self.arg, [given])

        elif isinstance(given, str):
            # behave like a dictionary and return the corresponding band
            if self.bandnames:
                if given not in self.bandnames:
                    raise ValueError("'%s' is not a known band. Known bands are: %s"
                                     % (given, ', '.join(list(self.bandnames.keys()))))
                if self.is_inmem:
521
                    return self.arr if self.ndim == 2 else self.arr[:, :, self.bandnames[given]]
522
523
524
525
                else:
                    return self.from_path(self.arg, [self.bandnames[given]])
            else:
                raise ValueError('String indices are only supported if %s has been instanced with bandnames given.'
526
                                 % self.__class__.__name__)
527
528
529
530
531
532
533

        elif isinstance(given, (tuple, list)):
            # handle requests like geoArr[[1,2],[3,4]  -> not implemented in from_path if array is not in mem
            types = [type(i) for i in given]
            if list in types or tuple in types:
                self.to_mem()

534
            if len(given) == 3:
535
536

                # handle strings in the 3rd dim of 'given' -> convert them to a band index
537
                if isinstance(given[2], str):
538
539
540
541
542
543
544
545
                    if self.bandnames:
                        if given[2] not in self.bandnames:
                            raise ValueError("'%s' is not a known band. Known bands are: %s"
                                             % (given[2], ', '.join(list(self.bandnames.keys()))))

                        band_idx = self.bandnames[given[2]]
                        # NOTE: the string in the 3rd is ignored if ndim==2 and band_idx==0
                        if self.is_inmem:
546
                            return self.arr if (self.ndim == 2 and band_idx == 0) else self.arr[:, :, band_idx]
547
                        else:
548
549
                            getitem_params = \
                                given[:2] if (self.ndim == 2 and band_idx == 0) else given[:2] + (band_idx,)
550
551
552
553
554
555
556
                            return self.from_path(self.arg, getitem_params)
                    else:
                        raise ValueError(
                            'String indices are only supported if %s has been instanced with bandnames given.'
                            % self.__class__.__name__)

                # in case a third dim is requested from 2D-array -> ignore 3rd dim if 3rd dim is 0
557
                elif self.ndim == 2 and given[2] == 0:
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
                    if self.is_inmem:
                        return self.arr[given[:2]]
                    else:
                        return self.from_path(self.arg, given[:2])

        # if nothing has been returned until here -> behave like a numpy array
        if self.is_inmem:
            return self.arr[given]
        else:
            getitem_params = [given] if isinstance(given, slice) else given
            return self.from_path(self.arg, getitem_params)

    def __setitem__(self, idx, array2set):
        """Overwrites the pixel values of GeoArray.arr with the given array.

        :param idx:         <int, list, slice> the index position to overwrite
        :param array2set:   <np.ndarray> array to be set. Must be compatible to the given index position.
        :return:
        """

        if self.is_inmem:
            self.arr[idx] = array2set
        else:
            raise NotImplementedError('Item assignment for %s instances that are not in memory is not yet supported.'
582
                                      % self.__class__.__name__)
583
584
585

    def __getattr__(self, attr):
        # check if the requested attribute can not be present because GeoArray has been instanced with an array
586
587
        if attr not in self.__dir__() and not self.is_inmem and attr in ['shape', 'dtype', 'geotransform',
                                                                         'projection']:
588
589
            self.set_gdalDataset_meta()

590
591
592
        if attr in self.__dir__():  # __dir__() includes also methods and properties
            return self.__getattribute__(attr)  # __getattribute__ avoids infinite loop
        elif hasattr(np.array([]), attr):
593
594
            return self[:].__getattribute__(attr)
        else:
595
            raise AttributeError("%s object has no attribute '%s'." % (self.__class__.__name__, attr))
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622

    def __getstate__(self):
        """Defines how the attributes of GMS object are pickled."""

        # clean array cache in order to avoid cache pickling
        self.flush_cache()

        return self.__dict__

    def __setstate__(self, state):
        """Defines how the attributes of GMS object are unpickled.
        NOTE: This method has been implemented because otherwise pickled and unpickled instances show recursion errors
        within __getattr__ when requesting any attribute.
        """

        self.__dict__ = state

    def calc_mask_nodata(self, fromBand=None, overwrite=False):
        """Calculates a no data mask with (values: 0=nodata; 1=data)

        :param fromBand:  <int> index of the band to be used (if None, all bands are used)
        :param overwrite: <bool> whether to overwrite existing nodata mask that has already been calculated
        :return:
        """

        if self._mask_nodata is None or overwrite:
            assert self.ndim in [2, 3], "Only 2D or 3D arrays are supported. Got a %sD array." % self.ndim
623
            arr = self[:, :, fromBand] if self.ndim == 3 and fromBand is not None else self[:]
624
625
626
627
628

            if self.nodata is None:
                self.mask_nodata = np.ones((self.rows, self.cols), np.bool)
            else:
                self.mask_nodata = np.where(arr == self.nodata, 0, 1).astype(np.bool) if arr.ndim == 2 else \
629
                    np.all(np.where(arr == self.nodata, 0, 1), axis=2).astype(np.bool)
630

631
632
633
634
635
636
637
638
    def find_noDataVal(self, bandIdx=0, sz=3):
        """Tries to derive no data value from homogenious corner pixels within 3x3 windows (by default).
        :param bandIdx:
        :param sz: window size in which corner pixels are analysed
        """
        wins = [self[0:sz, 0:sz, bandIdx], self[0:sz, -sz:, bandIdx],
                self[-sz:, -sz:, bandIdx], self[-sz:, 0:sz, bandIdx]]  # UL, UR, LR, LL

639
640
        means, stds = [np.mean(win) for win in wins], [np.std(win) for win in wins]
        possVals = [mean for mean, std in zip(means, stds) if std == 0 or np.isnan(std)]
641
642
643
644
        # possVals==[]: all corners are filled with data; np.std(possVals)==0: noDataVal clearly identified

        if possVals:
            if np.std(possVals) != 0:
645
646
647
648
649
650
                if np.isnan(np.std(possVals)):
                    # at least one of the possible values is np.nan
                    nodata = np.nan
                else:
                    # different possible nodata values have been found in the image corner
                    nodata = 'ambiguous'
651
652
653
654
655
            else:
                if len(possVals) <= 2:
                    # each window in each corner
                    warnings.warn("\nAutomatic nodata value detection returned the value %s for GeoArray '%s' but this "
                                  "seems to be unreliable (occurs in only %s). To avoid automatic detection, just pass "
656
657
658
                                  "the correct nodata value."
                                  % (possVals[0], self.basename, ('2 image corners' if len(possVals) == 2 else
                                                                  '1 image corner')))
659
                nodata = possVals[0]
660
        else:
661
662
663
            nodata = None

        return nodata
664

665
666
667
668
669
670
671
672
673
674
    def set_gdalDataset_meta(self):
        """Retrieves GDAL metadata from file. This function is only executed once to avoid overwriting of user defined
         attributes, that are defined after object instanciation.

        :return:
        """

        if not self._gdalDataset_meta_already_set:
            assert self.filePath
            ds = gdal.Open(self.filePath)
675
676
677
            if not ds:
                raise Exception('Error reading file:  ' + gdal.GetLastErrorMsg())

678
            # set private class variables (in order to avoid recursion error)
679
680
            self._shape = tuple([ds.RasterYSize, ds.RasterXSize] + ([ds.RasterCount] if ds.RasterCount > 1 else []))
            self._dtype = gdal_array.GDALTypeCodeToNumericTypeCode(ds.GetRasterBand(1).DataType)
681
            self._geotransform = ds.GetGeoTransform()
682
683
684
            # temp conversion to EPSG needed because GDAL seems to modify WKT string when writing file to disk
            # (e.g. using gdal_merge) -> conversion to EPSG and back undos that
            self._projection = EPSG2WKT(WKT2EPSG(ds.GetProjection()))
685

686
687
688
689
            if 'nodata' not in self._initParams or self._initParams['nodata'] is None:
                band = ds.GetRasterBand(1)
                # FIXME this does not support different nodata values within the same file
                self._nodata = band.GetNoDataValue()
690

691
            # read global domain metadata
692
693
            # TODO check to specifically use the 'ENVI' metadata domain ds.GetMetadata('ENVI')
            global_meta = ds.GetMetadata()
694

695
696
            # read band domain metadata
            for b in range(self.bands):
697
698
                band = ds.GetRasterBand(b + 1)
                meta_gs = GeoSeries(band.GetMetadata())
699

700
701
702
703
704
705
706
                # add band names if available
                if 'Band_%s' % str(b + 1) in global_meta.keys():
                    meta_gs['band_name'] = global_meta['Band_%s' % str(b + 1)]

                # TODO add the remaining global metadata

                self.metadata[b] = meta_gs
707

708
            del ds, band
709
710
711
712
713
714
715
716
717
718
719
720
721
722

        self._gdalDataset_meta_already_set = True

    def from_path(self, path, getitem_params=None):
        # type: (str, list) -> np.ndarray
        """Read a GDAL compatible raster image from disk, with respect to the given image position.
        NOTE: If the requested array position is already in cache, it is returned from there.

        :param path:            <str> the file path of the image to read
        :param getitem_params:  <list> a list of slices in the form [row_slice, col_slice, band_slice]
        :return out_arr:        <np.ndarray> the output array
        """

        ds = gdal.Open(path)
723
724
725
        if not ds:
            raise Exception('Error reading file:  ' + gdal.GetLastErrorMsg())

726
        R, C, B = ds.RasterYSize, ds.RasterXSize, ds.RasterCount
727
        del ds
728

729
        # convert getitem_params to subset area to be read ##
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
        rS, rE, cS, cE, bS, bE, bL = [None] * 7

        # populate rS, rE, cS, cE, bS, bE, bL
        if getitem_params:
            if len(getitem_params) >= 2:
                givenR, givenC = getitem_params[:2]
                if isinstance(givenR, slice):
                    rS = givenR.start
                    rE = givenR.stop - 1 if givenR.stop is not None else None
                elif isinstance(givenR, int):
                    rS = givenR
                    rE = givenR
                if isinstance(givenC, slice):
                    cS = givenC.start
                    cE = givenC.stop - 1 if givenC.stop is not None else None
                elif isinstance(givenC, int):
                    cS = givenC
                    cE = givenC
            if len(getitem_params) in [1, 3]:
                givenB = getitem_params[2] if len(getitem_params) == 3 else getitem_params[0]
                if isinstance(givenB, slice):
                    bS = givenB.start
                    bE = givenB.stop - 1 if givenB.stop is not None else None
                elif isinstance(givenB, int):
                    bS = givenB
                    bE = givenB
756
                elif isinstance(givenB, (tuple, list)):
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
                    typesInGivenB = [type(i) for i in givenB]
                    assert len(list(set(typesInGivenB))) == 1, \
                        'Mixed data types within the list of bands are not supported.'
                    if isinstance(givenB[0], int):
                        bL = list(givenB)
                    elif isinstance(givenB[0], str):
                        bL = [self.bandnames[i] for i in givenB]
                elif type(givenB) in [str]:
                    bL = [self.bandnames[givenB]]

        # set defaults for not given values
        rS = rS if rS is not None else 0
        rE = rE if rE is not None else R - 1
        cS = cS if cS is not None else 0
        cE = cE if cE is not None else C - 1
        bS = bS if bS is not None else 0
        bE = bE if bE is not None else B - 1
        bL = list(range(bS, bE + 1)) if not bL else bL

        # convert negative to positive ones
        rS = rS if rS >= 0 else self.rows + rS
        rE = rE if rE >= 0 else self.rows + rE
        cS = cS if cS >= 0 else self.columns + cS
        cE = cE if cE >= 0 else self.columns + cE
        bS = bS if bS >= 0 else self.bands + bS
        bE = bE if bE >= 0 else self.bands + bE
783
        bL = [b if b >= 0 else (self.bands + b) for b in bL]
784
785

        # validate subset area bounds to be read
786
787
788
789
790
791
792
        def msg(v, idx, sz):
            # FIXME numpy raises that error ONLY for the 2nd axis
            return '%s is out of bounds for axis %s with size %s' % (v, idx, sz)

        for val, axIdx, axSize in zip([rS, rE, cS, cE, bS, bE], [0, 0, 1, 1, 2, 2], [R, R, C, C, B, B]):
            if not 0 <= val <= axSize - 1:
                raise ValueError(msg(val, axIdx, axSize))
793
794

        # summarize requested array position in arr_pos
795
        # NOTE: # bandlist must be string because truth value of an array with more than one element is ambiguous
796
797
798
        arr_pos = dict(rS=rS, rE=rE, cS=cS, cE=cE, bS=bS, bE=bE, bL=bL)

        # check if the requested array position is already in cache -> if yes, return it from there
799
        if self._arr_cache is not None and self._arr_cache['pos'] == arr_pos:
800
            out_arr = self._arr_cache['arr_cached']
801
802
803
804
805
806
807

        else:
            # TODO insert a multiprocessing.Lock here in order to prevent IO bottlenecks?
            # read subset area from disk
            if bL == list(range(0, B)):
                tempArr = gdalnumeric.LoadFile(path, cS, rS, cE - cS + 1, rE - rS + 1)
                out_arr = np.swapaxes(np.swapaxes(tempArr, 0, 2), 0, 1) if B > 1 else tempArr
808
809
                if out_arr is None:
                    raise Exception('Error reading file:  ' + gdal.GetLastErrorMsg())
810
811
812
813
            else:
                ds = gdal.Open(path)
                if len(bL) == 1:
                    band = ds.GetRasterBand(bL[0] + 1)
814
                    out_arr = band.ReadAsArray(cS, rS, cE - cS + 1, rE - rS + 1)
815
816
                    if out_arr is None:
                        raise Exception('Error reading file:  ' + gdal.GetLastErrorMsg())
817
                    del band
818
819
820
821
822
                else:
                    out_arr = np.empty((rE - rS + 1, cE - cS + 1, len(bL)))
                    for i, bIdx in enumerate(bL):
                        band = ds.GetRasterBand(bIdx + 1)
                        out_arr[:, :, i] = band.ReadAsArray(cS, rS, cE - cS + 1, rE - rS + 1)
823
824
                        if out_arr is None:
                            raise Exception('Error reading file:  ' + gdal.GetLastErrorMsg())
825
                        del band
826

827
                del ds
828
829

            # only set self.arr if the whole cube has been read (in order to avoid sudden shape changes)
830
            if out_arr.shape == self.shape:
831
832
833
834
835
                self.arr = out_arr

            # write _arr_cache
            self._arr_cache = dict(pos=arr_pos, arr_cached=out_arr)

836
837
        return out_arr  # TODO implement check of returned datatype (e.g. NoDataMask should always return np.bool
        # TODO -> would be np.int8 if an int8 file is read from disk
838
839
840
841
842
843
844
845

    def save(self, out_path, fmt='ENVI', creationOptions=None):
        # type: (str, str, list) -> None
        """Write the raster data to disk.

        :param out_path:        <str> output path
        :param fmt:             <str> the output format / GDAL driver code to be used for output creation, e.g. 'ENVI'
                                Refer to http://www.gdal.org/formats_list.html to get a full list of supported formats.
846
847
        :param creationOptions: <list> GDAL creation options,
                                e.g., ["QUALITY=80", "REVERSIBLE=YES", "WRITE_METADATA=YES"]
848
849
850
        """

        if not self.q:
851
852
            print('Writing GeoArray of size %s to %s.' % (self.shape, out_path))
        assert self.ndim in [2, 3], 'Only 2D- or 3D arrays are supported.'
853
854
855
856
857
858
859
860
861
862

        driver = gdal.GetDriverByName(fmt)
        if driver is None:
            raise Exception("'%s' is not a supported GDAL driver. Refer to www.gdal.org/formats_list.html for full "
                            "list of GDAL driver codes." % fmt)

        if not os.path.isdir(os.path.dirname(out_path)):
            os.makedirs(os.path.dirname(out_path))

        if self.is_inmem:
863
864
            ds = get_GDAL_ds_inmem(self.arr, self.geotransform, self.projection,
                                   self.nodata)  # expects rows,columns,bands
865
866
867

            # set metadata
            if not self.metadata.empty:
868
869
870
                global_meta = {}

                # set band domain metadata
871
                for bidx in range(self.bands):
872
                    band = ds.GetRasterBand(bidx + 1)
873
                    meta2write = self.metadata[bidx].to_dict()
874
                    meta2write = dict((k, v) for k, v in meta2write.items() if v is not np.nan)
875
876

                    if 'band_name' in meta2write:
877
                        global_meta['Band_%s' % str(bidx + 1)] = meta2write['band_name']
878
879
                        del meta2write['band_name']

880
                    band.SetMetadata(meta2write)
881
                    del band
882

883
884
885
                # set global domain metadata
                ds.SetMetadata(global_meta)

886
887
888
889
890
                # get ENVI metadata domain
                # ds_orig = gdal.Open(self.filePath)
                # envi_meta_domain = ds_orig.GetMetadata('ENVI')
                # ds.SetMetadata(envi_meta_domain, 'ENVI')
                # ds_orig = None
891

892
893
            driver.CreateCopy(out_path, ds, options=creationOptions if creationOptions else [])

894
895
896
897
            # rows, columns, bands => bands, rows, columns
            # out_arr = self.arr if self.ndim == 2 else np.swapaxes(np.swapaxes(self.arr, 0, 2), 1, 2)
            # gdalnumeric.SaveArray(out_arr, out_path, format=fmt, prototype=ds) # expects bands,rows,columns
            del ds
898
899
900

        else:
            src_ds = gdal.Open(self.filePath)
901
902
903
            if not src_ds:
                raise Exception('Error reading file:  ' + gdal.GetLastErrorMsg())

904
905
            gdal_Translate = get_gdal_func('Translate')
            gdal_Translate(out_path, src_ds, format=fmt, creationOptions=creationOptions)
906
            del src_ds
907
908
909
910
911
912
913
914
915

        if not os.path.exists(out_path):
            raise Exception(gdal.GetLastErrorMsg())

    def dump(self, out_path):
        # type: (str) -> None
        """Serialize the whole object instance to disk using dill."""

        import dill
916
917
        with open(out_path, 'wb') as outF:
            dill.dump(self, outF)
918
919
920
921
922
923
924
925
926

    def _get_plottable_image(self, xlim=None, ylim=None, band=None, boundsMap=None, boundsMapPrj=None, res_factor=None,
                             nodataVal=None, out_prj=None):
        # handle limits
        if boundsMap:
            boundsMapPrj = boundsMapPrj if boundsMapPrj else self.prj
            image2plot, gt, prj = self.get_mapPos(boundsMap, boundsMapPrj, band2get=band,
                                                  fillVal=nodataVal if nodataVal is not None else self.nodata)
        else:
927
928
            cS, cE = xlim if isinstance(xlim, (tuple, list)) else (0, self.columns)
            rS, rE = ylim if isinstance(ylim, (tuple, list)) else (0, self.rows)
929
930

            image2plot = self[rS:rE, cS:cE, band] if band is not None else self[rS:rE, cS:cE]
931
            gt, prj = self.geotransform, self.projection
932

933
        transOpt = ['SRC_METHOD=NO_GEOTRANSFORM'] if tuple(gt) == (0, 1, 0, 0, 0, -1) else None
934
        xdim, ydim = None, None
935
        nodataVal = nodataVal if nodataVal is not None else self.nodata
936
937

        if res_factor != 1. and image2plot.shape[0] * image2plot.shape[1] > 1e6:  # shape > 1000*1000
938
939
940
941
            # sample image down / normalize
            xdim, ydim = \
                (self.columns * res_factor, self.rows * res_factor) if res_factor else \
                tuple((np.array([self.columns, self.rows]) / (np.array([self.columns, self.rows]).max() / 1000)))
942
943
944
            xdim, ydim = int(xdim), int(ydim)

        if xdim or ydim or out_prj:
945
            from py_tools_ds.geo.raster.reproject import warp_ndarray
946
947
948
949
950
            image2plot, gt, prj = warp_ndarray(image2plot, self.geotransform, self.projection,
                                               out_XYdims=(xdim, ydim), in_nodata=nodataVal, out_nodata=nodataVal,
                                               transformerOptions=transOpt, out_prj=out_prj, q=True)
            if transOpt and 'NO_GEOTRANSFORM' in ','.join(transOpt):
                image2plot = np.flipud(image2plot)
951
952
                gt = list(gt)
                gt[3] = 0
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974

            if xdim or ydim:
                print('Note: array has been downsampled to %s x %s for faster visualization.' % (xdim, ydim))

        return image2plot, gt, prj

    def show(self, xlim=None, ylim=None, band=None, boundsMap=None, boundsMapPrj=None, figsize=None,
             interpolation='none', vmin=None, vmax=None, cmap=None, nodataVal=None, res_factor=None, interactive=False):
        """Plots the desired array position into a figure.

        :param xlim:            [start_column, end_column]
        :param ylim:            [start_row, end_row]
        :param band:            the band index of the band to be plotted (if None and interactive==True all bands are
                                shown, otherwise the first band is chosen)
        :param boundsMap:       xmin, ymin, xmax, ymax
        :param boundsMapPrj:
        :param figsize:
        :param interpolation:
        :param vmin:
        :param vmax:
        :param cmap:
        :param nodataVal:
Daniel Scheffler's avatar
Daniel Scheffler committed
975
976
        :param res_factor:      <float> resolution factor for downsampling of the image to be plotted in order to save
                                plotting time and memory (default=None -> downsampling is performed to 1000x1000)
977
978
979
980
981
982
983
984
        :param interactive:     <bool> activates interactive plotting based on 'holoviews' library.
                                NOTE: this deactivates the magic '% matplotlib inline' in Jupyter Notebook
        :return:
        """

        band = (band if band is not None else 0) if not interactive else band

        # get image to plot
985
        nodataVal = nodataVal if nodataVal is not None else self.nodata
Daniel Scheffler's avatar
Daniel Scheffler committed
986
987
988
        image2plot, gt, prj = \
            self._get_plottable_image(xlim, ylim, band, boundsMap=boundsMap, boundsMapPrj=boundsMapPrj,
                                      res_factor=res_factor, nodataVal=nodataVal)
989
990

        # set color palette
991
992
        palette = cmap if cmap else plt.cm.gray
        if nodataVal is not None and np.std(image2plot) != 0:  # do not show nodata
993
            image2plot = np.ma.masked_equal(image2plot, nodataVal)
994
            vmin_auto, vmax_auto = np.percentile(image2plot.compressed(), 2), np.percentile(image2plot.compressed(), 98)
995
996
997
998
999
1000
1001
            palette.set_bad('aqua', 0)
        else:
            vmin_auto, vmax_auto = np.percentile(image2plot, 2), np.percentile(image2plot, 98)

        vmin = vmin if vmin is not None else vmin_auto
        vmax = vmax if vmax is not None else vmax_auto

1002
        palette.set_over('1')
1003
1004
        palette.set_under('0')

1005
        if interactive and image2plot.ndim == 3:
1006
1007
1008
1009
1010
1011
1012
            import holoviews as hv
            from skimage.exposure import rescale_intensity
            hv.notebook_extension('matplotlib')

            cS, cE = xlim if isinstance(xlim, (tuple, list)) else (0, self.columns - 1)
            rS, rE = ylim if isinstance(ylim, (tuple, list)) else (0, self.rows - 1)

1013
            image2plot = np.array(rescale_intensity(image2plot, in_range=(vmin, vmax)))
1014

1015
1016
1017
1018
1019
1020
1021
            def get_hv_image(b):
                # FIXME ylabels have the wrong order
                return hv.Image(image2plot[:, :, b] if b is not None else image2plot,
                                bounds=(cS, rS, cE, rE))(
                    style={'cmap': 'gray'}, plot={'fig_inches': 4 if figsize is None else figsize, 'show_grid': True})

            # hvIm = hv.Image(image2plot)(style={'cmap': 'gray'}, figure_inches=figsize)
1022
1023
1024
1025
1026
1027
1028
            hmap = hv.HoloMap([(band, get_hv_image(band)) for band in range(image2plot.shape[2])], kdims=['band'])

            return hmap

        else:
            if interactive:
                warnings.warn('Currently there is no interactive mode for single-band arrays. '
1029
                              'Switching to standard matplotlib figure..')  # TODO implement zoomable fig
1030
1031
1032
1033
1034

            # show image
            plt.figure(figsize=figsize)
            rows, cols = image2plot.shape[:2]
            plt.imshow(image2plot, palette, interpolation=interpolation, extent=(0, cols, rows, 0),
1035
                       vmin=vmin, vmax=vmax, )  # compressed excludes nodata values
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
            plt.show()

    def show_map(self, xlim=None, ylim=None, band=0, boundsMap=None, boundsMapPrj=None, ax=None, figsize=None,
                 interpolation='none', vmin=None, vmax=None, cmap=None, nodataVal=None, res_factor=None,
                 return_map=False, zoomable=False):
        """

        :param xlim:
        :param ylim:
        :param band:            band index (starting with 0)
        :param boundsMap:       xmin, ymin, xmax, ymax
        :param boundsMapPrj:
        :param ax:              allows to pass a matplotlib axis object where figure is plotted into
        :param figsize:
        :param interpolation:
        :param vmin:
        :param vmax:
        :param cmap:
        :param nodataVal:
Daniel Scheffler's avatar
Daniel Scheffler committed
1055
1056
        :param res_factor:      <float> resolution factor for downsampling of the image to be plotted in order to save
                                plotting time and memory (default=None -> downsampling is performed to 1000x1000)
1057
1058
1059
1060
1061
        :param return_map:
        :param zoomable:        <bool> enable or disable zooming via mpld3
        :return:
        """

1062
1063
1064
        try:
            from mpl_toolkits.basemap import Basemap
        except ImportError:
1065
1066
            warnings.warn('This function requires Basemap. You need to install basemap manually (see www./'
                          'matplotlib.org/basemap) if you want to plot maps. It is not automatically installed.')
1067
            raise
1068
1069
1070
1071
1072
1073
        try:
            import mpld3
            if zoomable:
                mpld3.enable_notebook()
            else:
                mpld3.disable_notebook()
1074
        except Exception:
1075
1076
1077
1078
            if zoomable:
                zoomable = False
                warnings.warn('mpld3 is not available. Zooming disabled.')

1079
1080
1081
        assert self.geotransform and tuple(self.geotransform) != (0, 1, 0, 0, 0, -1), \
            'A valid geotransform is needed for a map visualization. Got %s.' % list(self.geotransform)
        assert self.projection, 'A projection is needed for a map visualization. Got %s.' % self.projection
1082
1083

        # get image to plot
1084
        nodataVal = nodataVal if nodataVal is not None else self.nodata
1085
1086
1087
1088
1089
        image2plot, gt, prj = self._get_plottable_image(xlim, ylim, band, boundsMap=boundsMap,
                                                        boundsMapPrj=boundsMapPrj, res_factor=res_factor,
                                                        nodataVal=nodataVal, out_prj='epsg:4326')

        # calculate corner coordinates of plot
1090
        # if boundsMap:
1091
1092
1093
1094
1095
        #    boundsMapPrj = boundsMapPrj if boundsMapPrj else self.prj
        #    if not prj_equal(boundsMapPrj, 4326):
        #        boundsMap = reproject_shapelyGeometry(box(*boundsMap), boundsMapPrj, 4626).bounds
        #    xmin, ymin, xmax, ymax = boundsMap
        #    UL_XY, UR_XY, LR_XY, LL_XY = (xmin,ymax), (xmax, ymax), (xmax,ymin), (xmin, ymin)
1096
1097
1098
        # else:
        UL_XY, UR_XY, LR_XY, LL_XY = [(YX[1], YX[0]) for YX in GeoArray(image2plot, gt, prj).box.boxMapYX]
        center_lon, center_lat = (UL_XY[0] + UR_XY[0]) / 2., (UL_XY[1] + LL_XY[1]) / 2.
1099
1100
1101
1102
1103
1104

        # create map
        fig = plt.figure(figsize=figsize)
        plt.subplots_adjust(left=0.05, right=0.95, top=0.90, bottom=0.05, wspace=0.15, hspace=0.05)
        ax = ax if ax is not None else plt.subplot(111)

1105
        m = Basemap(projection='tmerc', resolution=None, lon_0=center_lon, lat_0=center_lat,
1106
1107
1108
1109
                    urcrnrlon=UR_XY[0], urcrnrlat=UR_XY[1], llcrnrlon=LL_XY[0], llcrnrlat=LL_XY[1])

        # set color palette
        palette = cmap if cmap else plt.cm.gray
1110
        if nodataVal is not None and np.std(image2plot) != 0:  # do not show nodata
1111
1112
1113
1114
1115
1116
1117
            image2plot = np.ma.masked_equal(image2plot, nodataVal)
            vmin_auto, vmax_auto = np.percentile(image2plot.compressed(), 2), np.percentile(image2plot.compressed(), 98)
            palette.set_bad('aqua', 0)
        else:
            vmin_auto, vmax_auto = np.percentile(image2plot, 2), np.percentile(image2plot, 98)
        vmin = vmin if vmin is not None else vmin_auto
        vmax = vmax if vmax is not None else vmax_auto
1118
        palette.set_over('1')
1119
1120
1121
1122
1123
1124
1125
1126
1127
        palette.set_under('0')

        # add image to map (y-axis must be inverted for basemap)
        if zoomable:
            m.imshow(image2plot, palette, interpolation=interpolation, vmin=vmin, vmax=vmax)
        else:
            m.imshow(np.flipud(image2plot), palette, interpolation=interpolation, vmin=vmin, vmax=vmax)

        # add coordinate grid lines
1128
1129
        parallels = np.arange(-90, 90., 0.25)  # TODO make this adjustable
        # parallels = np.arange(-90, 90., 0.1)
1130
1131
1132
        m.drawparallels(parallels, labels=[1, 0, 0, 0], fontsize=12, linewidth=0.4)

        meridians = np.arange(-180., 180., 0.25)
1133
        # meridians = np.arange(-180., 180., 0.1)
1134
1135
1136
        m.drawmeridians(meridians, labels=[0, 0, 0, 1], fontsize=12, linewidth=0.4)

        if return_map:
1137
            return fig, ax, m
1138
1139
1140
1141
1142
1143
        else:
            plt.show()

    def show_map_utm(self, xlim=None, ylim=None, band=0, figsize=None, interpolation='none', cmap=None,
                     nodataVal=None, vmin=None, vmax=None, res_factor=None, return_map=False):

1144
1145
1146
        try:
            from mpl_toolkits.basemap import Basemap
        except ImportError:
1147
1148
            warnings.warn('This function requires Basemap. You need to install basemap manually (see www./'
                          'matplotlib.org/basemap) if you want to plot maps. It is not automatically installed.')
1149
            raise
1150
1151
1152
1153
        warnings.warn(UserWarning('This function is still under construction and may not work as expected!'))
        # TODO debug this function

        # get image to plot
1154
        nodataVal = nodataVal if nodataVal is not None else self.nodata
1155
1156
1157
1158
        image2plot, gt, prj = self._get_plottable_image(xlim, ylim, band, res_factor, nodataVal)

        # calculate corner coordinates of plot
        box2plot = GeoArray(image2plot, gt, prj).box
1159
        # UL_XY, UR_XY, LR_XY, LL_XY = [(YX[1], YX[0]) for YX in GeoArray(image2plot, gt, prj).box.boxMapYX]
1160
        # Xarr, Yarr = self.box.get_coordArray_MapXY(prj=EPSG2WKT(4326))
1161
1162
        UL_XY, UR_XY, LR_XY, LL_XY = [transform_any_prj(self.projection, 'epsg:4326', x, y) for y, x in
                                      box2plot.boxMapYX]
1163
        center_X, center_Y = (UL_XY[0] + UR_XY[0]) / 2., (UL_XY[1] + LL_XY[1]) / 2.
1164
        center_lon, center_lat = transform_any_prj(prj, 'epsg:4326', center_X, center_Y)
1165
1166
1167
1168
1169
1170
1171
        print(center_lon, center_lat)

        # create map
        fig = plt.figure(figsize=figsize)
        plt.subplots_adjust(left=0.05, right=0.95, top=0.90, bottom=0.05, wspace=0.15, hspace=0.05)
        ax = plt.subplot(111)
        print(UL_XY, UR_XY, LR_XY, LL_XY)
1172
1173
        #        m = Basemap(projection='tmerc', resolution=None, lon_0=center_lon, lat_0=center_lat,
        #                    urcrnrx=UR_XY[0], urcrnry=UR_XY[1], llcrnrx=LL_XY[0], llcrnry=LL_XY[1])
1174
1175
        m = Basemap(projection='tmerc', resolution=None, lon_0=center_lon, lat_0=center_lat,
                    urcrnrlon=UR_XY[0], urcrnrlat=UR_XY[1], llcrnrlon=LL_XY[0], llcrnrlat=LL_XY[1],
1176
                    k_0=0.9996, rsphere=(6378137.00, 6356752.314245179), suppress_ticks=False)
1177
1178
1179
1180
        # m.pcolormesh(Xarr, Yarr, self[:], cmap=plt.cm.jet)

        # set color palette
        palette = cmap if cmap else plt.cm.gray
1181
        if nodataVal is not None:  # do not show nodata
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
            image2plot = np.ma.masked_equal(image2plot, nodataVal)
            vmin_auto, vmax_auto = np.percentile(image2plot.compressed(), 2), np.percentile(image2plot.compressed(), 98)
            palette.set_bad('aqua', 0)
        else:
            vmin_auto, vmax_auto = np.percentile(image2plot, 2), np.percentile(image2plot, 98)
        vmin = vmin if vmin is not None else vmin_auto
        vmax = vmax if vmax is not None else vmax_auto
        palette.set_over('1')
        palette.set_under('0')

        # add image to map (y-axis must be inverted for basemap)
        m.imshow(np.flipud(image2plot), palette, interpolation=interpolation, vmin=vmin, vmax=vmax)

        # add coordinate grid lines
        parallels = np.arange(-90, 90., 0.25)
        m.drawparallels(parallels, labels=[1, 0, 0, 0], fontsize=12, linewidth=0.4)

        meridians = np.arange(-180., 180., 0.25)
        m.drawmeridians(meridians, labels=[0, 0, 0, 1], fontsize=12, linewidth=0.4)

        if return_map:
            return fig, ax, m
        else:
            plt.show()

    def show_footprint(self):
        """This method is intended to be called from Jupyter Notebook and shows a web map containing the calculated
        footprint of GeoArray.
        """

        try:
1213
1214
            import folium
            import geojson
1215
1216
1217
1218
1219
1220
1221
1222
1223
        except ImportError:
            folium, geojson = None, None
        if not folium or not geojson:
            raise ImportError(
                "This method requires the libraries 'folium' and 'geojson'. They can be installed with "
                "the shell command 'pip install folium geojson'.")

        lonlatPoly = reproject_shapelyGeometry(self.footprint_poly, self.epsg, 4326)

1224
        m = folium.Map(location=tuple(np.array(lonlatPoly.centroid.coords.xy).flatten())[::-1])
1225
1226
1227
1228
        gjs = geojson.Feature(geometry=lonlatPoly, properties={})
        folium.GeoJson(gjs).add_to(m)
        return m

1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
    def show_histogram(self, band=1, bins=200, normed=False, exclude_nodata=True, vmin=None, vmax=None, figsize=None):
        # type: (int, int, bool, bool, float, float, tuple) -> None

        """Show a histogram of a given band.

        :param band:            the band to be used to plot the histogram
        :param bins:            number of bins to plot (default: 200)
        :param normed:          whether to normalize the y-axis or not (default: False)
        :param exclude_nodata:  whether tp exclude nodata value from the histogram
        :param vmin:            minimum value for the x-axis of the histogram
        :param vmax:            maximum value for the x-axis of the histogram
        :param figsize:         figure size (tuple)
        """

        if self.nodata is not None and exclude_nodata:
            data = np.ma.masked_equal(self[band], self.nodata)
            data = data.compressed()
        else:
            data = self[band]

        vmin = vmin if vmin is not None else np.percentile(data, 1)
        vmax = vmax if vmax is not None else np.percentile(data, 99)
        image2plot = data

        plt.figure(figsize=figsize)
        plt.hist(list(image2plot.flat), normed=normed, bins=bins, color='gray', range=[vmin, vmax])
        plt.xlabel('Pixel value')
        plt.ylabel('Probabilty' if normed else 'Count')
        plt.show()

        if not self.q:
            print('STD:', np.std(data))
            print('MEAN:', np.mean(data))
            print('2 % percentile:', np.percentile(data, 2))
            print('98 % percentile:', np.percentile(data, 98))

1265
    def get_mapPos(self, mapBounds, mapBounds_prj, band2get=None, out_prj=None, arr_gt=None, arr_prj=None, fillVal=None,
1266
                   rspAlg='near', progress=None, v=False):  # TODO implement slice for indexing bands
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
        # type: (tuple, str, int, str, tuple, str, int, str, bool, bool) -> (np.ndarray, tuple, str)
        """Returns the array data of GeoArray at a given geographic position.

        NOTE: The given mapBounds are snapped to the pixel grid of GeoArray. If the given mapBounds include areas
        outside of the extent of GeoArray, these areas are filled with the fill value of GeoArray.

        :param mapBounds:       xmin, ymin, xmax, ymax
        :param mapBounds_prj:   WKT projection string corresponding to mapBounds
        :param band2get:        band index of the band to be returned (full array if not given)
        :param out_prj:         output projection as WKT string. If not given, the self.projection is used.
        :param arr_gt:          GDAL GeoTransform (taken from self if not given)
        :param arr_prj:         WKT projection string (taken from self if not given)
        :param fillVal:         nodata value
        :param rspAlg:          <str> Resampling method to use. Available methods are:
                                near, bilinear, cubic, cubicspline, lanczos, average, mode, max, min, med, q1, q2
        :param progress:        whether to show progress bars or not
        :param v:               verbose mode (not related to GeoArray.v; must be explicitly set)
        :return:
        """

1287
1288
1289
1290
        arr_gt = arr_gt if arr_gt else self.geotransform
        arr_prj = arr_prj if arr_prj else self.projection
        out_prj = out_prj if out_prj else arr_prj
        fillVal = fillVal if fillVal is not None else self.nodata
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
        progress = progress if progress is not None else self.progress

        if self.is_inmem and (not arr_gt or not arr_prj):
            raise ValueError('In case of in-mem arrays the respective geotransform and projection of the array '
                             'has to be passed.')

        if v:
            print('%s.get_mapPos() input parameters:')
            print('\tmapBounds', mapBounds, '<==>', self.box.boundsMap)
            print('\tEPSG', WKT2EPSG(mapBounds_prj), self.epsg)
            print('\tarr_gt', arr_gt, self.gt)
            print('\tarr_prj', WKT2EPSG(arr_prj), self.epsg)
            print('\tfillVal', fillVal, self.nodata, '\n')

        sub_arr, sub_gt, sub_prj = get_array_at_mapPos(self, arr_gt, arr_prj,
1306
1307
1308
1309
1310
1311
1312
1313
                                                       out_prj=out_prj,
                                                       mapBounds=mapBounds,
                                                       mapBounds_prj=mapBounds_prj,
                                                       fillVal=fillVal,
                                                       rspAlg=rspAlg,
                                                       out_gsd=(self.xgsd, self.ygsd),
                                                       band2get=band2get,
                                                       progress=progress)
1314
1315
        return sub_arr, sub_gt, sub_prj

1316
1317
1318
    def get_subset(self, xslice=None, yslice=None, zslice=None, return_GeoArray=True):
        # type: (slice, slice, slice, bool) -> GeoArray
        """Returns a new instatnce of GeoArray representing a subset of the initial one wit respect to given array position.
Daniel Scheffler's avatar
Daniel Scheffler committed
1319

1320
1321
1322
1323
        :param xslice:          a slice providing the X-position for the subset in the form slice(xstart, xend, xstep)
        :param yslice:          a slice providing the Y-position for the subset in the form slice(ystart, yend, ystep)
        :param zslice:          a slice providing the Z-position for the subset in the form slice(zstart, zend, zstep)
        :param return_GeoArray: whether to return an instance of GeoArray (default) or a tuple(np.ndarray, gt, prj)
Daniel Scheffler's avatar
Daniel Scheffler committed
1324
        :return: