forked from Unidata/netcdf4-python
-
Notifications
You must be signed in to change notification settings - Fork 0
/
netCDF4.pyx
3835 lines (3440 loc) · 164 KB
/
netCDF4.pyx
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
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
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
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
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
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
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
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
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
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
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
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
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
756
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
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
"""
Introduction
============
Python interface to the netCDF version 4 library. U{netCDF version 4
<http://www.unidata.ucar.edu/software/netcdf/netcdf-4>} has many features
not found in earlier versions of the library and is implemented on top of
U{HDF5 <http://www.hdfgroup.org/HDF5>}. This module can read and write
files in both the new netCDF 4 and the old netCDF 3 format, and can create
files that are readable by HDF5 clients. The API modelled after
U{Scientific.IO.NetCDF
<http://dirac.cnrs-orleans.fr/plone/software/scientificpython/>}, and should be
familiar to users of that module.
Most new features of netCDF 4 are implemented, such as multiple
unlimited dimensions, groups and zlib data compression. All the new
numeric data types (such as 64 bit and unsigned integer types) are
implemented. Compound and variable length (vlen) data types are supported,
but the enum and opaque data types are not. Mixtures of compound and vlen
data types (compound types containing vlens, and vlens containing compound
types) are not supported.
Download
========
- Latest bleeding-edge code from the U{subversion repository <http://code.google.com/p/netcdf4-python/source>}.
- Latest U{releases <http://code.google.com/p/netcdf4-python/downloads/list>}
(source code and windows installers).
Requires
========
- numpy array module U{http://numpy.scipy.org}, version 1.3.0 or later (1.5.1
or higher recommended, required if using python 3).
- U{Cython <http://cython.org>} is optional - if it is installed setup.py will
use it to recompile the Cython source code into C, using conditional compilation
to enable features in the netCDF API that have been added since version 4.1.1. If
Cython is not installed, these features (such as the ability to rename Group objects)
will be disabled to preserve backward compatibility with older versions of the netCDF
library.
- For python < 2.7, the ordereddict module U{http://python.org/pypi/ordereddict}.
- The HDF5 C library version 1.8.4-patch1 or higher (1.8.8 or higher
recommended) from U{ftp://ftp.hdfgroup.org/HDF5/current/src}.
Be sure to build with 'C{--enable-hl --enable-shared}'.
- U{Libcurl <http://curl.haxx.se/libcurl/>}, if you want
U{OPeNDAP<http://opendap.org/>} support.
- U{HDF4 <http://www.hdfgroup.org/products/hdf4/>}, if you want
to be able to read HDF4 "Scientific Dataset" (SD) files.
- The netCDF-4 C library from U{ftp://ftp.unidata.ucar.edu/pub/netcdf}.
Version 4.1.1 or higher is required (4.2 or higher recommended).
Be sure to build with 'C{--enable-netcdf-4 --enable-shared}', and set
C{CPPFLAGS="-I $HDF5_DIR/include"} and C{LDFLAGS="-L $HDF5_DIR/lib"},
where C{$HDF5_DIR} is the directory where HDF5 was installed.
If you want U{OPeNDAP<http://opendap.org/>} support, add 'C{--enable-dap}'.
If you want HDF4 SD support, add 'C{--enable-hdf4}' and add
the location of the HDF4 headers and library to C{CPPFLAGS} and C{LDFLAGS}.
Install
=======
- install the requisite python modules and C libraries (see above). It's
easiest if all the C libs are built as shared libraries.
- optionally, set the C{HDF5_DIR} environment variable to point to where HDF5
is installed (the libs in C{$HDF5_DIR/lib}, the headers in
C{$HDF5_DIR/include}). If the headers and libs are installed in different
places, you can use C{HDF5_INCDIR} and C{HDF5_LIBDIR} to define the locations
of the headers and libraries independently.
- optionally, set the C{NETCDF4_DIR} (or C{NETCDF4_INCDIR} and C{NETCDF4_LIBDIR})
environment variable(s) to point to
where the netCDF version 4 library and headers are installed.
- If the locations of the HDF5 and netCDF libs and headers are not specified
with environment variables, some standard locations will be searched.
- if HDF5 was built as a static library with U{szip
<http://www.hdfgroup.org/doc_resource/SZIP/>} support,
you may also need to set the C{SZIP_DIR} (or C{SZIP_INCDIR} and C{SZIP_LIBDIR})
environment variable(s) to point to where szip is installed. Note that
the netCDF library does not support creating szip compressed files, but can read szip
compressed files if the HDF5 lib is configured to support szip.
- if netCDF lib was built as a static library with HDF4 and/or OpenDAP
support, you may also need to set C{HDF4_DIR}, C{JPEG_DIR} and/or
C{CURL_DIR}.
- Instead of using environment variables to specify the locations of the
required libraries, you can either let setup.py try to auto-detect their
locations, or use the file C{setup.cfg} to specify them. To use this
method, copy the file C{setup.cfg.template} to C{setup.cfg},
then open C{setup.cfg} in a text editor and follow the instructions in the
comments for editing. If you use C{setup.cfg}, environment variables will be
ignored.
- If you are using netcdf 4.1.2 or higher, instead of setting all those
enviroment variables defining where libs are installed, you can just set one
environment variable, USE_NCCONFIG, to 1. This will tell python to run the
netcdf nc-config utility to determine where all the dependencies live.
- run C{python setup.py build}, then C{python setup.py install} (as root if
necessary).
- If using environment variables to specify build options, be sure to run
'python setup.py build' *without* using sudo. sudo does not pass environment
variables. If you run 'setup.py build' first without sudo, you can run
'setup.py install' with sudo.
- run the tests in the 'test' directory by running C{python run_all.py}.
Tutorial
========
1) Creating/Opening/Closing a netCDF file
-----------------------------------------
To create a netCDF file from python, you simply call the L{Dataset}
constructor. This is also the method used to open an existing netCDF
file. If the file is open for write access (C{w, r+} or C{a}), you may
write any type of data including new dimensions, groups, variables and
attributes. netCDF files come in several flavors (C{NETCDF3_CLASSIC,
NETCDF3_64BIT, NETCDF4_CLASSIC}, and C{NETCDF4}). The first two flavors
are supported by version 3 of the netCDF library. C{NETCDF4_CLASSIC}
files use the version 4 disk format (HDF5), but do not use any features
not found in the version 3 API. They can be read by netCDF 3 clients
only if they have been relinked against the netCDF 4 library. They can
also be read by HDF5 clients. C{NETCDF4} files use the version 4 disk
format (HDF5) and use the new features of the version 4 API. The
C{netCDF4} module can read and write files in any of these formats. When
creating a new file, the format may be specified using the C{format}
keyword in the C{Dataset} constructor. The default format is
C{NETCDF4}. To see how a given file is formatted, you can examine the
C{data_model} L{Dataset} attribute. Closing the netCDF file is
accomplished via the L{close<Dataset.close>} method of the L{Dataset}
instance.
Here's an example:
>>> from netCDF4 import Dataset
>>> rootgrp = Dataset('test.nc', 'w', format='NETCDF4')
>>> print rootgrp.data_model
NETCDF4
>>>
>>> rootgrp.close()
Remote U{OPeNDAP<http://opendap.org>}-hosted datasets can be accessed for
reading over http if a URL is provided to the L{Dataset} constructor instead of a
filename. However, this requires that the netCDF library be built with
OPenDAP support, via the C{--enable-dap} configure option (added in
version 4.0.1).
2) Groups in a netCDF file
--------------------------
netCDF version 4 added support for organizing data in hierarchical
groups, which are analagous to directories in a filesystem. Groups serve
as containers for variables, dimensions and attributes, as well as other
groups. A C{netCDF4.Dataset} defines creates a special group, called
the 'root group', which is similar to the root directory in a unix
filesystem. To create L{Group} instances, use the
L{createGroup<Dataset.createGroup>} method of a L{Dataset} or L{Group}
instance. L{createGroup<Dataset.createGroup>} takes a single argument, a
python string containing the name of the new group. The new L{Group}
instances contained within the root group can be accessed by name using
the C{groups} dictionary attribute of the L{Dataset} instance. Only
C{NETCDF4} formatted files support Groups, if you try to create a Group
in a netCDF 3 file you will get an error message.
>>> rootgrp = Dataset('test.nc', 'a')
>>> fcstgrp = rootgrp.createGroup('forecasts')
>>> analgrp = rootgrp.createGroup('analyses')
>>> print rootgrp.groups
OrderedDict([('forecasts', <netCDF4.Group object at 0x1b4b7b0>),
('analyses', <netCDF4.Group object at 0x1b4b970>)])
>>>
Groups can exist within groups in a L{Dataset}, just as directories
exist within directories in a unix filesystem. Each L{Group} instance
has a C{'groups'} attribute dictionary containing all of the group
instances contained within that group. Each L{Group} instance also has a
C{'path'} attribute that contains a simulated unix directory path to
that group.
Here's an example that shows how to navigate all the groups in a
L{Dataset}. The function C{walktree} is a Python generator that is used
to walk the directory tree. Note that printing the L{Dataset} or L{Group}
object yields summary information about it's contents.
>>> fcstgrp1 = fcstgrp.createGroup('model1')
>>> fcstgrp2 = fcstgrp.createGroup('model2')
>>> def walktree(top):
>>> values = top.groups.values()
>>> yield values
>>> for value in top.groups.values():
>>> for children in walktree(value):
>>> yield children
>>> print rootgrp
>>> for children in walktree(rootgrp):
>>> for child in children:
>>> print child
<type 'netCDF4.Dataset'>
root group (NETCDF4 file format):
dimensions:
variables:
groups: forecasts, analyses
<type 'netCDF4.Group'>
group /forecasts:
dimensions:
variables:
groups: model1, model2
<type 'netCDF4.Group'>
group /analyses:
dimensions:
variables:
groups:
<type 'netCDF4.Group'>
group /forecasts/model1:
dimensions:
variables:
groups:
<type 'netCDF4.Group'>
group /forecasts/model2:
dimensions:
variables:
groups:
>>>
3) Dimensions in a netCDF file
------------------------------
netCDF defines the sizes of all variables in terms of dimensions, so
before any variables can be created the dimensions they use must be
created first. A special case, not often used in practice, is that of a
scalar variable, which has no dimensions. A dimension is created using
the L{createDimension<Dataset.createDimension>} method of a L{Dataset}
or L{Group} instance. A Python string is used to set the name of the
dimension, and an integer value is used to set the size. To create an
unlimited dimension (a dimension that can be appended to), the size
value is set to C{None} or 0. In this example, there both the C{time} and
C{level} dimensions are unlimited. Having more than one unlimited
dimension is a new netCDF 4 feature, in netCDF 3 files there may be only
one, and it must be the first (leftmost) dimension of the variable.
>>> level = rootgrp.createDimension('level', None)
>>> time = rootgrp.createDimension('time', None)
>>> lat = rootgrp.createDimension('lat', 73)
>>> lon = rootgrp.createDimension('lon', 144)
All of the L{Dimension} instances are stored in a python dictionary.
>>> print rootgrp.dimensions
OrderedDict([('level', <netCDF4.Dimension object at 0x1b48030>),
('time', <netCDF4.Dimension object at 0x1b481c0>),
('lat', <netCDF4.Dimension object at 0x1b480f8>),
('lon', <netCDF4.Dimension object at 0x1b48a08>)])
>>>
Calling the python C{len} function with a L{Dimension} instance returns
the current size of that dimension.
The L{isunlimited<Dimension.isunlimited>} method of a L{Dimension} instance
can be used to determine if the dimensions is unlimited, or appendable.
>>> print len(lon)
144
>>> print len.is_unlimited()
False
>>> print time.is_unlimited()
True
>>>
Printing the L{Dimension} object
provides useful summary info, including the name and length of the dimension,
and whether it is unlimited.
>>> for dimobj in rootgrp.dimensions.values():
>>> print dimobj
<type 'netCDF4.Dimension'> (unlimited): name = 'level', size = 0
<type 'netCDF4.Dimension'> (unlimited): name = 'time', size = 0
<type 'netCDF4.Dimension'>: name = 'lat', size = 73
<type 'netCDF4.Dimension'>: name = 'lon', size = 144
<type 'netCDF4.Dimension'> (unlimited): name = 'time', size = 0
>>>
L{Dimension} names can be changed using the
L{renameDimension<Dataset.renameDimension>} method of a L{Dataset} or
L{Group} instance.
4) Variables in a netCDF file
-----------------------------
netCDF variables behave much like python multidimensional array objects
supplied by the U{numpy module <http://numpy.scipy.org>}. However,
unlike numpy arrays, netCDF4 variables can be appended to along one or
more 'unlimited' dimensions. To create a netCDF variable, use the
L{createVariable<Dataset.createVariable>} method of a L{Dataset} or
L{Group} instance. The L{createVariable<Dataset.createVariable>} method
has two mandatory arguments, the variable name (a Python string), and
the variable datatype. The variable's dimensions are given by a tuple
containing the dimension names (defined previously with
L{createDimension<Dataset.createDimension>}). To create a scalar
variable, simply leave out the dimensions keyword. The variable
primitive datatypes correspond to the dtype attribute of a numpy array.
You can specify the datatype as a numpy dtype object, or anything that
can be converted to a numpy dtype object. Valid datatype specifiers
include: C{'f4'} (32-bit floating point), C{'f8'} (64-bit floating
point), C{'i4'} (32-bit signed integer), C{'i2'} (16-bit signed
integer), C{'i8'} (64-bit singed integer), C{'i1'} (8-bit signed
integer), C{'u1'} (8-bit unsigned integer), C{'u2'} (16-bit unsigned
integer), C{'u4'} (32-bit unsigned integer), C{'u8'} (64-bit unsigned
integer), or C{'S1'} (single-character string). The old Numeric
single-character typecodes (C{'f'},C{'d'},C{'h'},
C{'s'},C{'b'},C{'B'},C{'c'},C{'i'},C{'l'}), corresponding to
(C{'f4'},C{'f8'},C{'i2'},C{'i2'},C{'i1'},C{'i1'},C{'S1'},C{'i4'},C{'i4'}),
will also work. The unsigned integer types and the 64-bit integer type
can only be used if the file format is C{NETCDF4}.
The dimensions themselves are usually also defined as variables, called
coordinate variables. The L{createVariable<Dataset.createVariable>}
method returns an instance of the L{Variable} class whose methods can be
used later to access and set variable data and attributes.
>>> times = rootgrp.createVariable('time','f8',('time',))
>>> levels = rootgrp.createVariable('level','i4',('level',))
>>> latitudes = rootgrp.createVariable('latitude','f4',('lat',))
>>> longitudes = rootgrp.createVariable('longitude','f4',('lon',))
>>> # two dimensions unlimited.
>>> temp = rootgrp.createVariable('temp','f4',('time','level','lat','lon',))
All of the variables in the L{Dataset} or L{Group} are stored in a
Python dictionary, in the same way as the dimensions:
>>> print rootgrp.variables
OrderedDict([('time', <netCDF4.Variable object at 0x1b4ba70>),
('level', <netCDF4.Variable object at 0x1b4bab0>),
('latitude', <netCDF4.Variable object at 0x1b4baf0>),
('longitude', <netCDF4.Variable object at 0x1b4bb30>),
('temp', <netCDF4.Variable object at 0x1b4bb70>)])
>>>
To get summary info on a L{Variable} instance in an interactive session, just print it.
>>> print rootgrp.variables['temp']
<type 'netCDF4.Variable'>
float32 temp(time, level, lat, lon)
least_significant_digit: 3
units: K
unlimited dimensions: time, level
current shape = (0, 0, 73, 144)
>>>
L{Variable} names can be changed using the
L{renameVariable<Dataset.renameVariable>} method of a L{Dataset}
instance.
5) Attributes in a netCDF file
------------------------------
There are two types of attributes in a netCDF file, global and variable.
Global attributes provide information about a group, or the entire
dataset, as a whole. L{Variable} attributes provide information about
one of the variables in a group. Global attributes are set by assigning
values to L{Dataset} or L{Group} instance variables. L{Variable}
attributes are set by assigning values to L{Variable} instances
variables. Attributes can be strings, numbers or sequences. Returning to
our example,
>>> import time
>>> rootgrp.description = 'bogus example script'
>>> rootgrp.history = 'Created ' + time.ctime(time.time())
>>> rootgrp.source = 'netCDF4 python module tutorial'
>>> latitudes.units = 'degrees north'
>>> longitudes.units = 'degrees east'
>>> levels.units = 'hPa'
>>> temp.units = 'K'
>>> times.units = 'hours since 0001-01-01 00:00:00.0'
>>> times.calendar = 'gregorian'
The L{ncattrs<Dataset.ncattrs>} method of a L{Dataset}, L{Group} or
L{Variable} instance can be used to retrieve the names of all the netCDF
attributes. This method is provided as a convenience, since using the
built-in C{dir} Python function will return a bunch of private methods
and attributes that cannot (or should not) be modified by the user.
>>> for name in rootgrp.ncattrs():
>>> print 'Global attr', name, '=', getattr(rootgrp,name)
Global attr description = bogus example script
Global attr history = Created Mon Nov 7 10.30:56 2005
Global attr source = netCDF4 python module tutorial
The C{__dict__} attribute of a L{Dataset}, L{Group} or L{Variable}
instance provides all the netCDF attribute name/value pairs in a python
dictionary:
>>> print rootgrp.__dict__
OrderedDict([(u'description', u'bogus example script'),
(u'history', u'Created Thu Mar 3 19:30:33 2011'),
(u'source', u'netCDF4 python module tutorial')])
Attributes can be deleted from a netCDF L{Dataset}, L{Group} or
L{Variable} using the python C{del} statement (i.e. C{del grp.foo}
removes the attribute C{foo} the the group C{grp}).
6) Writing data to and retrieving data from a netCDF variable
-------------------------------------------------------------
Now that you have a netCDF L{Variable} instance, how do you put data
into it? You can just treat it like an array and assign data to a slice.
>>> import numpy
>>> lats = numpy.arange(-90,91,2.5)
>>> lons = numpy.arange(-180,180,2.5)
>>> latitudes[:] = lats
>>> longitudes[:] = lons
>>> print 'latitudes =\\n',latitudes[:]
latitudes =
[-90. -87.5 -85. -82.5 -80. -77.5 -75. -72.5 -70. -67.5 -65. -62.5
-60. -57.5 -55. -52.5 -50. -47.5 -45. -42.5 -40. -37.5 -35. -32.5
-30. -27.5 -25. -22.5 -20. -17.5 -15. -12.5 -10. -7.5 -5. -2.5
0. 2.5 5. 7.5 10. 12.5 15. 17.5 20. 22.5 25. 27.5
30. 32.5 35. 37.5 40. 42.5 45. 47.5 50. 52.5 55. 57.5
60. 62.5 65. 67.5 70. 72.5 75. 77.5 80. 82.5 85. 87.5
90. ]
>>>
Unlike NumPy's array objects, netCDF L{Variable}
objects with unlimited dimensions will grow along those dimensions if you
assign data outside the currently defined range of indices.
>>> # append along two unlimited dimensions by assigning to slice.
>>> nlats = len(rootgrp.dimensions['lat'])
>>> nlons = len(rootgrp.dimensions['lon'])
>>> print 'temp shape before adding data = ',temp.shape
temp shape before adding data = (0, 0, 73, 144)
>>>
>>> from numpy.random import uniform
>>> temp[0:5,0:10,:,:] = uniform(size=(5,10,nlats,nlons))
>>> print 'temp shape after adding data = ',temp.shape
temp shape after adding data = (6, 10, 73, 144)
>>>
>>> # levels have grown, but no values yet assigned.
>>> print 'levels shape after adding pressure data = ',levels.shape
levels shape after adding pressure data = (10,)
>>>
Note that the size of the levels variable grows when data is appended
along the C{level} dimension of the variable C{temp}, even though no
data has yet been assigned to levels.
>>> # now, assign data to levels dimension variable.
>>> levels[:] = [1000.,850.,700.,500.,300.,250.,200.,150.,100.,50.]
However, that there are some differences between NumPy and netCDF
variable slicing rules. Slices behave as usual, being specified as a
C{start:stop:step} triplet. Using a scalar integer index C{i} takes the ith
element and reduces the rank of the output array by one. Boolean array and
integer sequence indexing behaves differently for netCDF variables
than for numpy arrays. Only 1-d boolean arrays and integer sequences are
allowed, and these indices work independently along each dimension (similar
to the way vector subscripts work in fortran). This means that
>>> temp[0, 0, [0,1,2,3], [0,1,2,3]]
returns an array of shape (4,4) when slicing a netCDF variable, but for a
numpy array it returns an array of shape (4,).
Similarly, a netCDF variable of shape C{(2,3,4,5)} indexed
with C{[0, array([True, False, True]), array([False, True, True, True]), :]}
would return a C{(2, 3, 5)} array. In NumPy, this would raise an error since
it would be equivalent to C{[0, [0,1], [1,2,3], :]}. While this behaviour can
cause some confusion for those used to NumPy's 'fancy indexing' rules, it
provides a very powerful way to extract data from multidimensional netCDF
variables by using logical operations on the dimension arrays to create slices.
For example,
>>> tempdat = temp[::2, [1,3,6], lats>0, lons>0]
will extract time indices 0,2 and 4, pressure levels
850, 500 and 200 hPa, all Northern Hemisphere latitudes and Eastern
Hemisphere longitudes, resulting in a numpy array of shape (3, 3, 36, 71).
>>> print 'shape of fancy temp slice = ',tempdat.shape
shape of fancy temp slice = (3, 3, 36, 71)
>>>
Time coordinate values pose a special challenge to netCDF users. Most
metadata standards (such as CF and COARDS) specify that time should be
measure relative to a fixed date using a certain calendar, with units
specified like C{hours since YY:MM:DD hh-mm-ss}. These units can be
awkward to deal with, without a utility to convert the values to and
from calendar dates. The functione called L{num2date} and L{date2num} are
provided with this package to do just that. Here's an example of how they
can be used:
>>> # fill in times.
>>> from datetime import datetime, timedelta
>>> from netCDF4 import num2date, date2num
>>> dates = [datetime(2001,3,1)+n*timedelta(hours=12) for n in range(temp.shape[0])]
>>> times[:] = date2num(dates,units=times.units,calendar=times.calendar)
>>> print 'time values (in units %s): ' % times.units+'\\n',times[:]
time values (in units hours since January 1, 0001):
[ 17533056. 17533068. 17533080. 17533092. 17533104.]
>>>
>>> dates = num2date(times[:],units=times.units,calendar=times.calendar)
>>> print 'dates corresponding to time values:\\n',dates
dates corresponding to time values:
[2001-03-01 00:00:00 2001-03-01 12:00:00 2001-03-02 00:00:00
2001-03-02 12:00:00 2001-03-03 00:00:00]
>>>
L{num2date} converts numeric values of time in the specified C{units}
and C{calendar} to datetime objects, and L{date2num} does the reverse.
All the calendars currently defined in the U{CF metadata convention
<http://cf-pcmdi.llnl.gov/documents/cf-conventions/>} are supported.
A function called L{date2index} is also provided which returns the indices
of a netCDF time variable corresponding to a sequence of datetime instances.
7) Reading data from a multi-file netCDF dataset.
-------------------------------------------------
If you want to read data from a variable that spans multiple netCDF files,
you can use the L{MFDataset} class to read the data as if it were
contained in a single file. Instead of using a single filename to create
a L{Dataset} instance, create a L{MFDataset} instance with either a list
of filenames, or a string with a wildcard (which is then converted to
a sorted list of files using the python glob module).
Variables in the list of files that share the same unlimited
dimension are aggregated together, and can be sliced across multiple
files. To illustrate this, let's first create a bunch of netCDF files with
the same variable (with the same unlimited dimension). The files
must in be in C{NETCDF3_64BIT}, C{NETCDF3_CLASSIC} or
C{NETCDF4_CLASSIC format} (C{NETCDF4} formatted multi-file
datasets are not supported).
>>> for nfile in range(10):
>>> f = Dataset('mftest'+repr(nfile)+'.nc','w',format='NETCDF4_CLASSIC')
>>> f.createDimension('x',None)
>>> x = f.createVariable('x','i',('x',))
>>> x[0:10] = numpy.arange(nfile*10,10*(nfile+1))
>>> f.close()
Now read all the files back in at once with L{MFDataset}
>>> from netCDF4 import MFDataset
>>> f = MFDataset('mftest*nc')
>>> print f.variables['x'][:]
[ 0 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 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 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 73 74
75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99]
>>>
Note that MFDataset can only be used to read, not write, multi-file
datasets.
8) Efficient compression of netCDF variables
--------------------------------------------
Data stored in netCDF 4 L{Variable} objects can be compressed and
decompressed on the fly. The parameters for the compression are
determined by the C{zlib}, C{complevel} and C{shuffle} keyword arguments
to the L{createVariable<Dataset.createVariable>} method. To turn on
compression, set C{zlib=True}. The C{complevel} keyword regulates the
speed and efficiency of the compression (1 being fastest, but lowest
compression ratio, 9 being slowest but best compression ratio). The
default value of C{complevel} is 4. Setting C{shuffle=False} will turn
off the HDF5 shuffle filter, which de-interlaces a block of data before
compression by reordering the bytes. The shuffle filter can
significantly improve compression ratios, and is on by default. Setting
C{fletcher32} keyword argument to
L{createVariable<Dataset.createVariable>} to C{True} (it's C{False} by
default) enables the Fletcher32 checksum algorithm for error detection.
It's also possible to set the HDF5 chunking parameters and endian-ness
of the binary data stored in the HDF5 file with the C{chunksizes}
and C{endian} keyword arguments to
L{createVariable<Dataset.createVariable>}. These keyword arguments only
are relevant for C{NETCDF4} and C{NETCDF4_CLASSIC} files (where the
underlying file format is HDF5) and are silently ignored if the file
format is C{NETCDF3_CLASSIC} or C{NETCDF3_64BIT},
If your data only has a certain number of digits of precision (say for
example, it is temperature data that was measured with a precision of
0.1 degrees), you can dramatically improve zlib compression by
quantizing (or truncating) the data using the C{least_significant_digit}
keyword argument to L{createVariable<Dataset.createVariable>}. The least
significant digit is the power of ten of the smallest decimal place in
the data that is a reliable value. For example if the data has a
precision of 0.1, then setting C{least_significant_digit=1} will cause
data the data to be quantized using C{numpy.around(scale*data)/scale}, where
scale = 2**bits, and bits is determined so that a precision of 0.1 is
retained (in this case bits=4). Effectively, this makes the compression
'lossy' instead of 'lossless', that is some precision in the data is
sacrificed for the sake of disk space.
In our example, try replacing the line
>>> temp = rootgrp.createVariable('temp','f4',('time','level','lat','lon',))
with
>>> temp = dataset.createVariable('temp','f4',('time','level','lat','lon',),zlib=True)
and then
>>> temp = dataset.createVariable('temp','f4',('time','level','lat','lon',),zlib=True,least_significant_digit=3)
and see how much smaller the resulting files are.
9) Beyond homogenous arrays of a fixed type - compound data types
-----------------------------------------------------------------
Compound data types map directly to numpy structured (a.k.a 'record'
arrays). Structured arrays are akin to C structs, or derived types
in Fortran. They allow for the construction of table-like structures
composed of combinations of other data types, including other
compound types. Compound types might be useful for representing multiple
parameter values at each point on a grid, or at each time and space
location for scattered (point) data. You can then access all the
information for a point by reading one variable, instead of reading
different parameters from different variables. Compound data types
are created from the corresponding numpy data type using the
L{createCompoundType<Dataset.createCompoundType>} method of a L{Dataset} or L{Group} instance.
Since there is no native complex data type in netcdf, compound types are handy
for storing numpy complex arrays. Here's an example:
>>> f = Dataset('complex.nc','w')
>>> size = 3 # length of 1-d complex array
>>> # create sample complex data.
>>> datac = numpy.exp(1j*(1.+numpy.linspace(0, numpy.pi, size)))
>>> # create complex128 compound data type.
>>> complex128 = numpy.dtype([('real',numpy.float64),('imag',numpy.float64)])
>>> complex128_t = f.createCompoundType(complex128,'complex128')
>>> # create a variable with this data type, write some data to it.
>>> f.createDimension('x_dim',None)
>>> v = f.createVariable('cmplx_var',complex128_t,'x_dim')
>>> data = numpy.empty(size,complex128) # numpy structured array
>>> data['real'] = datac.real; data['imag'] = datac.imag
>>> v[:] = data # write numpy structured array to netcdf compound var
>>> # close and reopen the file, check the contents.
>>> f.close(); f = Dataset('complex.nc')
>>> v = f.variables['cmplx_var']
>>> datain = v[:] # read in all the data into a numpy structured array
>>> # create an empty numpy complex array
>>> datac2 = numpy.empty(datain.shape,numpy.complex128)
>>> # .. fill it with contents of structured array.
>>> datac2.real = datain['real']; datac2.imag = datain['imag']
>>> print datac.dtype,datac # original data
complex128 [ 0.54030231+0.84147098j -0.84147098+0.54030231j -0.54030231-0.84147098j]
>>>
>>> print datac2.dtype,datac2 # data from file
complex128 [ 0.54030231+0.84147098j -0.84147098+0.54030231j -0.54030231-0.84147098j]
>>>
Compound types can be nested, but you must create the 'inner'
ones first. All of the compound types defined for a L{Dataset} or L{Group} are stored in a
Python dictionary, just like variables and dimensions. As always, printing
objects gives useful summary information in an interactive session:
>>> print f
<type 'netCDF4.Dataset'>
root group (NETCDF4 file format):
dimensions: x_dim
variables: cmplx_var
groups:
<type 'netCDF4.Variable'>
>>> print f.variables['cmplx_var']
compound cmplx_var(x_dim)
compound data type: [('real', '<f8'), ('imag', '<f8')]
unlimited dimensions: x_dim
current shape = (3,)
>>> print f.cmptypes
OrderedDict([('complex128', <netCDF4.CompoundType object at 0x1029eb7e8>)])
>>> print f.cmptypes['complex128']
<type 'netCDF4.CompoundType'>: name = 'complex128', numpy dtype = [(u'real','<f8'), (u'imag', '<f8')]
>>>
10) Variable-length (vlen) data types.
--------------------------------------
NetCDF 4 has support for variable-length or "ragged" arrays. These are arrays
of variable length sequences having the same type. To create a variable-length
data type, use the L{createVLType<Dataset.createVLType>} method
method of a L{Dataset} or L{Group} instance.
>>> f = Dataset('tst_vlen.nc','w')
>>> vlen_t = f.createVLType(numpy.int32, 'phony_vlen')
The numpy datatype of the variable-length sequences and the name of the
new datatype must be specified. Any of the primitive datatypes can be
used (signed and unsigned integers, 32 and 64 bit floats, and characters),
but compound data types cannot.
A new variable can then be created using this datatype.
>>> x = f.createDimension('x',3)
>>> y = f.createDimension('y',4)
>>> vlvar = f.createVariable('phony_vlen_var', vlen_t, ('y','x'))
Since there is no native vlen datatype in numpy, vlen arrays are represented
in python as object arrays (arrays of dtype C{object}). These are arrays whose
elements are Python object pointers, and can contain any type of python object.
For this application, they must contain 1-D numpy arrays all of the same type
but of varying length.
In this case, they contain 1-D numpy C{int32} arrays of random length betwee
1 and 10.
>>> import random
>>> data = numpy.empty(len(y)*len(x),object)
>>> for n in range(len(y)*len(x)):
>>> data[n] = numpy.arange(random.randint(1,10),dtype='int32')+1
>>> data = numpy.reshape(data,(len(y),len(x)))
>>> vlvar[:] = data
>>> print 'vlen variable =\\n',vlvar[:]
vlen variable =
[[[ 1 2 3 4 5 6 7 8 9 10] [1 2 3 4 5] [1 2 3 4 5 6 7 8]]
[[1 2 3 4 5 6 7] [1 2 3 4 5 6] [1 2 3 4 5]]
[[1 2 3 4 5] [1 2 3 4] [1]]
[[ 1 2 3 4 5 6 7 8 9 10] [ 1 2 3 4 5 6 7 8 9 10]
[1 2 3 4 5 6 7 8]]]
>>> print f
<type 'netCDF4.Dataset'>
root group (NETCDF4 file format):
dimensions: x, y
variables: phony_vlen_var
groups:
>>> print f.variables['phony_vlen_var']
<type 'netCDF4.Variable'>
vlen phony_vlen_var(y, x)
vlen data type: int32
unlimited dimensions:
current shape = (4, 3)
>>> print f.VLtypes['phony_vlen']
<type 'netCDF4.VLType'>: name = 'phony_vlen', numpy dtype = int32
>>>
Numpy object arrays containing python strings can also be written as vlen
variables, For vlen strings, you don't need to create a vlen data type.
Instead, simply use the python C{str} builtin instead of a numpy datatype when
calling the L{createVariable<Dataset.createVariable>} method.
>>> z = f.createDimension('z',10)
>>> strvar = rootgrp.createVariable('strvar', str, 'z')
In this example, an object array is filled with random python strings with
random lengths between 2 and 12 characters, and the data in the object
array is assigned to the vlen string variable.
>>> chars = '1234567890aabcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ'
>>> data = NP.empty(10,'O')
>>> for n in range(10):
>>> stringlen = random.randint(2,12)
>>> data[n] = ''.join([random.choice(chars) for i in range(stringlen)])
>>> strvar[:] = data
>>> print 'variable-length string variable:\\n',strvar[:]
variable-length string variable:
[aDy29jPt jd7aplD b8t4RM jHh8hq KtaPWF9cQj Q1hHN5WoXSiT MMxsVeq td LUzvVTzj
5DS9X8S]
>>> print f
<type 'netCDF4.Dataset'>
root group (NETCDF4 file format):
dimensions: x, y, z
variables: phony_vlen_var, strvar
groups:
>>> print f.variables['strvar']
<type 'netCDF4.Variable'>
vlen strvar(z)
vlen data type: <type 'str'>
unlimited dimensions:
current size = (10,)
>>>
All of the code in this tutorial is available in C{examples/tutorial.py},
Unit tests are in the C{test} directory.
@contact: Jeffrey Whitaker <[email protected]>
@copyright: 2008 by Jeffrey Whitaker.
@license: Permission to use, copy, modify, and distribute this software and
its documentation for any purpose and without fee is hereby granted,
provided that the above copyright notice appear in all copies and that
both the copyright notice and this permission notice appear in
supporting documentation.
THE AUTHOR DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE,
INCLUDING ALL IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS. IN NO
EVENT SHALL THE AUTHOR BE LIABLE FOR ANY SPECIAL, INDIRECT OR
CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF
USE, DATA OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR
OTHER TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR
PERFORMANCE OF THIS SOFTWARE."""
__test__ = None
del __test__ # hack so epydoc doesn't show __test__
# Make changes to this file, not the c-wrappers that Pyrex generates.
# pure python utilities
from netCDF4_utils import _StartCountStride, _quantize, _find_dim, \
_out_array_shape, _sortbylist, _tostr
# try to use built-in ordered dict in python >= 2.7
try:
from collections import OrderedDict
except ImportError: # or else use drop-in substitute
try:
from ordereddict import OrderedDict
except ImportError:
raise ImportError('please install ordereddict (https://pypi.python.org/pypi/ordereddict)')
__version__ = "1.0.9"
# Initialize numpy
import posixpath
import netcdftime
import numpy
import weakref
import sys
import warnings
from glob import glob
from numpy import ma
from numpy import __version__ as _npversion
if _npversion.split('.')[0] < '1':
raise ImportError('requires numpy version 1.0rc1 or later')
import_array()
include "netCDF4.pxi"
# include pure python utility functions and MFDataset class.
# (use include instead of importing them so docstrings
# get included in C extension code).
include "utils.pyx"
include "constants.pyx"
# check for required version of netcdf-4 and hdf5.
def _gethdf5libversion():
majorvers = H5_VERS_MAJOR
minorvers = H5_VERS_MINOR
releasevers = H5_VERS_RELEASE
patchstring = H5_VERS_SUBRELEASE.decode('ascii')
if not patchstring:
return '%d.%d.%d' % (majorvers,minorvers,releasevers)
else:
return '%d.%d.%d-%s' % (majorvers,minorvers,releasevers,patchstring)
__netcdf4libversion__ = getlibversion().split()[0]
__hdf5libversion__ = _gethdf5libversion()
__has_rename_grp__ = HAS_RENAME_GRP
__has_nc_inq_path__ = HAS_NC_INQ_PATH
__has_nc_inq_format_extended__ = HAS_NC_INQ_FORMAT_EXTENDED
# numpy data type <--> netCDF 4 data type mapping.
_nptonctype = {'U1' : NC_CHAR,
'S1' : NC_CHAR,
'i1' : NC_BYTE,
'u1' : NC_UBYTE,
'i2' : NC_SHORT,
'u2' : NC_USHORT,
'i4' : NC_INT,
'u4' : NC_UINT,
'i8' : NC_INT64,
'u8' : NC_UINT64,
'f4' : NC_FLOAT,
'f8' : NC_DOUBLE}
default_fillvals = {#'S1':NC_FILL_CHAR,
'U1':'\0',
'S1':'\0',
'i1':NC_FILL_BYTE,
'u1':NC_FILL_UBYTE,
'i2':NC_FILL_SHORT,
'u2':NC_FILL_USHORT,
'i4':NC_FILL_INT,
'u4':NC_FILL_UINT,
'i8':NC_FILL_INT64,
'u8':NC_FILL_UINT64,
'f4':NC_FILL_FLOAT,
'f8':NC_FILL_DOUBLE}
is_native_little = numpy.dtype('<f4').byteorder == '='
is_native_big = numpy.dtype('>f4').byteorder == '='
# hard code this here, instead of importing from netcdf.h
# so it will compile with versions <= 4.2.
NC_DISKLESS = 0x0008
# encoding used to convert strings to bytes when writing text data
# to the netcdf file, and for converting bytes to strings when reading
# from the netcdf file.
default_encoding = 'utf-8'
# unicode decode/encode error handling. Replace bad chars with "?"
# can be set to 'strict' or 'ignore'.
unicode_error = 'replace'
python3 = sys.version_info[0] > 2
_nctonptype = {}
for _key,_value in _nptonctype.items():
_nctonptype[_value] = _key
_supportedtypes = _nptonctype.keys()
# make sure NC_CHAR points to S1
_nctonptype[NC_CHAR]='S1'
# internal C functions.
cdef _get_att_names(int grpid, int varid):
# Private function to get all the attribute names in a group
cdef int ierr, numatts, n
cdef char namstring[NC_MAX_NAME+1]
if varid == NC_GLOBAL:
ierr = nc_inq_natts(grpid, &numatts)
else:
ierr = nc_inq_varnatts(grpid, varid, &numatts)
if ierr != NC_NOERR:
raise RuntimeError((<char *>nc_strerror(ierr)).decode('ascii'))
attslist = []
for n from 0 <= n < numatts:
ierr = nc_inq_attname(grpid, varid, n, namstring)
if ierr != NC_NOERR:
raise RuntimeError((<char *>nc_strerror(ierr)).decode('ascii'))
attslist.append(namstring.decode(default_encoding,unicode_error))
return attslist
cdef _get_att(grp, int varid, name):
# Private function to get an attribute value given its name
cdef int ierr, n
cdef size_t att_len
cdef char *attname
cdef char *stratt
cdef nc_type att_type
cdef ndarray value_arr
bytestr = _strencode(name)
attname = bytestr
ierr = nc_inq_att(grp._grpid, varid, attname, &att_type, &att_len)
if ierr != NC_NOERR:
raise AttributeError((<char *>nc_strerror(ierr)).decode('ascii'))
# attribute is a character or string ...
if att_type == NC_CHAR:
value_arr = numpy.empty(att_len,'S1')
ierr = nc_get_att_text(grp._grpid, varid, attname, <char *>value_arr.data)
if ierr != NC_NOERR:
raise AttributeError((<char *>nc_strerror(ierr)).decode('ascii'))
pstring =\
value_arr.tostring().decode(default_encoding,unicode_error).replace('\x00','')
return pstring
elif att_type == NC_STRING:
if att_len == 1:
ierr = nc_get_att_string(grp._grpid, varid, attname, &stratt)
pstring = stratt.decode(default_encoding,unicode_error).replace('\x00','')
return pstring
else:
raise KeyError('vlen string array attributes not supported')
else:
# a regular numeric or compound type.
if att_type == NC_LONG:
att_type = NC_INT
try:
type_att = _nctonptype[att_type] # see if it is a primitive type
except KeyError:
# check if it's a compound
try:
type_att = _read_compound(grp, att_type)
except:
raise KeyError('attribute %s has unsupported datatype' % attname)
value_arr = numpy.empty(att_len,type_att)
ierr = nc_get_att(grp._grpid, varid, attname, value_arr.data)
if ierr != NC_NOERR:
raise AttributeError((<char *>nc_strerror(ierr)).decode('ascii'))
if value_arr.shape == ():
# return a scalar for a scalar array
return value_arr.item()
elif att_len == 1:
# return a scalar for a single element array
return value_arr[0]
else:
return value_arr
def _set_default_format(object format='NETCDF4'):
# Private function to set the netCDF file format
if format == 'NETCDF4':
nc_set_default_format(NC_FORMAT_NETCDF4, NULL)
elif format == 'NETCDF4_CLASSIC':
nc_set_default_format(NC_FORMAT_NETCDF4_CLASSIC, NULL)
elif format == 'NETCDF3_64BIT':
nc_set_default_format(NC_FORMAT_64BIT, NULL)
elif format == 'NETCDF3_CLASSIC':
nc_set_default_format(NC_FORMAT_CLASSIC, NULL)
else:
raise ValueError("format must be 'NETCDF4', 'NETCDF4_CLASSIC', 'NETCDF3_64BIT', or 'NETCDF3_CLASSIC', got '%s'" % format)
cdef _get_format(int grpid):
# Private function to get the netCDF file format
cdef int ierr, formatp
ierr = nc_inq_format(grpid, &formatp)
if ierr != NC_NOERR:
raise RuntimeError((<char *>nc_strerror(ierr)).decode('ascii'))
if formatp == NC_FORMAT_NETCDF4:
return 'NETCDF4'
elif formatp == NC_FORMAT_NETCDF4_CLASSIC:
return 'NETCDF4_CLASSIC'
elif formatp == NC_FORMAT_64BIT:
return 'NETCDF3_64BIT'
elif formatp == NC_FORMAT_CLASSIC:
return 'NETCDF3_CLASSIC'
cdef _get_full_format(int grpid):
# Private function to get the underlying disk format
cdef int ierr, formatp, modep
IF HAS_NC_INQ_FORMAT_EXTENDED:
ierr = nc_inq_format_extended(grpid, &formatp, &modep)
if ierr != NC_NOERR: