1 The NetCDF NCZarr Implementation
2 ============================
3 <!-- double header is needed to workaround doxygen bug -->
5 # The NetCDF NCZarr Implementation {#nczarr_head}
9 # NCZarr Introduction {#nczarr_introduction}
11 Beginning with netCDF version 4.8.0, the Unidata NetCDF group has extended the netcdf-c library to provide access to cloud storage (e.g. Amazon S3 <a href="#ref_aws">[1]</a> ).
12 This extension provides a mapping from a subset of the full netCDF Enhanced (aka netCDF-4) data model to a variant of the Zarr <a href="#ref_zarrv2">[4]</a> data model.
13 The NetCDF version of this storage format is called NCZarr <a href="#ref_nczarr">[4]</a>.
15 A note on terminology in this document.
16 1. The term "dataset" is used to refer to all of the Zarr objects constituting
17 the meta-data and data.
19 There are some important "caveats" of which to be aware when using this software.
20 1. NCZarr currently is not thread-safe. So any attempt to use it with parallelism, including MPIO, is likely to fail.
22 # The NCZarr Data Model {#nczarr_data_model}
24 NCZarr uses a data model <a href="#ref_nczarr">[4]</a> that, by design, extends the Zarr Version 2 Specification <a href="#ref_zarrv2">[6]</a> to add support for the NetCDF-4 data model.
26 __Note Carefully__: a legal _NCZarr_ dataset is also a legal _Zarr_ dataset under a specific assumption. This assumption is that within Zarr meta-data objects, like ''.zarray'', unrecognized dictionary keys are ignored.
27 If this assumption is true of an implementation, then the _NCZarr_ dataset is a legal _Zarr_ dataset and should be readable by that _Zarr_ implementation.
28 The inverse is true also. A legal _Zarr_ dataset is also a legal _NCZarr_
29 dataset, where "legal" means it conforms to the Zarr version 2 specification.
30 In addition, certain non-Zarr features are allowed and used.
31 Specifically the XArray ''\_ARRAY\_DIMENSIONS'' attribute is one such.
33 There are two other, secondary assumption:
35 1. The actual storage format in which the dataset is stored -- a zip file, for example -- can be read by the _Zarr_ implementation.
36 2. The compressors (aka filters) used by the dataset can be encoded/decoded by the implementation. NCZarr uses HDF5-style filters, so ensuring access to such filters is somewhat complicated. See [the companion document on
37 filters](./md_filters.html "filters") for details.
39 Briefly, the data model supported by NCZarr is netcdf-4 minus
40 the user-defined types. However, a restricted form of String type
41 is supported (see Appendix H).
42 As with netcdf-4 chunking is supported. Filters and compression
43 are also [supported](./md_filters.html "filters").
45 Specifically, the model supports the following.
46 - "Atomic" types: char, byte, ubyte, short, ushort, int, uint, int64, uint64, string.
47 - Shared (named) dimensions
48 - Attributes with specified types -- both global and per-variable
52 - N-Dimensional variables
54 - Per-variable endianness (big or little)
55 - Filters (including compression)
57 With respect to full netCDF-4, the following concepts are
58 currently unsupported.
59 - User-defined types (enum, opaque, VLEN, and Compound)
60 - Unlimited dimensions
61 - Contiguous or compact storage
63 Note that contiguous and compact are not actually supported
64 because they are HDF5 specific.
65 When specified, they are treated as chunked where the file consists of only one chunk.
66 This means that testing for contiguous or compact is not possible; the _nc_inq_var_chunking_ function will always return NC_CHUNKED and the chunksizes will be the same as the dimension sizes of the variable's dimensions.
68 Additionally, it should be noted that NCZarr supports scalar variables,
69 but Zarr does not; Zarr only supports dimensioned variables.
70 In order to support interoperability, NCZarr does the following.
71 1. A scalar variable is recorded in the Zarr metadata as if it has a shape of **[1]**.
72 2. A note is stored in the NCZarr metadata that this is actually a netCDF scalar variable.
74 These actions allow NCZarr to properly show scalars in its API while still
75 maintaining compatibility with Zarr.
77 # Enabling NCZarr Support {#nczarr_enable}
79 NCZarr support is enabled by default.
80 If the _--disable-nczarr_ option is used with './configure', then NCZarr (and Zarr) support is disabled.
81 If NCZarr support is enabled, then support for datasets stored as files in a directory tree is provided as the only guaranteed mechanism for storing datasets.
82 However, several addition storage mechanisms are available if additional libraries are installed.
84 1. Zip format -- if _libzip_ is installed, then it is possible to directly read and write datasets stored in zip files.
85 2. If the AWS C++ SDK is installed, and _libcurl_ is installed, then it is possible to directly read and write datasets stored in the Amazon S3 cloud storage.
87 # Accessing Data Using the NCZarr Prototocol {#nczarr_accessing_data}
89 In order to access a NCZarr data source through the netCDF API, the file name normally used is replaced with a URL with a specific format.
90 Note specifically that there is no NC_NCZARR flag for the mode argument of _nc_create_ or _nc_open_.
91 In this case, it is indicated by the URL path.
94 The URL is the usual format.
96 scheme:://host:port/path?query#fragment format
98 There are some details that are important.
99 - Scheme: this should be _https_ or _s3_,or _file_.
100 The _s3_ scheme is equivalent
101 to "https" plus setting "mode=nczarr,s3" (see below).
102 Specifying "file" is mostly used for testing, but is used to support
103 directory tree or zipfile format storage.
104 - Host: Amazon S3 defines three forms: _Virtual_, _Path_, and _S3_
105 + _Virtual_: the host includes the bucket name as in
106 __bucket.s3.<region>.amazonaws.com__
107 + _Path_: the host does not include the bucket name, but
108 rather the bucket name is the first segment of the path.
109 For example __s3.<region>.amazonaws.com/bucket__
110 + _S3_: the protocol is "s3:" and if the host is a single name,
111 then it is interpreted as the bucket. The region is determined
112 using the algorithm in Appendix E.
113 + _Other_: It is possible to use other non-Amazon cloud storage, but
114 that is cloud library dependent.
115 - Query: currently not used.
116 - Fragment: the fragment is of the form _key=value&key=value&..._.
117 Depending on the key, the _value_ part may be left out and some
118 default value will be used.
122 The fragment part of a URL is used to specify information that is interpreted to specify what data format is to be used, as well as additional controls for that data format.
123 For NCZarr support, the following _key=value_ pairs are allowed.
125 - mode=nczarr|zarr|noxarray|file|zip|s3
127 Typically one will specify two mode flags: one to indicate what format
128 to use and one to specify the way the dataset is to be stored.
129 For example, a common one is "mode=zarr,file"
131 Using _mode=nczarr_ causes the URL to be interpreted as a
132 reference to a dataset that is stored in NCZarr format.
133 The _zarr_ mode tells the library to
134 use NCZarr, but to restrict its operation to operate on pure
135 Zarr Version 2 datasets.
137 The modes _s3_, _file_, and _zip_ tell the library what storage
139 * The _s3_ driver is the default and indicates using Amazon S3 or some equivalent.
140 * The _file_ format stores data in a directory tree.
141 * The _zip_ format stores data in a local zip file.
143 Note that It should be the case that zipping a _file_
144 format directory tree will produce a file readable by the
145 _zip_ storage format, and vice-versa.
147 By default, the XArray convention is supported and used for
148 both NCZarr files and pure Zarr files. This
149 means that every variable in the root group whose named dimensions
150 are also in the root group will have an attribute called
151 *\_ARRAY\_DIMENSIONS* that stores those dimension names.
152 The _noxarray_ mode tells the library to disable the XArray support.
154 The netcdf-c library is capable of inferring additional mode flags based on the flags it finds. Currently we have the following inferences.
157 So for example: ````...#mode=zarr,zip```` is equivalent to this.
158 ````...#mode=nczarr,zarr,zip
161 - log=<output-stream>: this control turns on logging output,
162 which is useful for debugging and testing.
163 If just _log_ is used
164 then it is equivalent to _log=stderr_.
167 # NCZarr Map Implementation {#nczarr_mapimpl}
169 Internally, the nczarr implementation has a map abstraction that allows different storage formats to be used.
170 This is closely patterned on the same approach used in the Python Zarr implementation, which relies on the Python _MutableMap_ <a href="#ref_python">[5]</a> class.
172 In NCZarr, the corresponding type is called _zmap_.
173 The __zmap__ API essentially implements a simplified variant
174 of the Amazon S3 API.
176 As with Amazon S3, __keys__ are utf8 strings with a specific structure:
177 that of a path similar to those of a Unix path with '/' as the
178 separator for the segments of the path.
180 As with Unix, all keys have this BNF syntax:
183 keypath: '/' segment | keypath '/' segment ;
184 segment: <sequence of UTF-8 characters except control characters and '/'>
186 Obviously, one can infer a tree structure from this key structure.
187 A containment relationship is defined by key prefixes.
188 Thus one key is "contained" (possibly transitively)
189 by another if one key is a prefix (in the string sense) of the other.
190 So in this sense the key "/x/y/z" is contained by the key "/x/y".
192 In this model all keys "exist" but only some keys refer to
193 objects containing content -- aka _content bearing_.
194 An important restriction is placed on the structure of the tree,
195 namely that keys are only defined for content-bearing objects.
196 Further, all the leaves of the tree are these content-bearing objects.
197 This means that the key for one content-bearing object should not
198 be a prefix of any other key.
200 There several other concepts of note.
201 1. __Dataset__ - a dataset is the complete tree contained by the key defining
202 the root of the dataset.
203 Technically, the root of the tree is the key <dataset>/.zgroup, where .zgroup can be considered the _superblock_ of the dataset.
204 2. __Object__ - equivalent of the S3 object; Each object has a unique key
205 and "contains" data in the form of an arbitrary sequence of 8-bit bytes.
207 The zmap API defined here isolates the key-value pair mapping
208 code from the Zarr-based implementation of NetCDF-4.
209 It wraps an internal C dispatch table manager for implementing an
210 abstract data structure implementing the zmap key/object model.
211 Of special note is the "search" function of the API.
213 __Search__: The search function has two purposes:
214 1. Support reading of pure zarr datasets (because they do not explicitly track their contents).
215 2. Debugging to allow raw examination of the storage. See zdump for example.
217 The search function takes a prefix path which has a key syntax (see above).
218 The set of legal keys is the set of keys such that the key references a content-bearing object -- e.g. /x/y/.zarray or /.zgroup.
219 Essentially this is the set of keys pointing to the leaf objects of the tree of keys constituting a dataset.
220 This set potentially limits the set of keys that need to be examined during search.
222 The search function returns a limited set of names, where the set of names are immediate suffixes of a given prefix path.
223 That is, if _<prefix>_ is the prefix path, then search returnsnall _<name>_ such that _<prefix>/<name>_ is itself a prefix of a "legal" key.
224 This can be used to implement glob style searches such as "/x/y/*" or "/x/y/**"
226 This semantics was chosen because it appears to be the minimum required to implement all other kinds of search using recursion.
227 It was also chosen to limit the number of names returned from the search.
229 1. Avoid returning keys that are not a prefix of some legal key.
230 2. Avoid returning all the legal keys in the dataset because that set may be very large; although the implementation may still have to examine all legal keys to get the desired subset.
231 3. Allow for use of partial read mechanisms such as iterators, if available.
232 This can support processing a limited set of keys for each iteration.
233 This is a straighforward tradeoff of space over time.
235 As a side note, S3 supports this kind of search using common prefixes with a delimiter of '/', although its use is a bit tricky.
236 For the file system zmap implementation, the legal search keys can be obtained one level at a time, which directly implements the search semantics.
237 For the zip file implementation, this semantics is not possible, so the whole
238 tree must be obtained and searched.
242 1. S3 limits key lengths to 1024 bytes.
243 Some deeply nested netcdf files will almost certainly exceed this limit.
244 2. Besides content, S3 objects can have an associated small set
245 of what may be called tags, which are themselves of the form of
246 key-value pairs, but where the key and value are always text.
247 As far as it is possible to determine, Zarr never uses these tags,
248 so they are not included in the zmap data structure.
250 __A Note on Error Codes:__
252 The zmap API returns some distinguished error code:
253 1. NC_NOERR if a operation succeeded
254 2. NC_EEMPTY is returned when accessing a key that has no content.
255 3. NC_EOBJECT is returned when an object is found which should not exist
256 4. NC_ENOOBJECT is returned when an object is not found which should exist
258 This does not preclude other errors being returned such NC_EACCESS or NC_EPERM or NC_EINVAL if there are permission errors or illegal function arguments, for example.
259 It also does not preclude the use of other error codes internal to the zmap implementation.
260 So zmap_file, for example, uses NC_ENOTFOUND internally because it is possible to detect the existence of directories and files.
261 But this does not propagate outside the zmap_file implementation.
263 ## Zmap Implementatons
265 The primary zmap implementation is _s3_ (i.e. _mode=nczarr,s3_) and indicates that the Amazon S3 cloud storage -- or some related applicance -- is to be used.
266 Another storage format uses a file system tree of directories and files (_mode=nczarr,file_).
267 A third storage format uses a zip file (_mode=nczarr,zip_).
268 The latter two are used mostly for debugging and testing.
269 However, the _file_ and _zip_ formats are important because they are intended to match corresponding storage formats used by the Python Zarr implementation.
270 Hence it should serve to provide interoperability between NCZarr and the Python Zarr, although this interoperability has not been tested.
272 Examples of the typical URL form for _file_ and _zip_ are as follows.
274 file:///xxx/yyy/testdata.file#mode=nczarr,file
275 file:///xxx/yyy/testdata.zip#mode=nczarr,zip
278 Note that the extension (e.g. ".file" in "testdata.file")
279 is arbitraty, so this would be equally acceptable.
281 file:///xxx/yyy/testdata.anyext#mode=nczarr,file
283 As with other URLS (e.g. DAP), these kind of URLS can be passed as the path argument to, for example, __ncdump__.
285 # NCZarr versus Pure Zarr. {#nczarr_purezarr}
287 The NCZARR format extends the pure Zarr format by adding extra keys such as ''\_NCZARR\_ARRAY'' inside the ''.zarray'' object.
288 It is possible to suppress the use of these extensions so that the netcdf library can read and write a pure zarr formatted file.
289 This is controlled by using ''mode=zarr'', which is an alias for the
290 ''mode=nczarr,zarr'' combination.
291 The primary effects of using pure zarr are described in the [Translation Section](@ref nczarr_translation).
293 There are some constraints on the reading of Zarr datasets using the NCZarr implementation.
295 1. Zarr allows some primitive types not recognized by NCZarr.
296 Over time, the set of unrecognized types is expected to diminish.
297 Examples of currently unsupported types are as follows:
298 * "c" -- complex floating point
301 2. The Zarr dataset may reference filters and compressors unrecognized by NCZarr.
302 3. The Zarr dataset may store data in column-major order instead of row-major order. The effect of encountering such a dataset is to output the data in the wrong order.
304 Again, this list should diminish over time.
306 # Notes on Debugging NCZarr Access {#nczarr_debug}
308 The NCZarr support has a trace facility.
309 Enabling this can sometimes give important, but voluminous information.
310 Tracing can be enabled by setting the environment variable NCTRACING=n,
311 where _n_ indicates the level of tracing.
312 A good value of _n_ is 9.
314 # Zip File Support {#nczarr_zip}
316 In order to use the _zip_ storage format, the libzip [3] library must be installed.
317 Note that this is different from zlib.
319 # Amazon S3 Storage {#nczarr_s3}
321 The Amazon AWS S3 storage driver currently uses the Amazon AWS S3 Software Development Kit for C++ (aws-s3-sdk-cpp).
322 In order to use it, the client must provide some configuration information.
323 Specifically, the ''~/.aws/config'' file should contain something like this.
328 aws_access_key_id=XXXX...
329 aws_secret_access_key=YYYY...
331 See Appendix E for additional information.
335 The notion of "addressing style" may need some expansion.
336 Amazon S3 accepts two forms for specifying the endpoint for accessing the data.
338 1. Virtual -- the virtual addressing style places the bucket in the host part of a URL.
341 https://<bucketname>.s2.<region>.amazonaws.com/
343 2. Path -- the path addressing style places the bucket in at the front of the path part of a URL.
346 https://s2.<region>.amazonaws.com/<bucketname>/
349 The NCZarr code will accept either form, although internally, it is standardized on path style.
350 The reason for this is that the bucket name forms the initial segment in the keys.
352 # Zarr vs NCZarr {#nczarr_vs_zarr}
356 The NCZarr storage format is almost identical to that of the the standard Zarr version 2 format.
357 The data model differs as follows.
359 1. Zarr only supports anonymous dimensions -- NCZarr supports only shared (named) dimensions.
360 2. Zarr attributes are untyped -- or perhaps more correctly characterized as of type string.
364 Consider both NCZarr and Zarr, and assume S3 notions of bucket and object.
365 In both systems, Groups and Variables (Array in Zarr) map to S3 objects.
366 Containment is modeled using the fact that the dataset's key is a prefix of the variable's key.
367 So for example, if variable _v1_ is contained in top level group g1 -- _/g1 -- then the key for _v1_ is _/g1/v_.
368 Additional meta-data information is stored in special objects whose name start with ".z".
370 In Zarr, the following special objects exist.
372 1. Information about a group is kept in a special object named _.zgroup_;
373 so for example the object _/g1/.zgroup_.
374 2. Information about an array is kept as a special object named _.zarray_;
375 so for example the object _/g1/v1/.zarray_.
376 3. Group-level attributes and variable-level attributes are stored in a special object named _.zattr_;
377 so for example the objects _/g1/.zattr_ and _/g1/v1/.zattr_.
378 4. Chunk data is stored in objects named "<n1>.<n2>...,<nr>" where the ni are positive integers representing the chunk index for the ith dimension.
380 The first three contain meta-data objects in the form of a string representing a JSON-formatted dictionary.
381 The NCZarr format uses the same objects as Zarr, but inserts NCZarr
382 specific key-value pairs in them to hold NCZarr specific information
383 The value of each of these keys is a JSON dictionary containing a variety
384 of NCZarr specific information.
386 These keys are as follows:
388 _\_nczarr_superblock\__ -- this is in the top level group -- key _/.zarr_.
389 It is in effect the "superblock" for the dataset and contains
390 any netcdf specific dataset level information.
391 It is also used to verify that a given key is the root of a dataset.
392 Currently it contains the following key(s):
393 * "version" -- the NCZarr version defining the format of the dataset.
395 _\_nczarr_group\__ -- this key appears in every _.zgroup_ object.
396 It contains any netcdf specific group information.
397 Specifically it contains the following keys:
398 * "dims" -- the name and size of shared dimensions defined in this group.
399 * "vars" -- the name of variables defined in this group.
400 * "groups" -- the name of sub-groups defined in this group.
401 These lists allow walking the NCZarr dataset without having to use the potentially costly search operation.
403 _\_nczarr_array\__ -- this key appears in every _.zarray_ object.
404 It contains netcdf specific array information.
405 Specifically it contains the following keys:
406 * dimrefs -- the names of the shared dimensions referenced by the variable.
407 * storage -- indicates if the variable is chunked vs contiguous in the netcdf sense.
409 _\_nczarr_attr\__ -- this key appears in every _.zattr_ object.
410 This means that technically, it is attribute, but one for which access
411 is normally surpressed .
412 Specifically it contains the following keys:
413 * types -- the types of all of the other attributes in the _.zattr_ object.
415 ## Translation {#nczarr_translation}
417 With some constraints, it is possible for an nczarr library to read Zarr and for a zarr library to read the nczarr format.
418 The latter case, zarr reading nczarr is possible if the zarr library is willing to ignore keys whose name it does not recognize; specifically anything beginning with _\_NCZARR\__.
420 The former case, nczarr reading zarr is also possible if the nczarr can simulate or infer the contents of the missing _\_NCZARR\_XXX_ objects.
421 As a rule this can be done as follows.
422 1. _\_nczarr_group\__ -- The list of contained variables and sub-groups can be computed using the search API to list the keys "contained" in the key for a group.
423 The search looks for occurrences of _.zgroup_, _.zattr_, _.zarray_ to infer the keys for the contained groups, attribute sets, and arrays (variables).
424 Constructing the set of "shared dimensions" is carried out
425 by walking all the variables in the whole dataset and collecting
426 the set of unique integer shapes for the variables.
427 For each such dimension length, a top level dimension is created
428 named ".zdim_<len>" where len is the integer length.
429 2. _\_nczarr_array\__ -- The dimrefs are inferred by using the shape
430 in _.zarray_ and creating references to the simulated shared dimension.
431 netcdf specific information.
432 3. _\_nczarr_attr\__ -- The type of each attribute is inferred by trying to parse the first attribute value string.
434 # Compatibility {#nczarr_compatibility}
436 In order to accomodate existing implementations, certain mode tags are provided to tell the NCZarr code to look for information used by specific implementations.
440 The Xarray <a href="#ref_xarray">[7]</a> Zarr implementation uses its own mechanism for specifying shared dimensions.
441 It uses a special attribute named ''_ARRAY_DIMENSIONS''.
442 The value of this attribute is a list of dimension names (strings).
443 An example might be ````["time", "lon", "lat"]````.
444 It is essentially equivalent to the ````_nczarr_array "dimrefs" list````, except that the latter uses fully qualified names so the referenced dimensions can be anywhere in the dataset.
446 As of _netcdf-c_ version 4.8.2, The Xarray ''_ARRAY_DIMENSIONS'' attribute is supported for both NCZarr and pure Zarr.
447 If possible, this attribute will be read/written by default,
448 but can be suppressed if the mode value "noxarray" is specified.
449 If detected, then these dimension names are used to define shared dimensions.
450 The following conditions will cause ''_ARRAY_DIMENSIONS'' to not be written.
451 * The variable is not in the root group,
452 * Any dimension referenced by the variable is not in the root group.
454 # Examples {#nczarr_examples}
456 Here are a couple of examples using the _ncgen_ and _ncdump_ utilities.
458 1. Create an nczarr file using a local directory tree as storage.
460 ncgen -4 -lb -o "file:///home/user/dataset.file#mode=nczarr,file" dataset.cdl
462 2. Display the content of an nczarr file using a zip file as storage.
464 ncdump "file:///home/user/dataset.zip#mode=nczarr,zip"
466 3. Create an nczarr file using S3 as storage.
468 ncgen -4 -lb -o "s3://s3.us-west-1.amazonaws.com/datasetbucket" dataset.cdl
470 4. Create an nczarr file using S3 as storage and keeping to the pure zarr format.
472 ncgen -4 -lb -o "s3://s3.uswest-1.amazonaws.com/datasetbucket#mode=zarr" dataset.cdl
474 5. Create an nczarr file using the s3 protocol with a specific profile
476 ncgen -4 -lb -o "s3://datasetbucket/rootkey#mode=nczarr,awsprofile=unidata" dataset.cdl
478 Note that the URLis internally translated to this
480 https://s2.<region>.amazonaws.com/datasetbucket/rootkey#mode=nczarr,awsprofile=unidata" dataset.cdl
482 The region is from the algorithm described in Appendix E1.
484 # References {#nczarr_bib}
486 <a name="ref_aws">[1]</a> [Amazon Simple Storage Service Documentation](https://docs.aws.amazon.com/s3/index.html)<br>
487 <a name="ref_awssdk">[2]</a> [Amazon Simple Storage Service Library](https://github.com/aws/aws-sdk-cpp)<br>
488 <a name="ref_libzip">[3]</a> [The LibZip Library](https://libzip.org/)<br>
489 <a name="ref_nczarr">[4]</a> [NetCDF ZARR Data Model Specification](https://www.unidata.ucar.edu/blogs/developer/en/entry/netcdf-zarr-data-model-specification)<br>
490 <a name="ref_python">[5]</a> [Python Documentation: 8.3.
491 collections — High-performance dataset datatypes](https://docs.python.org/2/library/collections.html)<br>
492 <a name="ref_zarrv2">[6]</a> [Zarr Version 2 Specification](https://zarr.readthedocs.io/en/stable/spec/v2.html)<br>
493 <a name="ref_xarray">[7]</a> [XArray Zarr Encoding Specification](http://xarray.pydata.org/en/latest/internals.html#zarr-encoding-specification)<br>
494 <a name="dynamic_filter_loading">[8]</a> [Dynamic Filter Loading](https://support.hdfgroup.org/HDF5/doc/Advanced/DynamicallyLoadedFilters/HDF5DynamicallyLoadedFilters.pdf)<br>
495 <a name="official_hdf5_filters">[9]</a> [Officially Registered Custom HDF5 Filters](https://portal.hdfgroup.org/display/support/Registered+Filter+Plugins)<br>
496 <a name="blosc-c-impl">[10]</a> [C-Blosc Compressor Implementation](https://github.com/Blosc/c-blosc)<br>
497 <a name="ref_awssdk_conda">[11]</a> [Conda-forge / packages / aws-sdk-cpp](https://anaconda.org/conda-forge/aws-sdk-cpp)<br>
498 <a name="ref_gdal">[12]</a> [GDAL Zarr](https://gdal.org/drivers/raster/zarr.html)<br>
500 # Appendix A. Building NCZarr Support {#nczarr_build}
502 Currently the following build cases are known to work.
505 <tr><td><u>Operating System</u><td><u>Build System</u><td><u>NCZarr</u><td><u>S3 Support</u>
506 <tr><td>Linux <td> Automake <td> yes <td> yes
507 <tr><td>Linux <td> CMake <td> yes <td> yes
508 <tr><td>Cygwin <td> Automake <td> yes <td> no
509 <tr><td>OSX <td> Automake <td> unknown <td> unknown
510 <tr><td>OSX <td> CMake <td> unknown <td> unknown
511 <tr><td>Visual Studio <td> CMake <td> yes <td> tests fail
514 Note: S3 support includes both compiling the S3 support code as well as running the S3 tests.
518 There are several options relevant to NCZarr support and to Amazon S3 support.
519 These are as follows.
521 1. _--disable-nczarr_ -- disable the NCZarr support.
522 If disabled, then all of the following options are disabled or irrelevant.
523 2. _--enable-nczarr-s3_ -- Enable NCZarr S3 support.
524 3. _--enable-nczarr-s3-tests_ -- the NCZarr S3 tests are currently only usable by Unidata personnel, so they are disabled by default.
526 __A note about using S3 with Automake.__
527 If S3 support is desired, and using Automake, then LDFLAGS must be properly set, namely to this.
529 LDFLAGS="$LDFLAGS -L/usr/local/lib -laws-cpp-sdk-s3"
531 The above assumes that these libraries were installed in '/usr/local/lib', so the above requires modification if they were installed elsewhere.
533 Note also that if S3 support is enabled, then you need to have a C++ compiler installed because part of the S3 support code is written in C++.
535 ## CMake {#nczarr_cmake}
537 The necessary CMake flags are as follows (with defaults)
539 1. -DENABLE_NCZARR=off -- equivalent to the Automake _--disable-nczarr_ option.
540 2. -DENABLE_NCZARR_S3=off -- equivalent to the Automake _--enable-nczarr-s3_ option.
541 3. -DENABLE_NCZARR_S3_TESTS=off -- equivalent to the Automake _--enable-nczarr-s3-tests_ option.
543 Note that unlike Automake, CMake can properly locate C++ libraries, so it should not be necessary to specify _-laws-cpp-sdk-s3_ assuming that the aws s3 libraries are installed in the default location.
544 For CMake with Visual Studio, the default location is here:
547 C:/Program Files (x86)/aws-cpp-sdk-all
550 It is possible to install the sdk library in another location.
551 In this case, one must add the following flag to the cmake command.
553 cmake ... -DAWSSDK_DIR=<awssdkdir>
555 where "awssdkdir" is the path to the sdk installation.
556 For example, this might be as follows.
558 cmake ... -DAWSSDK_DIR="c:\tools\aws-cpp-sdk-all"
560 This can be useful if blanks in path names cause problems in your build environment.
562 ## Testing S3 Support {#nczarr_testing_S3_support}
564 The relevant tests for S3 support are in the _nczarr_test_ directory.
565 Currently, by default, testing of S3 with NCZarr is supported only for Unidata members of the NetCDF Development Group.
566 This is because it uses a Unidata-specific bucket is inaccessible to the general user.
568 # Appendix B. Building aws-sdk-cpp {#nczarr_s3sdk}
570 In order to use the S3 storage driver, it is necessary to install the Amazon [aws-sdk-cpp library](https://github.com/aws/aws-sdk-cpp.git).
572 Building this package from scratch has proven to be a formidable task.
573 This appears to be due to dependencies on very specific versions of,
574 for example, openssl.
578 For linux, the following context works. Of course your mileage may vary.
580 * aws-sdk-cpp version 1.9.96 (or later?)
581 * Required installed libraries: openssl, libcurl, cmake, ninja (ninja-build in apt)
583 ### AWS-SDK-CPP Build Recipe
586 git clone --recurse-submodules https://www.github.com/aws/aws-sdk-cpp
591 FLAGS="-DCMAKE_INSTALL_PREFIX=${PREFIX} \
592 -DCMAKE_INSTALL_LIBDIR=lib \
593 -DCMAKE_MODULE_PATH=${PREFIX}/lib/cmake \
594 -DCMAKE_POLICY_DEFAULT_CMP0075=NEW \
596 -DENABLE_UNITY_BUILD=ON \
597 -DENABLE_TESTING=OFF \
598 -DCMAKE_BUILD_TYPE=$CFG \
600 cmake -GNinja $FLAGS ..
609 In order to build netcdf-c with S3 sdk support,
610 the following options must be specified for ./configure.
614 If you have access to the Unidata bucket on Amazon, then you can
615 also test S3 support with this option.
617 --enable-nczarr-s3-tests
621 It is possible to build and install aws-sdk-cpp. It is also possible
622 to build netcdf-c using cmake. Unfortunately, testing currently fails.
624 For Windows, the following context work. Of course your mileage may vary.
625 * OS: Windows 10 64-bit with Visual Studio community edition 2019.
626 * aws-sdk-cpp version 1.9.96 (or later?)
627 * Required installed libraries: openssl, libcurl, cmake
629 ### AWS-SDK-CPP Build Recipe
631 This command-line build assumes one is using Cygwin or Mingw to provide
635 git clone --recurse-submodules https://www.github.com/aws/aws-sdk-cpp
640 PREFIX="c:/tools/aws-sdk-cpp"
642 FLAGS="-DCMAKE_INSTALL_PREFIX=${PREFIX} \
643 -DCMAKE_INSTALL_LIBDIR=lib" \
644 -DCMAKE_MODULE_PATH=${PREFIX}/cmake \
645 -DCMAKE_POLICY_DEFAULT_CMP0075=NEW \
647 -DENABLE_UNITY_BUILD=ON \
648 -DCMAKE_BUILD_TYPE=$CFG \
654 cmake -DCMAKE_BUILD_TYPE=${CFG} $FLAGS ..
655 cmake --build . --config ${CFG}
656 cmake --install . --config ${CFG}
660 Notice that the sdk is being installed in the directory "c:\tools\aws-sdk-cpp"
661 rather than the default location "c:\Program Files (x86)/aws-sdk-cpp-all"
662 This is because when using a command line, an install path that contains
665 ### NetCDF CMake Build
667 Enabling S3 support is controlled by these two cmake options:
669 -DENABLE_NCZARR_S3=ON
670 -DENABLE_NCZARR_S3_TESTS=OFF
673 However, to find the aws sdk libraries,
674 the following environment variables must be set:
676 AWSSDK_ROOT_DIR="c:/tools/aws-sdk-cpp"
677 AWSSDKBIN="/cygdrive/c/tools/aws-sdk-cpp/bin"
678 PATH="$PATH:${AWSSDKBIN}"
680 Then the following options must be specified for cmake.
682 -DAWSSDK_ROOT_DIR=${AWSSDK_ROOT_DIR}
683 -DAWSSDK_DIR=${AWSSDK_ROOT_DIR}/lib/cmake/AWSSDK"
686 # Appendix C. Amazon S3 Imposed Limits {#nczarr_s3limits}
688 The Amazon S3 cloud storage imposes some significant limits that are inherited by NCZarr (and Zarr also, for that matter).
690 Some of the relevant limits are as follows:
691 1. The maximum object size is 5 Gigabytes with a total for all objects limited to 5 Terabytes.
692 2. S3 key names can be any UNICODE name with a maximum length of 1024 bytes.
693 Note that the limit is defined in terms of bytes and not (Unicode) characters.
694 This affects the depth to which groups can be nested because the key encodes the full path name of a group.
696 # Appendix D. Alternative Mechanisms for Accessing Remote Datasets {#nczarr_altremote}
698 The NetCDF-C library contains an alternate mechanism for accessing traditional netcdf-4 files stored in Amazon S3: The byte-range mechanism.
699 The idea is to treat the remote data as if it was a big file.
700 This remote "file" can be randomly accessed using the HTTP Byte-Range header.
702 In the Amazon S3 context, a copy of a dataset, a netcdf-3 or netdf-4 file, is uploaded into a single object in some bucket.
703 Then using the key to this object, it is possible to tell the netcdf-c library to treat the object as a remote file and to use the HTTP Byte-Range protocol to access the contents of the object.
704 The dataset object is referenced using a URL with the trailing fragment containing the string ````#mode=bytes````.
706 An examination of the test program _nc_test/test_byterange.sh_ shows simple examples using the _ncdump_ program.
707 One such test is specified as follows:
709 https://s3.us-east-1.amazonaws.com/noaa-goes16/ABI-L1b-RadC/2017/059/03/OR_ABI-L1b-RadC-M3C13_G16_s20170590337505_e20170590340289_c20170590340316.nc#mode=bytes
711 Note that for S3 access, it is expected that the URL is in what is called "path" format where the bucket, _noaa-goes16_ in this case, is part of the URL path instead of the host.
713 The _#mode=bytes_ mechanism generalizes to work with most servers that support byte-range access.
715 Specifically, Thredds servers support such access using the HttpServer access method as can be seen from this URL taken from the above test program.
717 https://thredds-test.unidata.ucar.edu/thredds/fileServer/irma/metar/files/METAR_20170910_0000.nc#bytes
720 # Appendix E. AWS Selection Algorithms. {#nczarr_awsselect}
722 If byterange support is enabled, the netcdf-c library will parse the files
726 ${HOME}/.aws/credentials
728 to extract profile names plus a list
729 of key=value pairs. This example is typical.
732 aws_access_key_id=XXXXXXXXXXXXXXXXXXXX
733 aws_secret_access_key=YYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYYY
736 The keys in the profile will be used to set various parameters in the library
740 The algorithm for choosing the active profile to use is as follows:
742 1. If the "aws.profile" fragment flag is defined in the URL, then it is used. For example, see this URL.
744 https://...#mode=nczarr,s3&aws.profile=xxx
746 2. If the "AWS.PROFILE" entry in the .rc file (i.e. .netrc or .dodsrc) is set, then it is used.
747 3. Otherwise the profile "default" is used.
749 The profile named "none" is a special profile that the netcdf-c library automatically defines.
750 It should not be defined anywhere else. It signals to the library that no credentialas are to used.
751 It is equivalent to the "--no-sign-request" option in the AWS CLI.
752 Also, it must be explicitly specified by name. Otherwise "default" will be used.
756 If the specified URL is of the form
760 Then this is rebuilt to this form:
762 s3://s2.<region>.amazonaws.com>/key
764 However this requires figuring out the region to use.
765 The algorithm for picking an region is as follows.
767 1. If the "aws.region" fragment flag is defined in the URL, then it is used.
768 2. The active profile is searched for the "aws_region" key.
769 3. If the "AWS.REGION" entry in the .rc file (i.e. .netrc or .dodsrc) is set, then it is used.
770 4. Otherwise use "us-east-1" region.
772 ## Authorization Selection
774 Picking an access-key/secret-key pair is always determined
775 by the current active profile. To choose to not use keys
776 requires that the active profile must be "none".
778 # Appendix F. NCZarr Version 1 Meta-Data Representation. {#nczarr_version1}
780 In NCZarr Version 1, the NCZarr specific metadata was represented using new objects rather than as keys in existing Zarr objects.
781 Due to conflicts with the Zarr specification, that format is deprecated in favor of the one described above.
782 However the netcdf-c NCZarr support can still read the version 1 format.
784 The version 1 format defines three specific objects: _.nczgroup_, _.nczarray_,_.nczattr_.
785 These are stored in parallel with the corresponding Zarr objects. So if there is a key of the form "/x/y/.zarray", then there is also a key "/x/y/.nczarray".
786 The content of these objects is the same as the contents of the corresponding keys. So the value of the ''_NCZARR_ARRAY'' key is the same as the content of the ''.nczarray'' object. The list of connections is as follows:
788 * ''.nczarr'' <=> ''_nczarr_superblock_''
789 * ''.nczgroup <=> ''_nczarr_group_''
790 * ''.nczarray <=> ''_nczarr_array_''
791 * ''.nczattr <=> ''_nczarr_attr_''
793 # Appendix G. JSON Attribute Convention. {#nczarr_json}
795 The Zarr V2 specification is somewhat vague on what is a legal
796 value for an attribute. The examples all show one of two cases:
797 1. A simple JSON scalar atomic values (e.g. int, float, char, etc), or
798 2. A JSON array of such values.
800 However, the Zarr specification can be read to infer that the value
801 can in fact be any legal JSON expression.
802 This "convention" is currently used routinely to help support various
803 attributes created by other packages where the attribute is a
804 complex JSON expression. An example is the GDAL Driver
805 convention <a href="#ref_gdal">[12]</a>, where the value is a complex
808 In order for NCZarr to be as consistent as possible with Zarr Version 2,
809 it is desirable to support this convention for attribute values.
810 This means that there must be some way to handle an attribute
811 whose value is not either of the two cases above. That is, its value
812 is some more complex JSON expression. Ideally both reading and writing
813 of such attributes should be supported.
815 One more point. NCZarr attempts to record the associated netcdf
816 attribute type (encoded in the form of a NumPy "dtype") for each
817 attribute. This information is stored as NCZarr-specific
818 metadata. Note that pure Zarr makes no attempt to record such
821 The current algorithm to support JSON valued attributes
824 ## Writing an attribute:
825 There are mutiple cases to consider.
827 1. The netcdf attribute **is not** of type NC_CHAR and its value is a single atomic value.
828 * Convert to an equivalent JSON atomic value and write that JSON expression.
829 * Compute the Zarr equivalent dtype and store in the NCZarr metadata.
831 2. The netcdf attribute **is not** of type NC_CHAR and its value is a vector of atomic values.
832 * Convert to an equivalent JSON array of atomic values and write that JSON expression.
833 * Compute the Zarr equivalent dtype and store in the NCZarr metadata.
835 3. The netcdf attribute **is** of type NC_CHAR and its value – taken as a single sequence of characters –
836 **is** parseable as a legal JSON expression.
837 * Parse to produce a JSON expression and write that expression.
838 * Use "|U1" as the dtype and store in the NCZarr metadata.
840 4. The netcdf attribute **is** of type NC_CHAR and its value – taken as a single sequence of characters –
841 **is not** parseable as a legal JSON expression.
842 * Convert to a JSON string and write that expression
843 * Use "|U1" as the dtype and store in the NCZarr metadata.
845 ## Reading an attribute:
847 The process of reading and interpreting an attribute value requires two
848 pieces of information.
849 * The value of the attribute as a JSON expression, and
850 * The optional associated dtype of the attribute; note that this may not exist
851 if, for example, the file is pure zarr.
853 Given these two pieces of information, the read process is as follows.
855 1. The JSON expression is a simple JSON atomic value.
856 * If the dtype is defined, then convert the JSON to that type of data,
857 and then store it as the equivalent netcdf vector of size one.
858 * If the dtype is not defined, then infer the dtype based on the the JSON value,
859 and then store it as the equivalent netcdf vector of size one.
861 2. The JSON expression is an array of simple JSON atomic values.
862 * If the dtype is defined, then convert each JSON value in the array to that type of data,
863 and then store it as the equivalent netcdf vector.
864 * If the dtype is not defined, then infer the dtype based on the first JSON value in the array,
865 and then store it as the equivalent netcdf vector.
867 3. The JSON expression is an array some of whose values are dictionaries or (sub-)arrays.
868 * Un-parse the expression to an equivalent sequence of characters, and then store it as of type NC_CHAR.
870 3. The JSON expression is a dictionary.
871 * Un-parse the expression to an equivalent sequence of characters, and then store it as of type NC_CHAR.
875 1. If a character valued attributes's value can be parsed as a legal JSON expression, then it will be stored as such.
876 2. Reading and writing are *almost* idempotent in that the sequence of
877 actions "read-write-read" is equivalent to a single "read" and "write-read-write" is equivalent to a single "write".
878 The "almost" caveat is necessary because (1) whitespace may be added or lost during the sequence of operations,
879 and (2) numeric precision may change.
881 # Appendix H. Support for string types
883 Zarr supports a string type, but it is restricted to
884 fixed size strings. NCZarr also supports such strings,
885 but there are some differences in order to interoperate
886 with the netcdf-4/HDF5 variable length strings.
888 The primary issue to be addressed is to provide a way for user
889 to specify the maximum size of the fixed length strings. This is
890 handled by providing the following new attributes:
891 1. **_nczarr_default_maxstrlen** —
892 This is an attribute of the root group. It specifies the default
893 maximum string length for string types. If not specified, then
894 it has the value of 128 characters.
895 2. **_nczarr_maxstrlen** —
896 This is a per-variable attribute. It specifies the maximum
897 string length for the string type associated with the variable.
898 If not specified, then it is assigned the value of
899 **_nczarr_default_maxstrlen**.
901 Note that when accessing a string through the netCDF API, the
902 fixed length strings appear as variable length strings. This
903 means that they are stored as pointers to the string
904 (i.e. **char\***) and with a trailing nul character.
905 One consequence is that if the user writes a variable length
906 string through the netCDF API, and the length of that string
907 is greater than the maximum string length for a variable,
908 then the string is silently truncated.
909 Another consequence is that the user must reclaim the string storage.
911 Adding strings also requires some hacking to handle the existing
912 netcdf-c NC_CHAR type, which does not exist in Zarr. The goal
913 was to choose NumPY types for both the netcdf-c NC_STRING type
914 and the netcdf-c NC_CHAR type such that if a pure zarr
915 implementation reads them, it will still work.
917 For writing variables and NCZarr attributes, the type mapping is as follows:
919 * "|S1" for NC_STRING && MAXSTRLEN==1
920 * "|Sn" for NC_STRING && MAXSTRLEN==n
922 Admittedly, this encoding is a bit of a hack.
924 So when reading data with a pure zarr implementaion
925 the above types should always appear as strings,
926 and the type that signals NC_CHAR (in NCZarr)
927 would be handled by Zarr as a string of length 1.
929 # Change Log {#nczarr_changelog}
931 Note, this log was only started as of 8/11/2022 and is not
932 intended to be a detailed chronology. Rather, it provides highlights
933 that will be of interest to NCZarr users. In order to see exact changes,
934 It is necessary to use the 'git diff' command.
937 1. Zarr fixed-size string types are now supported.
940 1. The NCZarr specific keys have been converted to lower-case
941 (e.g. "_nczarr_attr" instead of "_NCZARR_ATTR"). Upper case is
942 accepted for back compatibility.
944 2. The legal values of an attribute has been extended to
945 include arbitrary JSON expressions; see Appendix G for more details.
947 # Point of Contact {#nczarr_poc}
949 __Author__: Dennis Heimbigner<br>
950 __Email__: dmh at ucar dot edu<br>
951 __Initial Version__: 4/10/2020<br>
952 __Last Revised__: 8/27/2022