PyangBind provides a set of helper classes which allow data to be loaded from, or serialised to a loaded data format. At the time of writing, the supported formats are:

  • IETF JSON - defined in draft-ietf-netmod-yang-json-08.
  • OpenConfig/PyangBind JSON - which does not currently have a published specification.

In the future, it is expected that an XML serialisation module may be required, given the current bias of devices towards this serialisation format.


Contents


Loading from JSON - Entire Module

The module pyangbind.lib.pyangbindJSON provides a wrapper around the encoder and decoder classes that are defined in pyangbind.lib.serialise. These methods are heavily biased towards loading an entire module into a new instance of the data tree.

The functions loads, load, loads_ietf and load_ietf are analagous to the Python json module’s functions. The loads.* functions load from a string which is expected to be valid JSON, whereas the load.* functions load from a file name that is specified to the modules.


Load Functions

loads(data, python_module, class_name) and loads_ietf(data, python_module, class_name)

The arguments to the loads functions are expected to be:
* data - a JSON-encoded string that can be loaded using Python’s json.load() function.
* python_module - the module within which the class that is to be instantiated is defined. In the case that -o <filename> is used, then this is filename as this code is a Python module that can be loaded with import. In the case that --split-class-dir <directory> has been used then it is directory.
* class_name - the name of the Python class that is found within module. This is the safe name (i.e., Python-safe - using no reserved Python keywords and substituting hyphens for underscores) of the YANG module that has been compiled - e.g., openconfig_bgp.

These functions return a PyangBind class instance with the data from data loaded into it.

load(filename, python_module, class_name) and load_ietf(filename, python_module, class_name)

The arguments to the load and load_ietf functions are identitical to those of the loads functions, other than the filename argument is a path to a file that can be loaded using open(filename, 'r').

These functions return a PyangBind class instance with the data from filename loaded.


Example Loading (Deserialisation)

With a simple module - such as the following:

 1module simple_serialise {
 2    yang-version "1";
 3    namespace "http://rob.sh/yang/test/ss";
 4    prefix "ss";
 5
 6    container a-container {
 7        leaf a-value {
 8            type int8;
 9        }
10    }
11
12    list a-list {
13        key "the-key";
14
15        leaf the-key {
16            type string;
17        }
18    }
19}

And bindings generated using:

$ pyang --plugindir $PYBINDPLUGIN -f pybind -o sbindings.py simple_serialise.yang

Loading OpenConfig/PyangBind JSON-encoded Data

An example instance of data for the above module encoded as PyangBind JSON looks like the following:

 1{
 2    "a-container": {
 3        "a-value": 8
 4    },
 5    "a-list": {
 6        "entry-one": {
 7            "the-key": "entry-one"
 8        },
 9        "entry-two": {
10            "the-key": "entry-two"
11        }
12    }
13}

To load this, using load and loads the following code can be used:

 1#!/usr/bin/env python
 2
 3import pprint
 4import pyangbind.lib.pybindJSON as pbJ
 5import sbindings
 6
 7pp = pprint.PrettyPrinter(indent=4)
 8
 9loaded_object = pbJ.load("json/simple-instance.json", sbindings, "simple_serialise")
10pp.pprint(loaded_object.get(filter=True))
11
12string_to_load = open('json/simple-instance.json', 'r').read().replace('\n', '')
13loaded_object_two = pbJ.loads(string_to_load, sbindings, "simple_serialise")
14pp.pprint(loaded_object_two.get(filter=True))

In both cases, the python_module name used is the name of the bindings file generated (sbindings) and the modulename is as specified in the YANG module (i.e., simple_serialise).

Both loading methods will return the same output:

python { 'a-container': { 'a-value': 8}, 'a-list': { u'entry-one': { 'the-key': u'entry-one'}, u'entry-two': { 'the-key': u'entry-two'}}} `


Loading IETF JSON-encoded Data

Loading IETF encoded data is almost identical, other than using the loads_ietf and load_ietf functions in place of loads and load respectively.

The corresponding example IETF-encoded JSON object for the above data is:

 1{
 2    "simple_serialise:a-container": {
 3        "a-value": 8
 4    },
 5    "simple_serialise:a-list": [
 6        {"the-key": "entry-one"},
 7        {"the-key": "entry-two"}
 8    ]
 9}

This can be loaded in the same way using the following Python:

 1# Load an instance from an IETF-JSON file
 2loaded_ietf_obj = pbJ.load_ietf("simple-instance-ietf.json", sbindings,
 3                                  "simple_serialise")
 4pp.pprint(loaded_ietf_obj.get(filter=True))
 5
 6# Load an instance from an IETF-JSON string
 7string_to_load = open('json/simple-instance-ietf.json', 'r').read().replace('\n',
 8                                                                         '')
 9loaded_ietf_obj_two = pbJ.loads_ietf(string_to_load, sbindings,
10                                                "simple_serialise")
11pp.pprint(loaded_ietf_obj_two.get(filter=True))

Again, the data loaded is idential to that shown above.


Loading Data into an Existing Instance

In a number of cases, it is desirable to load data from a serialised JSON input into an existing set of PyangBind classes - such as when accepting input from an external API. Doing this requires direct access of the pyangbind.lib.serialise classes, rather than the pyangbind.lib.pybindJSON helper functions. The relevant functions are those within pybindJSONDecoder - particularly load_ietf_json and load_json.

In order to use these functions (which are generally directly-called by the corresponding pyangbind.lib.pybindJSON functions - then there is a requirement to specify an already existing object, and skip the class instantiation stage of the loading functions.

When calling the load functions, the following format is expected to load into an existing object:

1load_json(input_data, None, None, obj=existing_object, path_helper=path_helper,
2                extmethods=extmethods, overwrite=overwrite):
3load_ietf_json(input_data, None, None, obj=existing_object, path_helper=path_helper,
4                                        extmethods=extmethods, overwrite=overwrite)

Where:
* input_data is a iterable object that corresponds to loaded JSON data.
* existing_object is the object that should be loaded into - i.e., an instantiated set of PyangBind classes.
* path_helper, extmethods - are the corresponding XPathHelper and extension methods that are to be used if required. In the case that these do not differ from the parent, they will be inherited.
* overwrite determines whether the existing instance’s data should be overwritten by the loaded data.


Example of Loading to Existing Instances

Using the same module as above, with the loaded instance in question (defining /a-list[the-key='entry-one'] and /a-list[the-key='entry-two'] then data can be loaded using the following load_json call:

1data_to_load = json.load(open('json/simple-instance-additional.json','r'))
2pybindJSONDecoder.load_json(data_to_load, None, None, obj=existing_instance)

The existing_instance object is modified in-place, and hence the list acquires the additional entry-three data defined in the simple-instance-additional.json file:

1pp.pprint(existing_instance.a_list.keys())
2# Outputs:
3# [u'entry-two', u'entry-three', u'entry-one']

Serialising Data into JSON

In order to serialise a PyangBind class instance into JSON, the dump, dumps functions defined in pyangbind.lib.pybindJSON are used. These functions take a mode keyword argument which determines whether they dump IETF-specified JSON or PyangBind JSON.

dump(obj, filename, indent=<int>, filter=<bool>, skip_subtrees=<list>, mode="default")

  • obj - which is a PyangBind class instance that is to be dumped. It is expected to havea get method, hence be a list or a container.
  • filename - the file to which the JSON should be written.
  • indent - the number of spaces to use to indent the JSON.
  • filter - analagous to the filter argument to a PyangBind class’ get method (see the documentation relating to generic methods), which determines whether the entire tree, or just the changed elements are to be dumped.
  • skip_subtrees - a list of paths (absolute rather than relative) that should be pruned from the output JSON. This is useful if multiple output files are used to save data instances.
  • mode - a string specifying the JSON encoding to be output – currently either “default” or “ietf”.

dumps(obj, indent=4, filter=True, skip_subtrees=[], select=False, mode="default")
* The obj, indent, filter, skip_subtrees and mode arguments of dumps are as per dump.
* select - when provided is expected to be a dictionary of the form { "element_name": value }. When this is specified only elements where obj.element_name == value are output. This is useful when using query parameters to select subsets of an object, or list.


Example Serialisation

In order to serialise an instance of the simple_serialise module used above - the following call is used:

1pyangbindJSON.dumps(existing_instance, filter=True)

This outputs the entire module as JSON:

 1{
 2    "a-list": {
 3        "entry-two": {
 4            "the-key": "entry-two"
 5        }, 
 6        "entry-three": {
 7            "the-key": "entry-three"
 8        }, 
 9        "entry-one": {
10            "the-key": "entry-one"
11        }
12    }
13}

The JSON format can be switched to IETF-encoded JSON by using the mode="ietf" argument to dumps.

To select an entry from the list where the the-key leaf is equal to “entry-one” (although this is a gratiutous example), the select dictionary can be used:

1pbJ.dumps(existing_instance.a_list, select={'the-key': 'entry-one'}, mode="ietf")

This outputs only the entry-one output of the list being shown:

1[
2    {
3        "simple_serialise:the-key": "entry-one"
4    }
5]

Example Code

The example used throughout this document is included under docs/example/simple-serialise.