json-cfg-rw 0.5.0

Config file examples

A traditional json config file:

{
    "servers": [
        {
            "ip_address": "127.0.0.1",
            "port": 8080
        },
        {
            "ip_address": "127.0.0.1",
            "port": 8081
        }
    ],
    "superuser_name": "tron"
}

Something similar but better with json-cfg:

{
    // Note that we can get rid of most quotation marks.
    servers: [
        {
            ip_address: "127.0.0.1",
            port: 8080
        },
        // We have commented out the block of the second server below.
        // Trailing commas are allowed so the comma after the
        // first block (above) doesn't cause any problems.
        /*
        {
            ip_address: "127.0.0.1",
            port: 8081
        },  // <-- optional trailing comma
        /**/
    ],
    superuser_name: "tron",  // <-- optional trailing comma
}

Note that json-cfg can load both config files because standard json is a subset of the extended syntax allowed by json-cfg.

Installation

pip install json-cfg

Alternatively you can download the zipped library from https://pypi.python.org/pypi/json-cfg

Quick-starter

The json-cfg library provides two modes when it comes to loading config files: One that is very similar to the standard json.loads() and another one that returns the json wrapped into special config nodes that make handling the config file much easier:

• jsoncfg.load() and jsoncfg.loads() are very similar to the standard json.loads(). These functions allow you to load config files into bare python representation of the json data (dictionaries, lists, numbers, etc…).

• jsoncfg.load_config() and jsoncfg.loads_config() load the json data into special wrapper objects that help you to query the config with much nicer syntax. At the same time if you are looking for a value that doesn’t exist in the config then these problems are handled with exceptions that contain line/column number info about the location of the error.

One of the biggest problems with loading the config into bare python objects with a simple json library is that the loaded json data doesn’t contain the line/column numbers for the loaded json nodes/elements. This means that by using a simple json library you can report the location of errors with config file line/column numbers only in case of json syntax errors (in best case). By loading the json nodes/elements into our wrapper objects we can retain the line/column numbers for the json nodes/elements and we can use them in our error messages in case of semantic errors.

I assume that you have already installed json-cfg and you have the previously shown server config example in a server.cfg file in the current directory.

This is how to load and process the above server configuration with a simple json library:

import json

with open('server.cfg') as f:
    config = json.load(f)
for server in config['servers']:
    listen_on_interface(server['ip_address'], server.get('port', 8000))
superuser_name = config['superuser_name']

The same with json-cfg:

import jsoncfg

config = jsoncfg.load_config('server.cfg')
for server in config.servers:
    listen_on_interface(server.ip_address(), server.port(8000))
superuser_name = config.superuser_name()

Seemingly the difference isn’t that big. With json-cfg you can use extended syntax in the config file and the code that loads/processes the config is also somewhat nicer but real difference is what happens when you encounter an error. With json-cfg you get an exception with a message that points to the problematic part of the json config file while the pure-json example can’t tell you line/column numbers in the config file. In case of larger configs this can cause headaches.

Open your server.cfg file and remove the required ip_address attribute from one of the server config blocks. This will cause an error when we try to load the config file with the above code examples. The above code snippets report the following error messages in this scenario:

json:

KeyError: 'ip_address'

json-cfg:

jsoncfg.config_classes.JSONConfigValueNotFoundError: Required config node not found. Missing query path: .ip_address (relative to error location) [line=3;col=9]

Detailed explanation of the library interface

When you load your json with jsoncfg.load_config() or jsoncfg.loads_config() the returned json data - the hierarchy - is a tree of wrapper objects provided by this library. These wrapper objects make it possible to store the column/line numbers for each json node/element (for error reporting) and these wrappers allow you to query the config with the nice syntax you’ve seen above.

This library differentiates 3 types of json nodes/elements and each of these have their own wrapper classes:

• json object (dictionary like stuff)

• json array (list like stuff)

• json scalar (I use “scalar” to refer any json value that isn’t a container)

I use json value to refer to any json node/element whose type is unknown or unimportant. The public API of the wrapper classes is very simple: they have no public methods. All they provide is a few magic methods that you can use to read/query the loaded json data. (These magic methods are __contains__, __getattr__, __getitem__, __len__, __iter__ and __call__ but don’t worry if you don’t know about these magic methods as I will demonstrate the usage with simple code examples that don’t assume that you know them.) The reason for having no public methods is simple: We allow querying json object keys with __getattr__ (with the dot or member access operator like config.myvalue) and we don’t want any public methods to conflict with the key values in your config file.

After loading the config you have a tree of wrapper object nodes and you have to perform these two operations to get values from the config:

1. querying/reading/traversing the json hierarchy: the result of querying is a wrapper object

2. fetching the python value from the selected wrapper object: this can be done by calling the queried wrapper object.

The following sections explain these two operations in detail.

import jsoncfg

config = jsoncfg.load_config('server.cfg')

# Using __getattr__ to get the servers key from the config json object.
# The result of this expression is a wrapper object that wraps the servers array/list.
server_array = config.servers

# The equivalent of the previous expression using __getitem__:
server_array = config['servers']

# Note that querying a non-existing key from an object doesn't raise an error. Instead
# it returns a special ValueNotFoundNode instance that you can continue using as a
# wrapper object. The error happens only if you try to fetch the value of this key
# without specifying a default value - but more on this later in the section where we
# discuss value fetching from wrapper objects.
special_value_not_found_node = config.non_existing_key

# Checking whether a key exists in a json object:
servers_exists = 'servers' in config

# Using __getitem__ to index into json array wrapper objects:
# Over-indexing the array would raise an exception with useful error message
# containing the location of the servers_array in the config file.
first_item_wrapper_object = servers_array[0]

# Getting the length of json object and json array wrappers:
num_config_key_value_pairs = len(config)
servers_array_len = len(servers_array)

# Iterating the items of a json object or array:
for key_string, value_wrapper_object in config:
    pass
for value_wrapper_object in config.servers:
    pass

import jsoncfg

config = jsoncfg.load_config('server.cfg')

# Fetching the value of the whole json object hierarchy.
# python_hierarchy now looks like something you normally
# get as a result of a standard ``json.load()``.
python_hierarchy = config()

# Converting only the servers array into python-object format:
python_server_list = config.servers()

# Getting the ip_address of the first server.
server_0_ip_address_str = config.servers[0].ip_address()

# If "optional_value" isn't in the config then return the default value (50).
v0 = config.optional_value(50)

# This raises an exception if "required_value" isn't in the config.
v1 = config.required_value()

from jsoncfg.value_mappers import RequireType
from jsoncfg.value_mappers import require_list, require_string, require_integer, require_number

# require_list is a jsoncfg.JSONValueMapper instance that checks if the fetched value is a list.
# If the "servers" key is missing form the config or its type isn't list then an exception is
# raised because we haven't specified a default value.
python_server_list = config.servers(require_list)

# If the "servers" key is missing from the config then the return value is None. If "servers"
# is in the config and it isn't a list instance then an exception is raised otherwise the
# return value is the servers list.
python_server_list = config.servers(None, require_list)

# Querying the required ip_address parameter with required string type.
ip_address = config.servers[0].ip_address(require_string)

# Querying the optional port parameter with a default value of 8000.
# If the optional port parameter is specified in the config then it has to be an integer.
ip_address = config.servers[0].port(8000, require_integer)

# An optional timeout parameter with a default value of 5. If the timeout parameter is in
# the config then it has to be a number (int, long, or float).
timeout = config.timeout(5, require_number)

# Getting a required guest_name parameter from the config. The parameter has to be either
# None (null in the json file) or a string.
guest_name = config.guest_name(RequireType(type(None), str))

import datetime
import jsoncfg
from jsoncfg import JSONValueMapper
from jsoncfg.value_mappers import require_integer

class OneOf(JSONValueMapper):
    def __init__(self, *enum_members):
        self.enum_members = set(enum_members)

    def __call__(self, v):
        if v not in self.enum_members:
            raise ValueError('%r is not one of these: %r' % (v, self.enum_members))
        return v

class RangeCheck(JSONValueMapper):
    def __init__(self, min_, max_):
        self.min = min_
        self.max = max_

    def __call__(self, v):
        if self.min <= v < self.max:
            return v
        raise ValueError('%r is not in range [%r, %r)' % (v, self.min, self.max))

class ToDateTime(JSONValueMapper):
    def __call__(self, v):
        if not isinstance(v, str):
            raise TypeError('Expected a naive iso8601 datetime string but found %r' % (v,))
        return datetime.datetime.strptime(v, '%Y-%m-%dT%H:%M:%S')

config = jsoncfg.load_config('server.cfg')

# Creating a value mapper instance for reuse.
require_cool_superuser_name = OneOf('tron', 'neo')
superuser_name = config.superuser_name(None, require_cool_superuser_name)

check_http_port_range = RangeCheck(8000, 9000)
port = config.servers[0].port(8000, check_http_port_range)

# Chaining value mappers. First require_integer receives the value of the port
# attribute, checks/transforms it and the output of require_integer goes
# to the check_http_port_range value mapper. What you receive as a result of
# value fetching is the output of check_http_port_range.
port = config.servers[0].port(require_integer, check_http_port_range)

# to_datetime converts a naive iso8601 datetime string into a datetime instance.
to_datetime = ToDateTime()
superuser_birthday = config.superuser_birthday(None, to_datetime)

# TODO : Unit tests based on this example

import jsoncfg

server_file_name = 'server.cfg'

config = jsoncfg.load_config(server_file_name)

# Create some values and store them
arr = [1, 2, "three"]
config.example_arr = arr

# As "subobj1" doesn't exist, it will be created as a json object, and then "val1"
# will be assigned to it
val1 = "This is a string value"
config.subobj1.val1 = val1

# The assignment is clever enough to accept json compatible structures
obj = {"key1":1,
       "key2":"two",
       "subobj1": {} }
c.obj1 = obj

# Now that we've created some data, we can save it.
jsoncfg.save_config(server_file_name, config)

# Now lets load it up again and check that the values match
config = jsoncfg.load_config(server_file_name)
assert config.example_arr == arr
assert config.subobj1.val1 == val1
assert str(config.obj1) == str(obj) # This comparison may not work

# TODO : Unit tests based on this example

import jsoncfg

server_file_name = 'server.cfg'
obj = {"key1":1,
       "key2":"two",
       "subobj1": {} }
arr = [1, 2, "three"]
val1 = "This is a string value"

config = jsoncfg.ConfigWithWrapper('server.cfg')

with config as c:
    # This will store an array as a property of the root config object
    c.example_arr = arr

    # As "subobj1" doesn't exist, it will be created as a json object, and then "val1"
    # will be assigned to it
    c.subobj1.val1 = val1

    # The assignment is clever enough to accept json compatible structures
    c.obj1 = obj

# The config file is automatically saved when we left the with block

# Now lets load it up again and check that the values match
config = jsoncfg.load_config(server_file_name)
assert config.example_arr == arr
assert config.subobj1.val1 == val1
assert str(config.obj1) == str(obj) # This comparison may not work

Error handling: exceptions

The base of all library exceptions is jsoncfg.JSONConfigException. If the parsed json contains a syntax error then you receive a jsoncfg.JSONConfigParserException - this exception has no subclasses. In case of config query errors you receive a jsoncfg.JSONConfigQueryError - this exception has several subclasses.

                 +---------------------+
                 | JSONConfigException |
                 +---------------------+
                    ^               ^
                    |               |
+-------------------+-------+       |
| JSONConfigParserException |       |
+---------------------------+       |
                              +-----+----------------+
      +---------------------->| JSONConfigQueryError |<------------------------+
      |                       +----------------------+                         |
      |                          ^                ^                            |
      |                          |                |                            |
      |   +----------------------+-----+    +-----+------------------------+   |
      |   | JSONConfigValueMapperError |    | JSONConfigValueNotFoundError |   |
      |   +----------------------------+    +------------------------------+   |
      |                                                                        |
+-----+-------------------+                                   +----------------+-----+
| JSONConfigNodeTypeError |                                   | JSONConfigIndexError |
+-------------------------+                                   +----------------------+

jsoncfg.JSONConfigException

This is the mother of all exceptions raised by the library (aside from some some ValueError``s and ``TypeErrors that are raised in case of trivial programming mistakes). Note that this exception is never raised directly - the library raises only exceptions that are derived from this.

jsoncfg.JSONConfigParserException

You receive this exception if there is a syntax error in the parsed json.

• error_message: The error message without the line/column number info. The standard Exception.message field contains this very same message but with the line/column info formatted into it as a postfix.

• line, column: line and column information to locate the error easily in the parsed json.

jsoncfg.JSONConfigQueryError

You receive this exception in case of errors you make while processing the parsed json. This exception class is never instantiated directly, only its subclasses are used.

• config_node: The json node/element that was processed when the error happened.

• line, column: line and column information to locate the error easily in the parsed json.

jsoncfg.JSONConfigValueMapperError

Raised when you query and fetch a value by specifying a value mapper but the value mapper instance raises an exception during while fetching the value.

• mapper_exception: The exception instance raised by the value mapper.

jsoncfg.JSONConfigValueNotFoundError

This is raised when you try to fetch a required (non-optional) value that doesn’t exist in the config file.

jsoncfg.JSONConfigNodeTypeError

You get this exception if you try to perform an operation on a node that is not allowed for that node type (object, array or scalar), for example indexing into an array with a string.

jsoncfg.JSONConfigIndexError

Over-indexing a json array results in this exception.

• index: The index used to over-index the array.

Utility functions

The config wrapper objects have no public methods but in some cases you may want to extract some info from them (for example line/column number, type of node). You can do that with utility functions that can be imported from the jsoncfg module.

jsoncfg.node_location(config_node)

Returns the location of the specified config node in the file it was parsed from. The returned location is a named tuple NodeLocation(line, column) containing the 1-based line and column numbers.

jsoncfg.node_exists(config_node)

The library doesn’t raise an error if you query a non-existing key. It raises error only when you try to fetch a value from it. Querying a non-existing key returns a special ValueNotFoundNode instance and this function actually checks whether the node is something else than a ValueNotFoundNode instance. A node can be any part of the json: an object/dict, a list, or any other json value. Before trying to fetch a value from the queried node you can test the result of a query with node_exists() whether it is an existing or non-existing node in order to handle missing/optional config blocks gracefully without exceptions.

from jsoncfg import load_config, node_exists

config = load_config('my_config.cfg')
if node_exists(config.whatever1.whatever2.whatever3):
    ...

# OR an equivalent piece of code:

node = config.whatever1.whatever2.whatever3
if node_exists(node):
    ...

# This node_exists() call returns True:
exists_1 = node_exists(config.existing_key1.existing_key2.existing_key3)

# This node_exists() call returns False:
exists_2 = node_exists(config.non_existing_key1.non_existing_key2)

jsoncfg.node_is_object(config_node)

Returns True if the specified config_node is a json object/dict.

jsoncfg.node_is_array(config_node)

Returns True if the specified config_node is a json array/list.

jsoncfg.node_is_scalar(config_node)

Returns True if the specified config_node is a json value other than an object or array - if it isn’t a container.

jsoncfg.ensure_exists(config_node)

Returns the specified config_node if it is an existing node, otherwise it raises a config error (with config file location info when possible).

jsoncfg.expect_object(config_node)

Returns the specified config_node if it is a json object/dict, otherwise it raises a config error (with config file location info when possible).

In many cases you can just query and fetch objects using jsoncfg without doing explicit error handling and jsoncfg provides useful error messages when an error occurs (like trying the fetch the value from a non-existing node, trying to map a non-integer value to an integer, etc…). There is however at least one exception when jsoncfg can’t really auto-detect problems in a smart way: When you iterate over a json object or array. A json object returns (key, value) pairs during iteration while an array returns simple items. If you just assume (without actually checking) that a config node is a json object/dict and you iterate over it with auto-unpacking the returned (key, value) pairs into two variables then you might get into trouble if your assumption is incorrect and the actual config node is a json array. If it is an array then it will return simple items and python fails to unpack it into two variables. The result is an ugly python runtime error and not a nice jsoncfg error that says that the config node is an array and not an object/dict that your code expected. To overcome this problem you can use this jsoncfg.expect_object() function to ensure that the node you iterate is a json object. The same is recommended in case of json arrays: it is recommended to check them with jsoncfg.expect_array() before iteration:

from jsoncfg import load_config, expect_object, expect_array

config = load_config('server.cfg')
for server in expect_array(config.servers):
    print('------------')
    for key, value in expect_object(server):
        print('%s: %r' % (key, value))

jsoncfg.expect_array(config_node)

Returns the specified config_node if it is a json array/list, otherwise it raises a config error (with config file location info when possible).

jsoncfg.expect_scalar(config_node)

Returns the specified config_node if it isn’t a json object or array, otherwise it raises a config error (with config file location info when possible).