Issue 6552 - Support arbitrary manifest values

chainedconfigparser.py

 # This Source Code Form is subject to the terms of the Mozilla Public
 # License, v. 2.0. If a copy of the MPL was not distributed with this
 # file, You can obtain one at http://mozilla.org/MPL/2.0/.
 
 import os
 import io
 import ConfigParser
 from StringIO import StringIO
+from collections import OrderedDict
 
 
 class Item(tuple):
     def __new__(cls, name, value, source):
         result = super(Item, cls).__new__(cls, (name, value))
         result.source = source
         return result
 
 
 class DiffForUnknownOptionError(ConfigParser.Error):
@@ skipping 131 matching lines @@
 
     def option_source(self, section, option):
         option = self.optionxform(option)
         try:
             return self._origin[(section, option)]
         except KeyError:
             if not self.has_section(section):
                 raise ConfigParser.NoSectionError(section)
             raise ConfigParser.NoOptionError(option, section)
 
+    def as_json_object(self, section):

[Sebastian Noack, 2018/04/14 00:33:07]

Actually there isn't anything specific to JSON in this function. It just returns
a dictionary which then is converted to JSON in the template.

[tlucas, 2018/04/14 08:36:23]

You are right - what would you suggest? How about merge_to_nested()?

[Sebastian Noack, 2018/04/14 08:54:26]

Maybe section_as_dict()?

[tlucas, 2018/04/18 14:46:04]

Done.

+        r"""Parse a given section into a JSON object.
+
+        Parse arbitrary key/value pairs from 'section' of the current
+        configuration into a nested JSON object.
+
+        The following rules need to be considered:
+
+        * An option's key may be declared as a series of nested dictionary keys,
+          seperated by '.'.
+        * An option's key must end with '[]', to mark this option as an array

[Sebastian Noack, 2018/04/14 00:33:06]

Nit: Please make all bullet point end with punctation, for consistence.

[Sebastian Noack, 2018/04/14 00:33:07]

I just noticed, we could identify arrays by the presence of newlines, rather
than by adding [] to the key:

  foo = just a string

  foo =
    1st item of array
    2nd item of array

  foo =
    array with only one item

What do you think?

[tlucas, 2018/04/14 08:36:23]

The bullet points: ack

The newlines: Nice idea, will do

[tlucas, 2018/04/18 14:46:04]

Done.

+        * When an option is marked as an array, no other nested objects may
+          follow
+        * An array is expandable by the ConfigParser's '+=' token.
+        * Values may be marked as `number` or `bool` by prefixing them
+          accordingly:
+          * bool:<value>
+          * number:<value>
+
+        Example:
+
+                                        {
+                                          "good": false,
+                                          "bar": {
+                                            "baz": ["bar", "bazinga"]
+        foo = foo                         },
+        bar.baz[] = bar                   "is": {
+        baz.foo = a                         "integer": 1,
+        baz.z[] = b bool:true               "float": 1.4
+        bar.baz[] += bazinga    ==>       },
+        bad = bool:true                   "baz": {
+        good = bool:false                   "foo": "a",
+        is.integer = number:1               "z": ["b", true]
+        is.float = number:1.4             },
+                                          "bad": true
+                                          "foo": "foo"
+                                        }
+        """
+        def parse_values(v):
+            if isinstance(v, list):
+                return [parse_values(x) for x in v]
+
+            if v.startswith('number:'):
+                v = v.split(':', 1)[1]
+                try:
+                    v = int(v)
+                except ValueError:
+                    v = float(v)
+            elif v == 'bool:true':
+                v = True
+            elif v == 'bool:false':
+                v = False
+            return v
+
+        def setdefault_recursive(target, arr):
+            if len(arr) == 2:
+                target.setdefault(arr[0], parse_values(arr[1]))
+            else:
+                current = target.setdefault(arr[0], OrderedDict())
+                setdefault_recursive(current, arr[1:])
+
+        data = self.items(section)
+        result = OrderedDict()
+
+        for k, v in data:
+            if k.endswith('[]'):
+                k = k[:-2]
+                v = v.split()

[Sebastian Noack, 2018/04/14 00:33:06]

We might want to strip by newlines, so that array items can have spaces.

[tlucas, 2018/04/14 08:36:23]

Acknowledged.

[tlucas, 2018/04/18 14:46:04]

Done.

+
+            setdefault_recursive(result, k.split('.') + [v])

[Sebastian Noack, 2018/04/14 00:33:06]

It seems weird/unnecessary to merge the key segments and value into one list.
Any reason to not pass them as separate arguments?

[tlucas, 2018/04/14 08:36:23]

Not at all, will change.

[tlucas, 2018/04/18 14:46:04]

Done.

+
+        return result
+
     def readfp(self, fp, filename=None):
         raise NotImplementedError
 
     def set(self, section, option, value=None):
         raise NotImplementedError
 
     def add_section(self, section):
         raise NotImplementedError
 
     def remove_option(self, section, option):
         raise NotImplementedError
 
     def remove_section(self, section):
         raise NotImplementedError