Show More
@@ -1,185 +1,200 b'' | |||
|
1 | 1 | """Utilities to manipulate JSON objects. |
|
2 | 2 | """ |
|
3 | 3 | #----------------------------------------------------------------------------- |
|
4 | 4 | # Copyright (C) 2010-2011 The IPython Development Team |
|
5 | 5 | # |
|
6 | 6 | # Distributed under the terms of the BSD License. The full license is in |
|
7 | 7 | # the file COPYING.txt, distributed as part of this software. |
|
8 | 8 | #----------------------------------------------------------------------------- |
|
9 | 9 | |
|
10 | 10 | #----------------------------------------------------------------------------- |
|
11 | 11 | # Imports |
|
12 | 12 | #----------------------------------------------------------------------------- |
|
13 | 13 | # stdlib |
|
14 | 14 | import re |
|
15 | 15 | import sys |
|
16 | 16 | import types |
|
17 | 17 | from base64 import encodestring |
|
18 | 18 | from datetime import datetime |
|
19 | 19 | |
|
20 | 20 | from IPython.utils import py3compat |
|
21 | 21 | from IPython.utils.encoding import DEFAULT_ENCODING |
|
22 | 22 | from IPython.utils import text |
|
23 | 23 | next_attr_name = '__next__' if py3compat.PY3 else 'next' |
|
24 | 24 | |
|
25 | 25 | #----------------------------------------------------------------------------- |
|
26 | 26 | # Globals and constants |
|
27 | 27 | #----------------------------------------------------------------------------- |
|
28 | 28 | |
|
29 | 29 | # timestamp formats |
|
30 | 30 | ISO8601="%Y-%m-%dT%H:%M:%S.%f" |
|
31 | 31 | ISO8601_PAT=re.compile(r"^\d{4}-\d{2}-\d{2}T\d{2}:\d{2}:\d{2}\.\d+$") |
|
32 | 32 | |
|
33 | 33 | #----------------------------------------------------------------------------- |
|
34 | 34 | # Classes and functions |
|
35 | 35 | #----------------------------------------------------------------------------- |
|
36 | 36 | |
|
37 | 37 | def rekey(dikt): |
|
38 | 38 | """Rekey a dict that has been forced to use str keys where there should be |
|
39 | 39 | ints by json.""" |
|
40 | 40 | for k in dikt.iterkeys(): |
|
41 | 41 | if isinstance(k, basestring): |
|
42 | 42 | ik=fk=None |
|
43 | 43 | try: |
|
44 | 44 | ik = int(k) |
|
45 | 45 | except ValueError: |
|
46 | 46 | try: |
|
47 | 47 | fk = float(k) |
|
48 | 48 | except ValueError: |
|
49 | 49 | continue |
|
50 | 50 | if ik is not None: |
|
51 | 51 | nk = ik |
|
52 | 52 | else: |
|
53 | 53 | nk = fk |
|
54 | 54 | if nk in dikt: |
|
55 | 55 | raise KeyError("already have key %r"%nk) |
|
56 | 56 | dikt[nk] = dikt.pop(k) |
|
57 | 57 | return dikt |
|
58 | 58 | |
|
59 | 59 | |
|
60 | 60 | def extract_dates(obj): |
|
61 | 61 | """extract ISO8601 dates from unpacked JSON""" |
|
62 | 62 | if isinstance(obj, dict): |
|
63 | 63 | obj = dict(obj) # don't clobber |
|
64 | 64 | for k,v in obj.iteritems(): |
|
65 | 65 | obj[k] = extract_dates(v) |
|
66 | 66 | elif isinstance(obj, (list, tuple)): |
|
67 | 67 | obj = [ extract_dates(o) for o in obj ] |
|
68 | 68 | elif isinstance(obj, basestring): |
|
69 | 69 | if ISO8601_PAT.match(obj): |
|
70 | 70 | obj = datetime.strptime(obj, ISO8601) |
|
71 | 71 | return obj |
|
72 | 72 | |
|
73 | 73 | def squash_dates(obj): |
|
74 | 74 | """squash datetime objects into ISO8601 strings""" |
|
75 | 75 | if isinstance(obj, dict): |
|
76 | 76 | obj = dict(obj) # don't clobber |
|
77 | 77 | for k,v in obj.iteritems(): |
|
78 | 78 | obj[k] = squash_dates(v) |
|
79 | 79 | elif isinstance(obj, (list, tuple)): |
|
80 | 80 | obj = [ squash_dates(o) for o in obj ] |
|
81 | 81 | elif isinstance(obj, datetime): |
|
82 | 82 | obj = obj.strftime(ISO8601) |
|
83 | 83 | return obj |
|
84 | 84 | |
|
85 | 85 | def date_default(obj): |
|
86 | 86 | """default function for packing datetime objects in JSON.""" |
|
87 | 87 | if isinstance(obj, datetime): |
|
88 | 88 | return obj.strftime(ISO8601) |
|
89 | 89 | else: |
|
90 | 90 | raise TypeError("%r is not JSON serializable"%obj) |
|
91 | 91 | |
|
92 | 92 | |
|
93 | 93 | # constants for identifying png/jpeg data |
|
94 | 94 | PNG = b'\x89PNG\r\n\x1a\n' |
|
95 | 95 | JPEG = b'\xff\xd8' |
|
96 | 96 | |
|
97 | 97 | def encode_images(format_dict): |
|
98 | 98 | """b64-encodes images in a displaypub format dict |
|
99 | 99 | |
|
100 | 100 | Perhaps this should be handled in json_clean itself? |
|
101 | ||
|
102 | Parameters | |
|
103 | ---------- | |
|
104 | ||
|
105 | format_dict : dict | |
|
106 | A dictionary of display data keyed by mime-type | |
|
107 | ||
|
108 | Returns | |
|
109 | ------- | |
|
110 | ||
|
111 | format_dict : dict | |
|
112 | A copy of the same dictionary, | |
|
113 | but binary image data ('image/png' or 'image/jpeg') | |
|
114 | is base64-encoded. | |
|
115 | ||
|
101 | 116 | """ |
|
102 | 117 | encoded = format_dict.copy() |
|
103 | 118 | pngdata = format_dict.get('image/png') |
|
104 | 119 | if isinstance(pngdata, bytes) and pngdata[:8] == PNG: |
|
105 | 120 | encoded['image/png'] = encodestring(pngdata).decode('ascii') |
|
106 | 121 | jpegdata = format_dict.get('image/jpeg') |
|
107 | 122 | if isinstance(jpegdata, bytes) and jpegdata[:2] == JPEG: |
|
108 | 123 | encoded['image/jpeg'] = encodestring(jpegdata).decode('ascii') |
|
109 | 124 | return encoded |
|
110 | 125 | |
|
111 | 126 | |
|
112 | 127 | def json_clean(obj): |
|
113 | 128 | """Clean an object to ensure it's safe to encode in JSON. |
|
114 | 129 | |
|
115 | 130 | Atomic, immutable objects are returned unmodified. Sets and tuples are |
|
116 | 131 | converted to lists, lists are copied and dicts are also copied. |
|
117 | 132 | |
|
118 | 133 | Note: dicts whose keys could cause collisions upon encoding (such as a dict |
|
119 | 134 | with both the number 1 and the string '1' as keys) will cause a ValueError |
|
120 | 135 | to be raised. |
|
121 | 136 | |
|
122 | 137 | Parameters |
|
123 | 138 | ---------- |
|
124 | 139 | obj : any python object |
|
125 | 140 | |
|
126 | 141 | Returns |
|
127 | 142 | ------- |
|
128 | 143 | out : object |
|
129 | 144 | |
|
130 | 145 | A version of the input which will not cause an encoding error when |
|
131 | 146 | encoded as JSON. Note that this function does not *encode* its inputs, |
|
132 | 147 | it simply sanitizes it so that there will be no encoding errors later. |
|
133 | 148 | |
|
134 | 149 | Examples |
|
135 | 150 | -------- |
|
136 | 151 | >>> json_clean(4) |
|
137 | 152 | 4 |
|
138 | 153 | >>> json_clean(range(10)) |
|
139 | 154 | [0, 1, 2, 3, 4, 5, 6, 7, 8, 9] |
|
140 | 155 | >>> sorted(json_clean(dict(x=1, y=2)).items()) |
|
141 | 156 | [('x', 1), ('y', 2)] |
|
142 | 157 | >>> sorted(json_clean(dict(x=1, y=2, z=[1,2,3])).items()) |
|
143 | 158 | [('x', 1), ('y', 2), ('z', [1, 2, 3])] |
|
144 | 159 | >>> json_clean(True) |
|
145 | 160 | True |
|
146 | 161 | """ |
|
147 | 162 | # types that are 'atomic' and ok in json as-is. bool doesn't need to be |
|
148 | 163 | # listed explicitly because bools pass as int instances |
|
149 | 164 | atomic_ok = (unicode, int, float, types.NoneType) |
|
150 | 165 | |
|
151 | 166 | # containers that we need to convert into lists |
|
152 | 167 | container_to_list = (tuple, set, types.GeneratorType) |
|
153 | 168 | |
|
154 | 169 | if isinstance(obj, atomic_ok): |
|
155 | 170 | return obj |
|
156 | 171 | |
|
157 | 172 | if isinstance(obj, bytes): |
|
158 | 173 | return obj.decode(DEFAULT_ENCODING, 'replace') |
|
159 | 174 | |
|
160 | 175 | if isinstance(obj, container_to_list) or ( |
|
161 | 176 | hasattr(obj, '__iter__') and hasattr(obj, next_attr_name)): |
|
162 | 177 | obj = list(obj) |
|
163 | 178 | |
|
164 | 179 | if isinstance(obj, list): |
|
165 | 180 | return [json_clean(x) for x in obj] |
|
166 | 181 | |
|
167 | 182 | if isinstance(obj, dict): |
|
168 | 183 | # First, validate that the dict won't lose data in conversion due to |
|
169 | 184 | # key collisions after stringification. This can happen with keys like |
|
170 | 185 | # True and 'true' or 1 and '1', which collide in JSON. |
|
171 | 186 | nkeys = len(obj) |
|
172 | 187 | nkeys_collapsed = len(set(map(str, obj))) |
|
173 | 188 | if nkeys != nkeys_collapsed: |
|
174 | 189 | raise ValueError('dict can not be safely converted to JSON: ' |
|
175 | 190 | 'key collision would lead to dropped values') |
|
176 | 191 | # If all OK, proceed by making the new dict that will be json-safe |
|
177 | 192 | out = {} |
|
178 | 193 | for k,v in obj.iteritems(): |
|
179 | 194 | out[str(k)] = json_clean(v) |
|
180 | 195 | return out |
|
181 | 196 | |
|
182 | 197 | # If we get here, we don't know how to handle the object, so we just get |
|
183 | 198 | # its repr and return that. This will catch lambdas, open sockets, class |
|
184 | 199 | # objects, and any other complicated contraption that json can't encode |
|
185 | 200 | return repr(obj) |
General Comments 0
You need to be logged in to leave comments.
Login now