|
1 """ |
|
2 Implementation of JSONEncoder |
|
3 """ |
|
4 import re |
|
5 |
|
6 ESCAPE = re.compile(r'[\x00-\x19\\"\b\f\n\r\t]') |
|
7 ESCAPE_ASCII = re.compile(r'([\\"/]|[^\ -~])') |
|
8 ESCAPE_DCT = { |
|
9 # escape all forward slashes to prevent </script> attack |
|
10 '/': '\\/', |
|
11 '\\': '\\\\', |
|
12 '"': '\\"', |
|
13 '\b': '\\b', |
|
14 '\f': '\\f', |
|
15 '\n': '\\n', |
|
16 '\r': '\\r', |
|
17 '\t': '\\t', |
|
18 } |
|
19 for i in range(0x20): |
|
20 ESCAPE_DCT.setdefault(chr(i), '\\u%04x' % (i,)) |
|
21 |
|
22 # assume this produces an infinity on all machines (probably not guaranteed) |
|
23 INFINITY = float('1e66666') |
|
24 |
|
25 def floatstr(o, allow_nan=True): |
|
26 # Check for specials. Note that this type of test is processor- and/or |
|
27 # platform-specific, so do tests which don't depend on the internals. |
|
28 |
|
29 if o != o: |
|
30 text = 'NaN' |
|
31 elif o == INFINITY: |
|
32 text = 'Infinity' |
|
33 elif o == -INFINITY: |
|
34 text = '-Infinity' |
|
35 else: |
|
36 return str(o) |
|
37 |
|
38 if not allow_nan: |
|
39 raise ValueError("Out of range float values are not JSON compliant: %r" |
|
40 % (o,)) |
|
41 |
|
42 return text |
|
43 |
|
44 |
|
45 def encode_basestring(s): |
|
46 """ |
|
47 Return a JSON representation of a Python string |
|
48 """ |
|
49 def replace(match): |
|
50 return ESCAPE_DCT[match.group(0)] |
|
51 return '"' + ESCAPE.sub(replace, s) + '"' |
|
52 |
|
53 def encode_basestring_ascii(s): |
|
54 def replace(match): |
|
55 s = match.group(0) |
|
56 try: |
|
57 return ESCAPE_DCT[s] |
|
58 except KeyError: |
|
59 return '\\u%04x' % (ord(s),) |
|
60 return '"' + str(ESCAPE_ASCII.sub(replace, s)) + '"' |
|
61 |
|
62 |
|
63 class JSONEncoder(object): |
|
64 """ |
|
65 Extensible JSON <http://json.org> encoder for Python data structures. |
|
66 |
|
67 Supports the following objects and types by default: |
|
68 |
|
69 +-------------------+---------------+ |
|
70 | Python | JSON | |
|
71 +===================+===============+ |
|
72 | dict | object | |
|
73 +-------------------+---------------+ |
|
74 | list, tuple | array | |
|
75 +-------------------+---------------+ |
|
76 | str, unicode | string | |
|
77 +-------------------+---------------+ |
|
78 | int, long, float | number | |
|
79 +-------------------+---------------+ |
|
80 | True | true | |
|
81 +-------------------+---------------+ |
|
82 | False | false | |
|
83 +-------------------+---------------+ |
|
84 | None | null | |
|
85 +-------------------+---------------+ |
|
86 |
|
87 To extend this to recognize other objects, subclass and implement a |
|
88 ``.default()`` method with another method that returns a serializable |
|
89 object for ``o`` if possible, otherwise it should call the superclass |
|
90 implementation (to raise ``TypeError``). |
|
91 """ |
|
92 __all__ = ['__init__', 'default', 'encode', 'iterencode'] |
|
93 item_separator = ', ' |
|
94 key_separator = ': ' |
|
95 def __init__(self, skipkeys=False, ensure_ascii=True, |
|
96 check_circular=True, allow_nan=True, sort_keys=False, |
|
97 indent=None, separators=None): |
|
98 """ |
|
99 Constructor for JSONEncoder, with sensible defaults. |
|
100 |
|
101 If skipkeys is False, then it is a TypeError to attempt |
|
102 encoding of keys that are not str, int, long, float or None. If |
|
103 skipkeys is True, such items are simply skipped. |
|
104 |
|
105 If ensure_ascii is True, the output is guaranteed to be str |
|
106 objects with all incoming unicode characters escaped. If |
|
107 ensure_ascii is false, the output will be unicode object. |
|
108 |
|
109 If check_circular is True, then lists, dicts, and custom encoded |
|
110 objects will be checked for circular references during encoding to |
|
111 prevent an infinite recursion (which would cause an OverflowError). |
|
112 Otherwise, no such check takes place. |
|
113 |
|
114 If allow_nan is True, then NaN, Infinity, and -Infinity will be |
|
115 encoded as such. This behavior is not JSON specification compliant, |
|
116 but is consistent with most JavaScript based encoders and decoders. |
|
117 Otherwise, it will be a ValueError to encode such floats. |
|
118 |
|
119 If sort_keys is True, then the output of dictionaries will be |
|
120 sorted by key; this is useful for regression tests to ensure |
|
121 that JSON serializations can be compared on a day-to-day basis. |
|
122 |
|
123 If indent is a non-negative integer, then JSON array |
|
124 elements and object members will be pretty-printed with that |
|
125 indent level. An indent level of 0 will only insert newlines. |
|
126 None is the most compact representation. |
|
127 |
|
128 If specified, separators should be a (item_separator, key_separator) |
|
129 tuple. The default is (', ', ': '). To get the most compact JSON |
|
130 representation you should specify (',', ':') to eliminate whitespace. |
|
131 """ |
|
132 |
|
133 self.skipkeys = skipkeys |
|
134 self.ensure_ascii = ensure_ascii |
|
135 self.check_circular = check_circular |
|
136 self.allow_nan = allow_nan |
|
137 self.sort_keys = sort_keys |
|
138 self.indent = indent |
|
139 self.current_indent_level = 0 |
|
140 if separators is not None: |
|
141 self.item_separator, self.key_separator = separators |
|
142 |
|
143 def _newline_indent(self): |
|
144 return '\n' + (' ' * (self.indent * self.current_indent_level)) |
|
145 |
|
146 def _iterencode_list(self, lst, markers=None): |
|
147 if not lst: |
|
148 yield '[]' |
|
149 return |
|
150 if markers is not None: |
|
151 markerid = id(lst) |
|
152 if markerid in markers: |
|
153 raise ValueError("Circular reference detected") |
|
154 markers[markerid] = lst |
|
155 yield '[' |
|
156 if self.indent is not None: |
|
157 self.current_indent_level += 1 |
|
158 newline_indent = self._newline_indent() |
|
159 separator = self.item_separator + newline_indent |
|
160 yield newline_indent |
|
161 else: |
|
162 newline_indent = None |
|
163 separator = self.item_separator |
|
164 first = True |
|
165 for value in lst: |
|
166 if first: |
|
167 first = False |
|
168 else: |
|
169 yield separator |
|
170 for chunk in self._iterencode(value, markers): |
|
171 yield chunk |
|
172 if newline_indent is not None: |
|
173 self.current_indent_level -= 1 |
|
174 yield self._newline_indent() |
|
175 yield ']' |
|
176 if markers is not None: |
|
177 del markers[markerid] |
|
178 |
|
179 def _iterencode_dict(self, dct, markers=None): |
|
180 if not dct: |
|
181 yield '{}' |
|
182 return |
|
183 if markers is not None: |
|
184 markerid = id(dct) |
|
185 if markerid in markers: |
|
186 raise ValueError("Circular reference detected") |
|
187 markers[markerid] = dct |
|
188 yield '{' |
|
189 key_separator = self.key_separator |
|
190 if self.indent is not None: |
|
191 self.current_indent_level += 1 |
|
192 newline_indent = self._newline_indent() |
|
193 item_separator = self.item_separator + newline_indent |
|
194 yield newline_indent |
|
195 else: |
|
196 newline_indent = None |
|
197 item_separator = self.item_separator |
|
198 first = True |
|
199 if self.ensure_ascii: |
|
200 encoder = encode_basestring_ascii |
|
201 else: |
|
202 encoder = encode_basestring |
|
203 allow_nan = self.allow_nan |
|
204 if self.sort_keys: |
|
205 keys = dct.keys() |
|
206 keys.sort() |
|
207 items = [(k, dct[k]) for k in keys] |
|
208 else: |
|
209 items = dct.iteritems() |
|
210 for key, value in items: |
|
211 if isinstance(key, basestring): |
|
212 pass |
|
213 # JavaScript is weakly typed for these, so it makes sense to |
|
214 # also allow them. Many encoders seem to do something like this. |
|
215 elif isinstance(key, float): |
|
216 key = floatstr(key, allow_nan) |
|
217 elif isinstance(key, (int, long)): |
|
218 key = str(key) |
|
219 elif key is True: |
|
220 key = 'true' |
|
221 elif key is False: |
|
222 key = 'false' |
|
223 elif key is None: |
|
224 key = 'null' |
|
225 elif self.skipkeys: |
|
226 continue |
|
227 else: |
|
228 raise TypeError("key %r is not a string" % (key,)) |
|
229 if first: |
|
230 first = False |
|
231 else: |
|
232 yield item_separator |
|
233 yield encoder(key) |
|
234 yield key_separator |
|
235 for chunk in self._iterencode(value, markers): |
|
236 yield chunk |
|
237 if newline_indent is not None: |
|
238 self.current_indent_level -= 1 |
|
239 yield self._newline_indent() |
|
240 yield '}' |
|
241 if markers is not None: |
|
242 del markers[markerid] |
|
243 |
|
244 def _iterencode(self, o, markers=None): |
|
245 if isinstance(o, basestring): |
|
246 if self.ensure_ascii: |
|
247 encoder = encode_basestring_ascii |
|
248 else: |
|
249 encoder = encode_basestring |
|
250 yield encoder(o) |
|
251 elif o is None: |
|
252 yield 'null' |
|
253 elif o is True: |
|
254 yield 'true' |
|
255 elif o is False: |
|
256 yield 'false' |
|
257 elif isinstance(o, (int, long)): |
|
258 yield str(o) |
|
259 elif isinstance(o, float): |
|
260 yield floatstr(o, self.allow_nan) |
|
261 elif isinstance(o, (list, tuple)): |
|
262 for chunk in self._iterencode_list(o, markers): |
|
263 yield chunk |
|
264 elif isinstance(o, dict): |
|
265 for chunk in self._iterencode_dict(o, markers): |
|
266 yield chunk |
|
267 else: |
|
268 if markers is not None: |
|
269 markerid = id(o) |
|
270 if markerid in markers: |
|
271 raise ValueError("Circular reference detected") |
|
272 markers[markerid] = o |
|
273 for chunk in self._iterencode_default(o, markers): |
|
274 yield chunk |
|
275 if markers is not None: |
|
276 del markers[markerid] |
|
277 |
|
278 def _iterencode_default(self, o, markers=None): |
|
279 newobj = self.default(o) |
|
280 return self._iterencode(newobj, markers) |
|
281 |
|
282 def default(self, o): |
|
283 """ |
|
284 Implement this method in a subclass such that it returns |
|
285 a serializable object for ``o``, or calls the base implementation |
|
286 (to raise a ``TypeError``). |
|
287 |
|
288 For example, to support arbitrary iterators, you could |
|
289 implement default like this:: |
|
290 |
|
291 def default(self, o): |
|
292 try: |
|
293 iterable = iter(o) |
|
294 except TypeError: |
|
295 pass |
|
296 else: |
|
297 return list(iterable) |
|
298 return JSONEncoder.default(self, o) |
|
299 """ |
|
300 raise TypeError("%r is not JSON serializable" % (o,)) |
|
301 |
|
302 def encode(self, o): |
|
303 """ |
|
304 Return a JSON string representation of a Python data structure. |
|
305 |
|
306 >>> JSONEncoder().encode({"foo": ["bar", "baz"]}) |
|
307 '{"foo":["bar", "baz"]}' |
|
308 """ |
|
309 # This doesn't pass the iterator directly to ''.join() because it |
|
310 # sucks at reporting exceptions. It's going to do this internally |
|
311 # anyway because it uses PySequence_Fast or similar. |
|
312 chunks = list(self.iterencode(o)) |
|
313 return ''.join(chunks) |
|
314 |
|
315 def iterencode(self, o): |
|
316 """ |
|
317 Encode the given object and yield each string |
|
318 representation as available. |
|
319 |
|
320 For example:: |
|
321 |
|
322 for chunk in JSONEncoder().iterencode(bigobject): |
|
323 mysocket.write(chunk) |
|
324 """ |
|
325 if self.check_circular: |
|
326 markers = {} |
|
327 else: |
|
328 markers = None |
|
329 return self._iterencode(o, markers) |
|
330 |
|
331 __all__ = ['JSONEncoder'] |