比“无法解码JSON对象”显示更好的错误消息

Python code to load data from some long complicated JSON file:

with open(filename, "r") as f:
  data = json.loads(f.read())

(note: the best code version should be:

with open(filename, "r") as f:
  data = json.load(f)

but both exhibit similar behavior)

For many types of JSON error (missing delimiters, incorrect backslashes in strings, etc), this prints a nice helpful message containing the line and column number where the JSON error was found.

However, for other types of JSON error (including the classic “using comma on the last item in a list”, but also other things like capitalising true/false), Python’s output is just:

Traceback (most recent call last):
  File "myfile.py", line 8, in myfunction
    config = json.loads(f.read())
  File "c:\python27\lib\json\__init__.py", line 326, in loads
    return _default_decoder.decode(s)
  File "c:\python27\lib\json\decoder.py", line 360, in decode
    obj, end = self.raw_decode(s, idx=_w(s, 0).end())
  File "c:\python27\lib\json\decoder.py", line 378, in raw_decode
    raise ValueError("No JSON object could be decoded")
ValueError: No JSON object could be decoded

For that type of ValueError, how do you get Python to tell you where is the error in the JSON file?

I’ve found that the simplejson module gives more descriptive errors in many cases where the built-in json module is vague. For instance, for the case of having a comma after the last item in a list:

json.loads('[1,2,]')
....
ValueError: No JSON object could be decoded

which is not very descriptive. The same operation with simplejson:

simplejson.loads('[1,2,]')
...
simplejson.decoder.JSONDecodeError: Expecting object: line 1 column 5 (char 5)

Much better! Likewise for other common errors like capitalizing True.

You wont be able to get python to tell you where the JSON is incorrect. You will need to use a linter online somewhere like this

This will show you error in the JSON you are trying to decode.

You could try the rson library found here: http://code.google.com/p/rson/ . I it also up on PYPI: https://pypi.python.org/pypi/rson/0.9 so you can use easy_install or pip to get it.

for the example given by tom:

>>> rson.loads('[1,2,]')
...
rson.base.tokenizer.RSONDecodeError: Unexpected trailing comma: line 1, column 6, text ']'

RSON is a designed to be a superset of JSON, so it can parse JSON files. It also has an alternate syntax which is much nicer for humans to look at and edit. I use it quite a bit for input files.

As for the capitalizing of boolean values: it appears that rson reads incorrectly capitalized booleans as strings.

>>> rson.loads('[true,False]')
[True, u'False']

I had a similar problem and it was due to singlequotes. The JSON standard(http://json.org) talks only about using double quotes so it must be that the python json library supports only double quotes.

For my particular version of this problem, I went ahead and searched the function declaration of load_json_file(path) within the packaging.py file, then smuggled a print line into it:

def load_json_file(path):
    data = open(path, 'r').read()
    print data
    try:
        return Bunch(json.loads(data))
    except ValueError, e:
        raise MalformedJsonFileError('%s when reading "%s"' % (str(e),
                                                               path))

That way it would print the content of the json file before entering the try-catch, and that way – even with my barely existing Python knowledge – I was able to quickly figure out why my configuration couldn’t read the json file.
(It was because I had set up my text editor to write a UTF-8 BOM … stupid)

Just mentioning this because, while maybe not a good answer to the OP’s specific problem, this was a rather quick method in determining the source of a very oppressing bug. And I bet that many people will stumble upon this article who are searching a more verbose solution for a MalformedJsonFileError: No JSON object could be decoded when reading …. So that might help them.

As to me, my json file is very large, when use common json in python it gets the above error.

After install simplejson by sudo pip install simplejson.

And then I solved it.

import json
import simplejson


def test_parse_json():
    f_path = '/home/hello/_data.json'
    with open(f_path) as f:
        # j_data = json.load(f)      # ValueError: No JSON object could be decoded
        j_data = simplejson.load(f)  # right
    lst_img = j_data['images']['image']
    print lst_img[0]


if __name__ == '__main__':
    test_parse_json()

I had a similar problem this was my code:

    json_file=json.dumps(pyJson)
    file = open("list.json",'w')
    file.write(json_file)  

    json_file = open("list.json","r")
    json_decoded = json.load(json_file)
    print json_decoded

the problem was i had forgotten to file.close() I did it and fixed the problem.

The accepted answer is the easiest one to fix the problem. But in case you are not allowed to install the simplejson due to your company policy, I propose below solution to fix the particular issue of “using comma on the last item in a list”:

1. Create a child class “JSONLintCheck” to inherite from class “JSONDecoder” and override the init method of the class “JSONDecoder” like below:

   def __init__(self, encoding=None, object_hook=None, parse_float=None,parse_int=None, parse_constant=None, strict=True,object_pairs_hook=None)        
           super(JSONLintCheck,self).__init__(encoding=None, object_hook=None,      parse_float=None,parse_int=None, parse_constant=None, strict=True,object_pairs_hook=None)
           self.scan_once = make_scanner(self)
   

2. make_scanner is a new function that used to override the ‘scan_once’ method of the above class. And here is code for it：

  1 #!/usr/bin/env python
  2 from json import JSONDecoder
  3 from json import decoder
  4 import re
  5
  6 NUMBER_RE = re.compile(
  7     r'(-?(?:0|[1-9]\d*))(\.\d+)?([eE][-+]?\d+)?',
  8     (re.VERBOSE | re.MULTILINE | re.DOTALL))
  9
 10 def py_make_scanner(context):
 11     parse_object = context.parse_object
 12     parse_array = context.parse_array
 13     parse_string = context.parse_string
 14     match_number = NUMBER_RE.match
 15     encoding = context.encoding
 16     strict = context.strict
 17     parse_float = context.parse_float
 18     parse_int = context.parse_int
 19     parse_constant = context.parse_constant
 20     object_hook = context.object_hook
 21     object_pairs_hook = context.object_pairs_hook
 22
 23     def _scan_once(string, idx):
 24         try:
 25             nextchar = string[idx]
 26         except IndexError:
 27             raise ValueError(decoder.errmsg("Could not get the next character",string,idx))
 28             #raise StopIteration
 29
 30         if nextchar == '"':
 31             return parse_string(string, idx + 1, encoding, strict)
 32         elif nextchar == '{':
 33             return parse_object((string, idx + 1), encoding, strict,
 34                 _scan_once, object_hook, object_pairs_hook)
 35         elif nextchar == '[':
 36             return parse_array((string, idx + 1), _scan_once)
 37         elif nextchar == 'n' and string[idx:idx + 4] == 'null':
 38             return None, idx + 4
 39         elif nextchar == 't' and string[idx:idx + 4] == 'true':
 40             return True, idx + 4
 41         elif nextchar == 'f' and string[idx:idx + 5] == 'false':
 42             return False, idx + 5
 43
 44         m = match_number(string, idx)
 45         if m is not None:
 46             integer, frac, exp = m.groups()
 47             if frac or exp:
 48                 res = parse_float(integer + (frac or '') + (exp or ''))
 49             else:
 50                 res = parse_int(integer)
 51             return res, m.end()
 52         elif nextchar == 'N' and string[idx:idx + 3] == 'NaN':
 53             return parse_constant('NaN'), idx + 3
 54         elif nextchar == 'I' and string[idx:idx + 8] == 'Infinity':
 55             return parse_constant('Infinity'), idx + 8
 56         elif nextchar == '-' and string[idx:idx + 9] == '-Infinity':
 57             return parse_constant('-Infinity'), idx + 9
 58         else:
 59             #raise StopIteration   # Here is where needs modification
 60             raise ValueError(decoder.errmsg("Expecting propert name enclosed in double quotes",string,idx))
 61     return _scan_once
 62
 63 make_scanner = py_make_scanner

3. Better put the ‘make_scanner’ function together with the new child class into a same file.

Just hit the same issue and in my case the problem was related to BOM (byte order mark) at the beginning of the file.

json.tool would refuse to process even empty file (just curly braces) until i removed the UTF BOM mark.

What I have done is:

• opened my json file with vim,
• removed byte order mark (set nobomb)
• save file

This resolved the problem with json.tool. Hope this helps!

When your file is created. Instead of creating a file with content is empty. Replace with:

json.dump({}, file)

You could use cjson, that claims to be up to 250 times faster than pure-python implementations, given that you have “some long complicated JSON file” and you will probably need to run it several times (decoders fail and report the first error they encounter only).