Validate and format JSON files

Yes, there are ways to validate that a JSON file is valid. One way is to use a JSON parsing library that will throw exceptions if the input you provide is not well-formatted.

try:
   load_json_file(filename)
except InvalidDataException: # or something
   # oops guess it's not valid

Of course, if you want to fix it, you naturally cannot use a JSON loader since, well, it's not valid JSON in the first place. Unless the library you're using will automatically fix things for you, in which case you probably wouldn't even have this question.

One way is to load the file manually and tokenize it and attempt to detect errors and try to fix them as you go, but I'm sure there are cases where the error is just not possible to fix automatically and would be better off throwing an error and asking the user to fix their files.

I have not written a JSON fixer myself so I can't provide any details on how you might go about actually fixing errors.

However I am not sure whether it would be a good idea to fix all errors, since then you'd have assume your fixes are what the user actually wants. If it's a missing comma or they have an extra trailing comma, then that might be OK, but there may be cases where it is ambiguous what the user wants.

The built-in JSON module can be used as a validator:

import json

def parse(text):
    try:
        return json.loads(text)
    except ValueError as e:
        print('invalid json: %s' % e)
        return None # or: raise

You can make it work with files by using:

with open(filename) as f:
    return json.load(f)

instead of json.loads and you can include the filename as well in the error message.

On Python 3.3.5, for {test: "foo"}, I get:

invalid json: Expecting property name enclosed in double quotes: line 1 column 2 (char 1)

and on 2.7.6:

invalid json: Expecting property name: line 1 column 2 (char 1)

This is because the correct json is {"test": "foo"}.

When handling the invalid files, it is best to not process them any further. You can build a skipped.txt file listing the files with the error, so they can be checked and fixed by hand.

If possible, you should check the site/program that generated the invalid json files, fix that and then re-generate the json file. Otherwise, you are going to keep having new files that are invalid JSON.

Failing that, you will need to write a custom json parser that fixes common errors. With that, you should be putting the original under source control (or archived), so you can see and check the differences that the automated tool fixes (as a sanity check). Ambiguous cases should be fixed by hand.