Duplicate keys aren't cool!

thejh commented 12 years ago

$ echo '{"a":1,"a":2}' | jsonlint
The "sys" module is now called "util". It should have a similar interface.
{
  "a": 2
}
$

zaach commented 12 years ago

By default we simply follow the spec, but it wouldn't hurt to have as an option.

thejh commented 12 years ago

I guess I expect something with "lint" in the name to do some plausibility checking.

Klortho commented 11 years ago

+1 Any chance this will get fixed? Also, I'd think it should complain about empty-string keys:

{ "": "foo" }

rngadam commented 10 years ago

+1 just got bitten by a bug caused by a duplicate key, would love to use a tool such as jsonlint to catch these errors in my automated tests.

dharasty commented 10 years ago

JavaScript, JSON, and Python all allow a key to be duplicated. (It is NOT a syntax error; the value of the last occurrence of the key is kept.)

JavaScript, JSON, and Python all allow the empty string as a valid key. (It is NOT a syntax error.)

So my vote is that these are NOT flagged as error or even warnings, unless the JSONLint user specifically calls for them via a commandline option.

rngadam commented 10 years ago

@dharasty : what's the point of a lint tool if it doesn't catch anything more than the real parser? I think it's more than likely that there's an issue when there are duplicate keys...

mattfenwick commented 10 years ago

@rngadam checking for duplicate keys was a big problem for me, and since this issue is still unresolved after 2 years, I had to bite the bullet and implement my own checker.

I agree with your position -- most people expect keys to be unique, key/value pairs to not be silently discarded, and the resultant object to match the JSON input. All of these reasonable expectations are broken, making debugging quite painful. :+1:

dharasty commented 10 years ago

I see your points (@rngadam and @mattfenwick) and understand that if this as "bitten" you, you'd like a checker.

I guess I just don't use linting tools that way: I tend to use them to validate syntax and to pretty-format.

I do scratch my head a bit at "most people expect ... the resultant object to match the JSON input". I'm not sure what that means if the JSON input in the presence of multiple identical keys, which is allowed by in the serialization defined in the JSON spec, but clearly not in the concept of a JSON (or JavaScript) object. Therefore I'm not sure what is "reasonable" to expect.

I -- for one -- expect that repeated keys silently trump prior instances of the same key... but just because it it my experience that that is true (in JSON, JavaScript, and Python).

Is this your view?: "because purposeful use duplicated keys are very rare, when seen, it JSONLint should emit a warning".

Delagen commented 9 years ago

+1

odino commented 9 years ago

+1

pixelastic commented 9 years ago

I would also like to have jsonlint emit a warning on duplicate keys.

itsikavidan commented 9 years ago

+1

amire80 commented 9 years ago

It actually happened to me at least once that a duplicate key written manually by mistake caused a real bug, so it would be nice to get a warning about that. In my case I'd prefer that to be a fatal error in the build that would force me to fix it before deployment.

hotoo commented 9 years ago

:+1:

mikeabiezzi commented 8 years ago

100% agree. There is no reason to purposefully hard code duplicate keys in a JSON file. A linter should definitely check for this. Via an option is reasonable, considering it's not part of the JSON spec. Any chance this will get added? Need help creating a PR?

azamat-sharapov commented 7 years ago

A pure JavaScript version of the service provided at jsonlint.com.

Judging from readme, this projects claims to work like jsonlint.com, but compared to jsonlint.com - it doesn't detect duplicates.

adbl commented 7 years ago

We use this library for linting manually written json files, and this problem caused confusion and time wasted more than once. I'd be happy to have it as an option if it shouldn't be the default behavior...