ChatGPT解决这个技术问题 Extra ChatGPT

What is the "right" JSON date format?

I've seen so many different standards for the JSON date format:

"\"\\/Date(1335205592410)\\/\""         .NET JavaScriptSerializer
"\"\\/Date(1335205592410-0500)\\/\""    .NET DataContractJsonSerializer
"2012-04-23T18:25:43.511Z"              JavaScript built-in JSON object
"2012-04-21T18:25:43-05:00"             ISO 8601

Which one is the right one? Or best? Is there any sort of standard on this?

There is no date format in JSON, there's only strings a de-/serializer decides to map to date values.
strings, numbers, true, false, null, objects and arrays
However, JavaScript built-in JSON object and ISO8601 contains all the information to be understand by human and computer and does not relies on the beginning of the computer era (1970-1-1).

C
Community

JSON itself does not specify how dates should be represented, but JavaScript does.

You should use the format emitted by Date's toJSON method:

2012-04-23T18:25:43.511Z

Here's why:

It's human readable but also succinct It sorts correctly It includes fractional seconds, which can help re-establish chronology It conforms to ISO 8601 ISO 8601 has been well-established internationally for more than a decade ISO 8601 is endorsed by W3C, RFC3339, and XKCD

That being said, every date library ever written can understand "milliseconds since 1970". So for easy portability, ThiefMaster is right.


This is also the preferred representations according to ECMA: JSON.stringify({'now': new Date()}) "{"now":"2013-10-21T13:28:06.419Z"}"
I would add another important reason to the list: it's locale-independent. If you had a date like 02-03-2014 you'd need additional information to know if it refers to the 3rd of Feb or the 2nd of March. It depends on whether US-format or other format is used.
upvote for mentioning and linking xkcd :D @ajorquera I usually use momentjs for this. I've also seen issues with IE in this regard
Regarding the second point, it does not sort correctly after the year 10000. We do have almost 8000 years to come up with a new format though, so it's probably not an issue.
Actually, @Erfa, since - comes before digits in ASCII, it will sort just fine until the year 100,000. ;P
P
Peter Mortensen

JSON does not know anything about dates. What .NET does is a non-standard hack/extension.

I would use a format that can be easily converted to a Date object in JavaScript, i.e. one that can be passed to new Date(...). The easiest and probably most portable format is the timestamp containing milliseconds since 1970.


If you go this route, make sure you don't need to deal with dates earlier than 1970!
As @BenDolman said, this "solution" doesn't deal appropriately with dates prior to Jan 1st, 1970 (Epoch). Also, there is a reason ISO8601 exists in the first place. Here on Earth we have things called "time zones." Where is that in milliseconds? JSON may not have a standard for dates, but dates exist outside of JSON, and there is a standard for that. funroll's answer is the correct one (see also: xkcd.com/1179).
It is maybe also worth mentioning that (milli)seconds from 1970 isn't predictable for dates in the future because we have leap seconds. So I wouldn't use if for inter-process communication and data storage. It is however nice to use internally in a program since it can be stored in a single integer which gives you some performance benefits.
The Unix timestamp is always UTC, you convert from your local timezone before generating the timestamp, and back again to the local timezone on display, there is no ambiguity there.
Every one of these comments saying that pre-1970s dates or future dates can't be represented is misunderstanding epoch time. All time is relative, and the date string will really break down once you get at/below year 1. Whatever the source clock for your time, it's almost certainly based on a representation from epoch-time anyway, so you're not making it more accurate by avoiding epoch time.
B
Basil Bourque

There is no right format; The JSON specification does not specify a format for exchanging dates which is why there are so many different ways to do it.

The best format is arguably a date represented in ISO 8601 format (see Wikipedia); it is a well known and widely used format and can be handled across many different languages, making it very well suited for interoperability. If you have control over the generated json, for example, you provide data to other systems in json format, choosing 8601 as the date interchange format is a good choice.

If you do not have control over the generated json, for example, you are the consumer of json from several different existing systems, the best way of handling this is to have a date parsing utility function to handle the different formats expected.


@mlissner but that's an opinion on which one is best. ISO-8601 is a standard, but it's not the standard for JSON (even though I'd be inclined to use it); for example, Microsoft decided not to use it (msdn.microsoft.com/en-us/library/…). The best practice is to stick with one (sensible) convention , whatever that is. As I stated in the answer, the best way of handling this is to define a date parsing utility function that can handle the expected formats. If you integrate with systems that use different formats, the function should handle each case.
@RussCam, we can go back and forth, but if somebody is asking the best way to encode dates in JSON, they're asking how to format dates when they make JSON (and the answer is generally ISO-8601). You're answering the opposite question: how to consume JSON dates once they're already made (though your advice is sound).
The JSON schema specification actually says that dates that are verified by a schema must be in 8601 format.
@gnasher729 do you have a link?
@vallismortis - That is a draft specification for defining a schema for a given json structure exchanged between parties, not the format for dates in the json specification. I'm going to revise my answer as based on the comments, it doesn't appear I've made it clear enough
C
Community

From RFC 7493 (The I-JSON Message Format ):

I-JSON stands for either Internet JSON or Interoperable JSON, depending on who you ask.

Protocols often contain data items that are designed to contain timestamps or time durations. It is RECOMMENDED that all such data items be expressed as string values in ISO 8601 format, as specified in RFC 3339, with the additional restrictions that uppercase rather than lowercase letters be used, that the timezone be included not defaulted, and that optional trailing seconds be included even when their value is "00". It is also RECOMMENDED that all data items containing time durations conform to the "duration" production in Appendix A of RFC 3339, with the same additional restrictions.


This is also the format produced by Date().toISOString() and Date().toJSON(), with the limitation that Date doesn't track a timezone value, and hence always emits the timestamps in the UTC (Z) timezone. The value can be parsed using new Date("...") and Date.parseDate.
b
bb216b3acfd8f72cbc8f899d4d6963

When in doubt simply go to the javascript web console of a modern browser by pressing F12 (Ctrl+Shift+K in Firefox) and write the following:

new Date().toISOString()

Will output:

"2019-07-04T13:33:03.969Z"

Ta-da!!


T
Tel

Just for reference I've seen this format used:

Date.UTC(2017,2,22)

It works with JSONP which is supported by the $.getJSON() function. Not sure I would go so far as to recommend this approach... just throwing it out there as a possibility because people are doing it this way.

FWIW: Never use seconds since epoch in a communication protocol, nor milliseconds since epoch, because these are fraught with danger thanks to the randomized implementation of leap seconds (you have no idea whether sender and receiver both properly implement UTC leap seconds).

Kind of a pet hate, but many people believe that UTC is just the new name for GMT -- wrong! If your system does not implement leap seconds then you are using GMT (often called UTC despite being incorrect). If you do fully implement leap seconds you really are using UTC. Future leap seconds cannot be known; they get published by the IERS as necessary and require constant updates. If you are running a system that attempts to implement leap seconds but contains and out-of-date reference table (more common than you might think) then you have neither GMT, nor UTC, you have a wonky system pretending to be UTC.

These date counters are only compatible when expressed in a broken down format (y, m, d, etc). They are NEVER compatible in an epoch format. Keep that in mind.


I wouldn't use that format, but the rest of the info you provided is very useful, thank you!
A
Aryan

"2014-01-01T23:28:56.782Z"

The date is represented in a standard and sortable format that represents a UTC time (indicated by the Z). ISO 8601 also supports time zones by replacing the Z with + or – value for the timezone offset:

"2014-02-01T09:28:56.321-10:00"

There are other variations of the timezone encoding in the ISO 8601 spec, but the –10:00 format is the only TZ format that current JSON parsers support. In general it’s best to use the UTC based format (Z) unless you have a specific need for figuring out the time zone in which the date was produced (possible only in server side generation).

NB:

    var date = new Date();
    console.log(date); // Wed Jan 01 2014 13:28:56 GMT- 
    1000 (Hawaiian Standard Time) 
        
    var json = JSON.stringify(date);
    console.log(json);  // "2014-01-01T23:28:56.782Z"

To tell you that's the preferred way even though JavaScript doesn't have a standard format for it

// JSON encoded date
var json = "\"2014-01-01T23:28:56.782Z\"";

var dateStr = JSON.parse(json);  
console.log(dateStr); // 2014-01-01T23:28:56.782Z

J
Justus Romijn

JSON itself has no date format, it does not care how anyone stores dates. However, since this question is tagged with javascript, I assume you want to know how to store javascript dates in JSON. You can just pass in a date to the JSON.stringify method, and it will use Date.prototype.toJSON by default, which in turns uses Date.prototype.toISOString (MDN on Date.toJSON):

const json = JSON.stringify(new Date());
const parsed = JSON.parse(json); //2015-10-26T07:46:36.611Z
const date = new Date(parsed); // Back to date object

I also found it useful to use the reviver parameter of JSON.parse (MDN on JSON.parse) to automatically convert ISO strings back to javascript dates whenever I read JSON strings.

const isoDatePattern = new RegExp(/\d{4}-[01]\d-[0-3]\dT[0-2]\d:[0-5]\d:[0-5]\d\.\d+([+-][0-2]\d:[0-5]\d|Z)/);

const obj = {
 a: 'foo',
 b: new Date(1500000000000) // Fri Jul 14 2017, etc...
}
const json = JSON.stringify(obj);

// Convert back, use reviver function:
const parsed = JSON.parse(json, (key, value) => {
    if (typeof value === 'string' &&  value.match(isoDatePattern)){
        return new Date(value); // isostring, so cast to js date
    }
    return value; // leave any other value as-is
});
console.log(parsed.b); // // Fri Jul 14 2017, etc...

nice, I like it!
A
Alireza

The prefered way is using 2018-04-23T18:25:43.511Z...

The picture below shows why this is the prefered way:

https://i.stack.imgur.com/HhBOj.png

So as you see Date has a native Method toJSON, which return in this format and can be easily converted to Date again...


Correct! The JSON Data Interchange Syntax does not specify the standard: ecma-international.org/publications/files/ECMA-ST/ECMA-404.pdf but in practice, ISO 8601 compatible formats are more desirable across platforms including JavaScript runtime.
i
irundaia

In Sharepoint 2013, getting data in JSON there is no format to convert date into date only format, because in that date should be in ISO format

yourDate.substring(0,10)

This may be helpful for you


C
Ciabaros

I believe that the best format for universal interoperability is not the ISO-8601 string, but rather the format used by EJSON:

{ "myDateField": { "$date" : <ms-since-epoch> } }

As described here: https://docs.meteor.com/api/ejson.html

Benefits

Parsing performance: If you store dates as ISO-8601 strings, this is great if you are expecting a date value under that particular field, but if you have a system which must determine value types without context, you're parsing every string for a date format. No Need for Date Validation: You need not worry about validation and verification of the date. Even if a string matches ISO-8601 format, it may not be a real date; this can never happen with an EJSON date. Unambiguous Type Declaration: as far as generic data systems go, if you wanted to store an ISO string as a string in one case, and a real system date in another, generic systems adopting the ISO-8601 string format will not allow this, mechanically (without escape tricks or similar awful solutions).

Conclusion

I understand that a human-readable format (ISO-8601 string) is helpful and more convenient for 80% of use cases, and indeed no-one should ever be told not to store their dates as ISO-8601 strings if that's what their applications understand, but for a universally accepted transport format which should guarantee certain values to for sure be dates, how can we allow for ambiguity and need for so much validation?


See this answer earlier in the thread for why milliseconds since epoch has caveats such as incorrect computation of leap seconds etc: stackoverflow.com/a/42480073/190476
@SudhanshuMishra The caveats you reference are general gotchas for extremely academic concerns of unix timestamps, mostly related to timestamp generation. This is even less of a concern with millisecond resolution. As was called out in another comment, most computer dates are internally represented as a unix timestamps, even if they are exposed and formatted otherwise. Nonetheless, there is nothing wrong with millisecond representation of any given date+time, especially when compared with other approaches, which can easily be affected by the same nano-impact caveats under the hood.
Just to add regarding concerns of "out of range" dates for unix timestamps: those are system storage issues, to be addressed in a much wider context than the transport format. For instance, this format does not need to be limited to integers that fit into 32 bits, nor does it need to be strictly positive numbers, but nobody is going to solve the "year 2038 problem" by dropping timestamps on a system/architecture level; they just need to be expanded (e.g. to 64-bit or beyond), and this does not affect this proposed transport format.
Too late. That might have been a good format to use for JSON, but now most systems in the wild use ISO8601 or straight time since epoch, anything else is an interoperability headache. EJSON reminds me of this: xkcd.com/927 It provide solutions for working with JSON in JS, but there is an awful lot of JSON that isn't processed using JS.
R
Rosdi Kasim

it is work for me with parse Server

{
    "ContractID": "203-17-DC0101-00003-10011",
    "Supplier":"Sample Co., Ltd",
    "Value":12345.80,
    "Curency":"USD",
    "StartDate": {
                "__type": "Date",
                "iso": "2017-08-22T06:11:00.000Z"
            }
}

C
Chad Wilson

There is only one correct answer to this and most systems get it wrong. Number of milliseconds since epoch, aka a 64 bit integer. Time Zone is a UI concern and has no business in the app layer or db layer. Why does your db care what time zone something is, when you know it's going to store it as a 64 bit integer then do the transformation calculations.

Strip out the extraneous bits and just treat dates as numbers up to the UI. You can use simple arithmetic operators to do queries and logic.


Now you have 2 problems: Which epoch should you choose, and which milliseconds should you count? Probably the most common choice is Unix time (1970-01-01T00:00:00 UTC and SI milliseconds except for those which fall within a leap second), but of course that makes future times undefined.
So how do you represent microseconds? RFC3339 works fine with any precision, you'll have a reader that parses the timezone and gives you the right time stamp, and it's additional information. Calendar apps usually care about time zones.
Timezone is not a UI concern, unless you don't mind missing your next flight. Flights are posted in local time and follow specific rules for DST changes. Losing the offset means losing important business information
Some further counterarguments include the ability to represent times before 1970 (assuming that particular epoch), and JSONs tendency to be somewhat human-readable.
Regarding the timezone comment pitchforks: while I agree that it's dangerous to say which layer's concern it is (UI, Business Logic as Data), as that is application specific -- a universal approach to STORE datetimes should NOT include the timezone as you can always derive any timezone specific variation of the datetime, through standard well understood means. I feel that any arguments for storing time zones universally baked into dates are misguided in trying to include LOCATION into the datetime, which should rather be stored as separate independent data.
P
Phil3992

The following code has worked for me. This code will print date in DD-MM-YYYY format.

DateValue=DateValue.substring(6,8)+"-"+DateValue.substring(4,6)+"-"+DateValue.substring(0,4);

else, you can also use:

DateValue=DateValue.substring(0,4)+"-"+DateValue.substring(4,6)+"-"+DateValue.substring(6,8);

M
Mukus

I think that really depends on the use case. In many cases it might be more beneficial to use a proper object model (instead of rendering the date to a string), like so:

{
"person" :
      {
 "name" : {
   "first": "Tom",
   "middle": "M",
  ...
}
 "dob" :  {
         "year": 2012,
         "month": 4,
         "day": 23,
         "hour": 18,
         "minute": 25,
         "second": 43,
         "timeZone": "America/New_York"
    }   
   }
}

Admittedly this is more verbose than RFC 3339 but:

it's human readable as well

it implements a proper object model (as in OOP, as far as JSON permits it)

it supports time zones (not just the UTC offset at the given date and time)

it can support smaller units like milliseconds, nanoseconds, ... or simply fractional seconds

it doesn't require a separate parsing step (to parse the date-time string), the JSON parser will do everything for you

easy creation with any date-time framework or implementation in any language

can easily be extended to support other calendar scales (Hebrew, Chinese, Islamic ...) and eras (AD, BC, ...)

it's year 10000 safe ;-) (RFC 3339 isn't)

supports all-day dates and floating times (Javascript's Date.toJSON() doesn't)

I don't think that correct sorting (as noted by funroll for RFC 3339) is a feature that's really needed when serializing a date to JSON. Also that's only true for date-times having the same time zone offset.


I doubt anyone will be using json in the year 10000, or even that by that time the year 10000 will still be year 10000. But if both of those things are still true by then, the format can simply be expanded to contain a 3 digit century component. So I'd say people can safely stick with RFC 3339, at least until the year 9900
@downvoters: According to Why is voting important? you should downvote if a post contains wrong information, is poorly researched, or fails to communicate information. Please explain for which one of these reasons you've downvoted this answer.
@Marten Two things. 1. You are never owed an explanation for downvotes, though I understand it can be helpful. 2. I did not downvote your answer, but I would guess that people don't like your answer because they think it is the wrong way to do this. That would qualify it as "Wrong information" since the question is looking for the best way to do something
I did not downvote you, but I can certainly understand how "invent yet another poorly specified format" (which is basically what you're saying) would be seen as wrong or poorly researched.
@Phil, UTC is not really a time zone (there is no place on earth which uses "UTC" as its official time zone), it's a time standard. Also time zone offsets are quite unpredictable. There is no way to say if in 2025 "12:00 Moscow time" is still "9:00 UTC" like it is today, it has been changed a couple of times during the last 30 years. If you want to express a future local time you need true time zones.