Parsing an arbitrary date format in dart

Question

I am developing an application that should accept arbitrarily formatted dates -- in other words, if it walks like date, talks like a date, it's a date.

I have attempted to adapt the date format parser from the moment.js package to generate a format string with some partial success but have run into a situation where the input string in it has some element that is not recognized.

For instance, there seems to be no proper format to deal with the string

Thu Nov 07 2019 13:50:03 GMT-0800 (PST)

-- or, at least one which seems clear to me but perhaps I'm missing something in the documentation.

I have looked at https://pub.dev/documentation/intl/latest/intl/DateFormat-class.html but I see no way of, for instance, ignoring fields that may be irrelevant or encoding the format correctly (for instance the GMT-0800 field).

I am using the jiffy package to do the date parsing if that matters.

Following up: I suspect what I'm looking for is something that acts more or less like strptime() in the C language. Or maybe not.

Answer 1

if it walks like date, talks like a date, it's a date

The Problem is that there is no such way to walk or such a date.

Dates are a hassle to implement properly in any language, as there are no rules defining exactly how a date should be formatted by human beings. The "arbitrarily formatted dates" you're asking for, well, they wouldn't be worth much as they wouldn't represent actual dates in a predictable manner. Check out the latest date/time related questions here on SO to get a perspective!

Using the the Dart DateTime class, you can only create a DateTime object by parsing a correctly formatted string which complies with a subset of ISO 8601

How you get to such a string from "arbitrary" input must be defined by you.

You can use some pre-defined constants to improve code readability, but using the DateTime class directly would require you to write some pretty cool code for extracting and formatting any possible perception of a date anywhere in the world into the correct syntax for creating a usable DateTime object. Maybe create a Date Input AI!

Luckily you can do a lot of heavy lifting using the class you mentioned: DateFormat Class, which is the way I would attack this.

It has ways to help you format localized date strings from input, then, together with some regular expressions you should be able to achieve what you need if you validate this based on "sane value ranges" for your application domain.

But the exact logic is something you'd be better off creating yourself, as this would vary from case to case / app to app.

Answer 2

The basic answer, as I mentioned in the comment to cseder was to encapsulate the uninteresting tokens of the date string in single quotes, per the documentation.

For instance, the format

EEE MMM DD yyyy H:mm:ss 'GMT-0800' '(PST)'

happily parses the string

Thu Nov 07 2019 13:50:03 GMT-0800 (PST)

I can probably also deal with the timezone issues but this reqolves the immediate question.

Answer 3

There are a number of ways to parse different timestamp formats. But note that accepting completely arbitrary formats can be ambiguous: does 01-02-2020 represent January 2, 2020 (typical U.S. convention) or February 1, 2020 (most other parts of the world)?

In the particular case you describe, I would break up the string with a RegExp first (or from .split(' ')) to separate the bits that DateFormat can handle (basically everything other than time zones, which DateFormat doesn't support). While you could use DateFormat's mechanism for escaping literal text to ignore the time zone, that would work only if the time zone is guaranteed to never change. Even if the strings come from a server in a fixed location, GMT-0800 (PST) might change to GMT-0700 (PDT) if daylight saving time was in effect.

The timezone itself you could parse manually with a RegExp, or you could use TZDateTime.parse from package:timezone.

I believe that the following code should parse the string as a UTC DateTime object:

DateTime parseDate(String timestamp) {
  final re = RegExp(''
      r'^(?<datetime>\w{3} \w{3} \d{2} \d{4} \d{2}:\d{2}:\d{2}) '
      r'\w{3}(?<offset>[+-]\d{4}) '
      r'.*$');
  final match = re.firstMatch(timestamp);
  if (match == null) {
    throw FormatException('Failed to parse timestamp: $timestamp');
  }

  final rawOffsetFromGMT = int.parse(match.namedGroup('offset'));
  final offsetFromGMT = Duration(
    hours: rawOffsetFromGMT ~/ 100,
    minutes: rawOffsetFromGMT.remainder(100),
  );

  return DateFormat('EEE MMM DD yyyy H:mm:ss')
      .parse(
        match.namedGroup('datetime'),
        true,
      )
      .subtract(offsetFromGMT);
}

void main() {
  print(parseDate('Thu Nov 07 2019 13:50:03 GMT-0800 (PST)'));
}