如何解决scala中的无效日期格式错误？

Question

I m getting follwing error:

val formatter = ISODateTimeFormat.dateTimeParser()
scala> val date2 = "Tue Dec 29 11:11:30 IST 2015"
 date2: String = Tue Dec 29 11:11:30 IST 2015

scala> formatter.parseDateTime(date2)
java.lang.IllegalArgumentException: Invalid format: "Tue Dec 29 11:11:30 IST 2015" is malformed at "ue Dec 29 11:11:30 IST 2015"

how to resolve following error??

Answer 1

I think you have the wrong format here - using SimpleDateFormat and a bit of googling this works:

scala> val formatter = new java.text.SimpleDateFormat("EEE MMM d HH:mm:ss Z yyyy")
formatter: java.text.SimpleDateFormat = java.text.SimpleDateFormat@73342172

scala> formatter.parse("Tue Dec 29 11:11:30 IST 2015")
res1: java.util.Date = Tue Dec 29 09:11:30 GMT 2015

edit: errr don't forget the timezone and year like I originally did ;-)

Answer 2

First thing to note:

Your input is not in ISO-Format and contains the name of a timezone (here: IST). You tried Joda-Time, but used an ISO-Format. This cannot work because the ISO-format pattern does not match the non-ISO-input . Second reason against Joda-Time is the fact that Joda-Time cannot parse timezone names . So following approach using a theoretically correct pattern will fail:

String input = "Tue Dec 29 11:11:30 IST 2015";
DateTimeFormatter dtf = 
  DateTimeFormat.forPattern("EEE MMM d HH:mm:ss z yyyy").withLocale(Locale.ENGLISH);
DateTime unparseable = dtf.parseDateTime(input); 
// java.lang.IllegalArgumentException: 
// Invalid format: "Tue Dec 29 11:11:30 IST 2015" is malformed at "IST 2015"

So you can only change the library. An obvious candidate using the (horrible) class SimpleDateFormat is:

String input = "Tue Dec 29 11:11:30 IST 2015";
String pattern = "EEE MMM d HH:mm:ss z yyyy";
SimpleDateFormat sdf = new SimpleDateFormat(pattern, Locale.ENGLISH);
java.util.Date jud = sdf.parse(input);
System.out.println(jud); // Tue Dec 29 10:11:30 CET 2015 (in my local tz CET=+01:00)

Okay no exception. But this does not mean that the result is automatically correct. The result can only be explained by having an offset of (+02:00 for IST). But is this true??? So let's have a closer look at the timezone involved:

TimeZone india = TimeZone.getTimeZone("Asia/Kolkata");
TimeZone israel = TimeZone.getTimeZone("Asia/Jerusalem"); 
System.out.println(israel.getDisplayName(false, TimeZone.SHORT)); // IST
System.out.println(india.getDisplayName(false, TimeZone.SHORT)); // IST

System.out.println(israel.getOffset(jud.getTime()) / 1000); // 7200 = +02:00
System.out.println(india.getOffset(jud.getTime()) / 1000); // 19800 = +05:30

This should trigger an alarm. Timezone names (especially abbreviations, here: IST) are often ambivalent and denote different timezones with different offsets.

So if you have got the input from Israel then you can be happy, but if from India then the result is wrong by 3:30 hours. Maybe your result will be such that it will match the India case instead of Israel. This will depend on your local timezone configuration. What so ever, don't blindly trust the parsed offsets.

The alternative Java-8:

String input = "Tue Dec 29 11:11:30 IST 2015";
ZoneId india = ZoneId.of("Asia/Kolkata");

DateTimeFormatterBuilder builder = new DateTimeFormatterBuilder();
builder.appendPattern("EEE MMM d HH:mm:ss ");
builder.appendZoneText(TextStyle.SHORT, Collections.singleton(india)); // preferred zone
builder.appendPattern(" yyyy");
DateTimeFormatter dtf = builder.toFormatter(Locale.ENGLISH);

ZonedDateTime zdt = ZonedDateTime.parse(input, dtf);
System.out.println(zdt); // 2015-12-29T11:11:30+05:30[Asia/Kolkata]

// compare dangerous standard approach (not specifying your zone preference)
String pattern = "EEE MMM d HH:mm:ss z yyyy";
zdt = ZonedDateTime.parse(input, DateTimeFormatter.ofPattern(pattern, Locale.ENGLISH));
System.out.println(zdt); // 2015-12-29T11:11:30+02:00[Asia/Jerusalem]

It will work if you specify your preferred timezone but you really need to think twice before you parse timezone names. The builder approach might appear a little bit awkward but cannot be avoided due to the difficulty of the problem. So Java-8 is very fine to give you a solution here.

By the way, if you use a strict style ( DateTimeFormatter.ofPattern(pattern, Locale.ENGLISH).withResolverStyle(ResolverStyle.STRICT) ) then the parser will throw an exception with the message:

java.time.format.DateTimeParseException: Text 'Tue Dec 29 11:11:30 IST 2015' could not be parsed: Unable to obtain ZonedDateTime from TemporalAccessor: {YearOfEra=2015, DayOfMonth=29, DayOfWeek=2, MonthOfYear=12},ISO,Asia/Jerusalem resolved to 11:11:30 of type java.time.format.Parsed

The message is somehow mysterious but I assume it is because of the ambivalent name IST.

If you are working on a platform with older JDK (Java 6 or 7) then you might consider ThreetenBP . ThreetenBP has the advantage to make a future migration easy (just changing the import statements) but my own experiments with the builder approach failed, unfortunately (even failed with the newest version v1.3.1 - maybe this depends on the underlying JDK???):

String input = "Tue Dec 29 11:11:30 IST 2015";
ZoneId india = ZoneId.of("Asia/Kolkata");

DateTimeFormatterBuilder builder = new DateTimeFormatterBuilder();
builder.appendPattern("EEE MMM d HH:mm:ss ");
builder.appendZoneText(TextStyle.SHORT, Collections.singleton(india)); // preferred zone
builder.appendPattern(" yyyy");
DateTimeFormatter dtf = builder.toFormatter(Locale.ENGLISH);

ZonedDateTime zdt = ZonedDateTime.parse(input, dtf);
System.out.println(zdt); // 2015-12-29T11:11:30+02:00[Israel] // why???

Else you can try my library Time4J which works on Java-6 or 7 (or later). It works similar to Java-8:

String input = "Tue Dec 29 11:11:30 IST 2015";
TZID india = ASIA.KOLKATA;

ChronoFormatter<Moment> f = ChronoFormatter.setUp(Moment.axis(),  Locale.ENGLISH)
    .addPattern("EEE MMM d HH:mm:ss ", PatternType.CLDR)
    .addShortTimezoneName(Collections.singleton(india)) // preferred zone
    .addPattern(" yyyy", PatternType.CLDR)
    .build();

 System.out.println(f.parse(input)); // 2015-12-29T05:41:30Z
 System.out.println(ZonalDateTime.parse(input, f)); // 2015-12-29T11:11:30UTC+05:30[Asia/Kolkata]

The error message in strict or smart parser style using

String input = "Tue Dec 29 11:11:30 IST 2015";
String pattern = "EEE MMM d HH:mm:ss z yyyy";

ChronoFormatter<Moment> f = // here smart standard style!
    ChronoFormatter.ofMomentPattern(
        pattern, PatternType.CLDR, Locale.ENGLISH, ZonalOffset.UTC);
f.parse(input);

will be:

Time zone name "IST" not found among preferred timezones in locale en, candidates=[Asia/Colombo, Asia/Jerusalem, Asia/Kolkata, Europe/Dublin]

Then you will immediately see that "IST" can be associated with different timezones.