Wildcard in DateTimeFormatter

Question

I need to parse a string into a LocalDate. The string looks like 31.* 03 2016 in regex terms (i.e. .* means that there may be 0 or more unknown characters after the day number).

Example input/output: 31xy 03 2016 ==> 2016-03-31

I was hoping to find a wildcard syntax in the DateTimeFormatter documentation to allow a pattern such as:

LocalDate.parse("31xy 03 2016", DateTimeFormatter.ofPattern("dd[.*] MM yyyy"));

but I could not find anything.

Is there a simple way to express optional unknown characters with a DateTimeFormatter?

ps: I can obviously modify the string before parsing it but that's not what I'm asking for.

Answer 1

I’d do it in two steps, use a regexp to get the original string into something that LocalDate can parse, for example:

String dateSource = "31xy 03 2016";
String normalizedDate = dateSource.replaceFirst("^(\\d+).*? (\\d+ \\d+)", "$1 $2");
LocalDate date = LocalDate.parse(normalizedDate, DateTimeFormatter.ofPattern("dd MM yyyy"));
System.out.println(date);

I know it’s not what you asked for.

Answer 2

There is no direct support for this in java.time.

The closest would be to use parse(CharSequence,ParsePosition) using two different formatters.

// create the formatter for the first half
DateTimeFormatter a = DateTimeFormatter.ofPattern("dd")

// setup a ParsePosition to keep track of where we are in the parse
ParsePosition pp = new ParsePosition();

// parse the date, which will update the index in the ParsePosition
String str = "31xy 03 2016";
int dom = a.parse(str, pp).get(DAY_OF_MONTH);

// some logic to skip the messy 'xy' part
// logic must update the ParsePosition to the start of the month section
pp.setIndex(???)

// use the parsed day-of-month in the formatter for the month and year
DateTimeFormatter b = DateTimeFormatter.ofPattern("MM yyyy")
    .parseDefaulting(DAY_OF_MONTH, dom);

// parse the date, using the *same* ParsePosition
LocalDate date = b.parse(str, pp).query(LocalDate::from);

While the above is untested it should basically work. However, it would be far easier parse it manually.