[#350] allow creation of unbounded Ranges

[Artus2b]

The class org.jooq.lambda.tuple.Range does not support nullable bounds.

For example in PostgreSQL a lower range bound of value null means -infinity.

See the Range Types documentation:

The lower bound of a range can be omitted, meaning that all points less than the upper bound are included in the range.

E.g. when testing that the range (,3) contains the value -12:

# SELECT '(,3)'::INT4RANGE @> -12;
 ?column?
----------
 t
(1 row)

What's left to do in this PR

I did not update the method Tuple2.intersect to handle null bounds, because it's a bit complicated, and I want to make sure that people support this PR before making this last effort 😄.

Other possible improvements

• support bound inclusiveness or exclusiveness (in documentation "Inclusive and Exclusive Bounds")
• port more functions than intersect and overlaps (in documentation "Range Functions and Operators")

Usage example

For PostgreSQL type tstzrange:

public class TsTzRangeBinding implements Binding<Object, Range<Instant>> {

    public class TsTzRangeConverter implements Converter<Object, Range<Instant>> {

        private static final Pattern PATTERN = Pattern.compile("\\[(.*?),(.*?)\\)");

        @Override
        public Range<Instant> from(Object t) {
            if (t == null) {
                return null;
            }

            Matcher m = PATTERN.matcher("" + t);
            if (m.find()) {
                return new Range<>(Instant.parse(m.group(1)), Instant.parse(m.group(2)));
            }

            throw new IllegalArgumentException("Unsupported range : " + t);
        }

        @Override
        public Object to(Range<Instant> u) {
            return u == null ? null : "[" + emptyIfNull(u.v1) + "," + emptyIfNull(u.v2) + ")";
        }

        private String emptyIfNull(Instant i) {
            return i == null ? "" : i.toString();
        }

        @Override
        public Class<Object> fromType() {
            return Object.class;
        }

        @SuppressWarnings({"unchecked", "rawtypes"})
        @Override
        public Class<Range<Instant>> toType() {
            return (Class) Range.class;
        }

    }

    @Override
    public Converter<Object, Range<Instant>> converter() {
        return new TsTzRangeConverter();
    }

    @Override
    public void sql(BindingSQLContext<Range<Instant>> ctx) {
        ctx.render()
                .visit(DSL.val(ctx.convert(converter()).value()))
                .sql("::TSTZRANGE");
    }

    // ...
}

... And register the binding:

ForcedType tstzrange = new ForcedType()
                .withUserType("org.jooq.lambda.tuple.Range<Instant>")
                .withBinding("org.artus.jooq.binding.TsTzRangeBinding")
                .withTypes("tstzrange");

[lukaseder]

PR rejected because it combines multiple, unrelated changes in a single change, and because it implements range related API in Tuple2 where it does not belong.

[lukaseder]

Thanks a lot for your suggestions. There are too many review findings for this to be merged as it is. But I will take the following bits from your commits and integrate them in slightly changed ways:

• The tests
• The contains method (but implemented differently: Add Range.contains(T) and Range.contains(Range) #354)
• Parts of the implementation

Thanks again for taking the time to suggest an implementation.

[Artus2b]

hey @lukaseder thanks for your time and comments/

So am I to submit corrected changes in another PR? or you are 'taking it from here' and use what's here to complete the change?
Is it the usual process on this project? I would have imagined that based on the comments received I could add 'fix' commits to this PR.

Edit: Oh I've just seen your change on the issue. all good then

[lukaseder]

So am I to submit corrected changes in another PR? or you are 'taking it from here' and use what's here to complete the change?

I'm taking it from here.

Is it the usual process on this project? I would have imagined that based on the comments received I could add 'fix' commits to this PR.

That's what I thought at first, until I saw how much needed fixing, at which point the ping pong of getting PRs, reviewing, getting PRs, reviewing takes much more of both our time than just doing it myself.

Edit: Oh I've just seen your change on the issue. all good then

Yep :)