Support BigInteger and BigDecimal creators in StdValueInstantiator

[eatdrinksleepcode]

Currently using 2.9.6; have searched and don't see any planned changes in this area.

Basically, I would like this test to pass by default:

private static class BigDecimalWrapper {
    public final BigDecimal value;
    public BigDecimalWrapper(BigDecimal value) { this.value = value; }
}
@Test
public void createFromBigDecimal() throws IOException {
    assertThat(new ObjectMapper().readValue("42.5", BigDecimalWrapper.class).value, is(new BigDecimal("42.5")));
}

the same way that this one does:

private static class DoubleWrapper {
    public final Double value;
    public DoubleWrapper(Double value) { this.value = value; }
}
@Test
public void createFromDouble() throws IOException {
    assertThat(new ObjectMapper().readValue("42.5", DoubleWrapper.class).value, is(new Double("42.5")));
}

Currently, StdValueInstantiator does not support BigDecimal (or BigInteger) at all. I would like to add support for BigDecimal creators to createFromDouble, createFromLong, and createFromInt (and BigInteger creators to createFromLong and createFromInt) in StdValueInstantiator as widening conversions, the same way that createFromInt supports using long creators as a widening conversion.

[cowtowncoder]

First of all, thank you for checking existing issues to see if there was previous work (there isn't AFAIK).
I can also see how this could be beneficial.

The only concern I have is regarding extension of auto-detection, and whether it is something that would be best supported by figuring out an even more general extension point for allowing single-argument auto-detected constructors (and factory methods -- currently only hard-coded valueOf() name, but could obviously be customizable as well).

But for 2.10, we could start by allowing BigDecimal as an "alias" of sort for double (and perhaps float too, if it is not yet), and similarly perhaps BigInteger.

One practical concern is that I would prefer not to add yet more _fromXxxCreator entries in StdValueInstantiator, if feasible -- but not sure if that can be avoided without adding other types of complexities.

[eatdrinksleepcode]

I think a more general extension point is definitely worth discussion. In trying to work around this, I have been trying for the last couple of days to define an extension that I can add to my ObjectMapper to do this automatically (I don't want to use @JsonCreator or similar because I have a lot of these types in my code). I have so far been unsuccessful; neither custom deserializers nor custom value instantiators seem capable of solving this at a general level.

I understand the concern about adding creators to StdValueInstantiator, but I'm not sure I see another way of achieving this that is remotely consistent with the current implementation.

[cowtowncoder]

Another thing I'll add as a note is the problem wrt double expanding to BigDecimal: that is, using 2-based double, with possible lossy value, into fully accurate BigDecimal.
Ideally value would be parsed as BigDecimal, passed, and no (possibly lossy) conversion occurs.
This for use cases like monetary values where 0.1 can be accurately represented as BigDecimal but NOT as 2-based double or float (although in practice conversion in Java may seem safe... but actual value is off by a bit).

But technically speaking it seems quite doable to just add new overload for createFromDouble() (or createFromBigDecimal()), as well as add createFromBigInteger(), add default implementation similar to existing methods.

In fact I'll tag this as "new contributor friendly"; if anyone wants to take a look.
If not I may come back to this before 2.12.0 finalized.

[verkhovin]

@cowtowncoder Hi! if you haven't picked it up yet, i would work on this

[cowtowncoder]

@verkhovin I have not, so you are welcome to try it!

[cowtowncoder]

Implementation contributed by @upsidedownsmile, just merged.