MaxValue for uint32 type validated against int32.max

[kdkavanagh]

Checking to see if this validation implementation is intentional. My expectation was that a uint32 max value would be checked against uint32.max, not int32.max

From CompositeType.java:

            final long maxInt = INT32.maxValue().longValue();
            if (primitiveType == UINT32 && longValue > maxInt)
            {
                XmlSchemaParser.handleError(node, String.format(
                    "maxValue greater than allowed for type: maxValue=%d allowed=%d", longValue, maxInt));
            }

[kdkavanagh]

Guess based on the related generated code for dimension en/decoders that there is some requirement that both blockLen and numInGrp are always <= int32.

Related to #609

[mjpt777]

This is correct. The spec recommends UINT8 and UINT16 for the types. We allow a bit of flexibility in going larger but given the limitations of the Java language we limit to max signed int.

[kdkavanagh]

SBE also enforces unsigned types for numInGrp. Should this restriction be relaxed to allow 2^31 (int32) objects in a repeating group? While semantically a signed count doesn't make a ton of sense, there is a big difference between 2^16 and 2^31

[mjpt777]

That is a question for the spec more so than this implementation.