NullProvider has been removed

[jorgheymans]

It seems that NullProvider was removed from jackson 2.8 onwards, this library is still using it. It's on their deprecated list for 2.7 (https://fasterxml.github.io/jackson-databind/javadoc/2.7/deprecated-list.html) but it doesn't say how to replace it.

[jhaber]

Thanks for the tip! I believe I found NullProvider by looking through the Jackson source and seeing how it handles null values, I'll take a look at the 2.8 source and see what it does now.

[cowtowncoder]

Worth noting that NullProvider has not been used sine 2.6, although class itself was left for bit longer.
I can probably help but I'd need to know what the intended usage was.

[jhaber]

I use it when I encounter a null value during deserialization of primitives, such as here. The intention was to avoid dealing directly with DeserializationFeature.FAIL_ON_NULL_FOR_PRIMITIVES; I thought going through NullProvider would be more robust than trying to duplicate the logic. In general I was trying to keep the protobuf deserializer behavior similar to BeanDeserializer (make sure we throw the same exception with the same message, for example), but it seems like duplicating the code may be the way to go.

[cowtowncoder]

Ah. So as a convenient checker for whether nulls may be coerced.
I did not realize anyone was using, outside of use by SettableBeanProperty, which is why it was removed.

Code duplication may indeed be the way to go, since the only (?) pluggable way I can think of is to access deserializers for relevant types (like int == Integer.TYPE), which is possible, but some work for maybe minor gain. Doing that would require implementing ResolvableDeserializer (method resolve(...) to fetch deserializers to delegate to (in theory it can be done from deserialize(...) but lookups are relatively expensive.
One benefit there would be to get fully standard checks and coercions (like unwrapping of single-element arrays).

Then again, handling of nulls is the same for all primitives so maybe amount of duplication is limited.

[stevenschlansker]

This issue can probably be closed now.