-
-
Notifications
You must be signed in to change notification settings - Fork 1.4k
Description
It turns out that buffering of token streams cannot always be done without some format-specific additions. For example:
- XML module may need to retain information on "attribute-ness", namespace(s); possible "unwrapped"
Collection
types - Ion and Avro probably have additional concepts that may need to retain
But currently databind will explicitly construct format-agnostic TokenBuffer
, which can support some aspects (native type, object ids) but with added overhead and complexity.
It should be relatively easy, however, to (eventually?) allow format modules to override this selection: the way to do that would be to first make allocations of buffer instances go through:
DeserializationContext
(for deserialization)SerializerProvider
(for serialization)
and then format module may override one or both as necessary (XML module already overrides both, for example); or, if this becomes more widely necessary, context objects can offer other means of pluggability.
Either way, the first step is to add this one redirection.