Methods defined on T are also available on *T. Thought this technically
changes the interface, because *T will be automatically dereferenced as
needed it shouldn't be a breaking change.
See a8802b16cc for similar change.
Methods defined on T are also available on *T. This change makes Value
consistent with database/sql Value implementations. It also makes Value,
EncodeBinary, and EncodeText more convenient to use because you can
pass T or *T as an argument to a query.
The MarshalJSON change is even more significant because without it
json.Marshal would generate the "%v" format instead of the implemented
MarshalJSON.
Thought this technically changes the interface, because *T will be
automatically dereferenced as needed it shouldn't be a breaking change.
See: https://github.com/jackc/pgx/issues/538 for initial discussion.
Because reading a record type requires the decoder to be able to look up oid
to type mapping and types such as hstore have types that are not fixed between
different PostgreSQL servers it was necessary to restructure the pgtype system
so all encoders and decodes take a *ConnInfo that includes oid/name/type
information.
Set having the capability to assign an object of the same type was
inconsistently implemented. Some places it was not implemented at all, some
places it was a shallow copy, some places a deep copy. Given that it doesn't
seem likely to ever be used, and if it is needed it is easy enough to do
outside of the library this code has been removed.
To aid in composability, these methods no longer write their own length. This
is especially useful for text formatted arrays and may be useful for future
database/sql compatibility. It also makes the code a little simpler as the
types no longer have to compute their own size.
Along with this, these methods cannot encode NULL. They now return a boolean
if they are NULL. This also benefits text array encoding as numeric arrays
require NULL to be exactly `NULL` while string arrays require NULL to be
`"NULL"`.
Jack Christensen