Bulk insert

[Mark Rotteveel]

On 10-10-2018 12:06, Arthur McGibbon wrote:
> Would it be possible to add bulk insert functionality to the spec?
> 
> Many drivers already implement this in their own way.  Their procedure is
> generally to define the column metadata, define some properties about the
> table (e.g. locking) and then fire a stream at it.
> 
> For reference...
> 
> MS SQL: com.microsoft.sqlserver.jdbc.SQLServerBulkCopy
> 
> PostgreSQL:org.postgresql.copy.CopyManager
> 
> MySQL:com.mysql.cj.jdbc.StatementImpl#setLocalInfileInputStream

The problem with this is that these are all specifically native 
solutions, as in: the file format, configuration options, etc are 
specific to the underlying database and usually have different tools for 
that database that support the same options and format.

Having such a thing for JDBC, in my opinion, would require a common 
(JDBC-defined) input format, common config, common api, etc, and not 
just provide a basic entrypoint to get access. If you don't do that, 
you're not really providing the database-independent API that JDBC tries 
to be.

And trying to unify an API that would be suitable to cover the existing 
options would probably be complex. My guess is that it would likely not 
gain much traction, because these tools already exist and they map well 
to equivalent features in other tools for that database system.

And if you just provide the common denominator or simplest thing 
possible, then people will probably still end up using the 
database-specific tooling, as they will likely be richer in features.

I personally think this is one of those things where it is better to 
just escape JDBC, and use the database specific tools.

That said, I assume that if you can come up with a proposal and a 
workable API, it could be considered by the EG (could you confirm, Lance).

Mark
-- 
Mark Rotteveel