Do all the things like ++ or -- rants, post your own rants, comment on others' rants and build your customized dev avatarSign Up
Yep, Oracle is the hautest of haute garbage.
Try layering dapper over it. It should allow you to use the managed client and at least have types on the right side of the operation. Run the database out of docker for local and it should dramatically reduce your entropy.
We hit similar issues when we began autogenerating our data connectors for our client integrations. This was our solve.
The perf hit of ODBC is the greater evil, at least to my mind. For most non-MSSQL dbs, the happy path will unfortunately be to allow the left side to be a dataframe, and the output to be whatever your coercable T is. Dapper will at least eliminate the boilerplate, so the workflow becomes query/sproc -> dataframe -> T.