Prepared statements exist for good reasons. That includes performance (if the same query it executed multiple times, often with different values) and security (proper escaping is hard!).
In some cases placeholder substitution makes sense, but this should be an optional feature.

Maybe it should be just an exported helper function, which allows the user of the driver to build an parameter-less query, which then can be executed by db.Exec or db.Query. Because of the Execer / Queryer interface implementation, this also saves the extra round-trips.

This would be a good feature. We wrote about our adventures with it here:
https://vividcortex.com/blog/2014/11/19/analyzing-prepared-statement-performance-with-vividcortex/

Question: if you could enable such a setting, perhaps in the DSN string,
could the driver do it internally so you wouldn't need to alter the
interface or calling code in any way? Or does the driver not get a chance
to intercept and "emulate" such calls?

@julienschmidt Placement holder substitution takes some config from mysqlConn: mc.cfg.loc, mc.maxPamaxPacketAllowed and mc.status I've added in this pull request.
Exposing helper function is possible, but integrating it is safer and usuful.

@xaprb I think adding an option to DSN is good idea, while I don't think PHP's "EMULATED PREPARE" is good word.

Another use: SphinxQL, which does not support prepared statements at all (#221).
We should maybe investigate if we can somehow find out during connecting if the server supports prepared statements.

Try to prepare

SELECT 'go driver testing prepared statements'