Fix MySQL database name handling

[ethan-readyset]

Description

When we connect to ReadySet as configured with a MySQL upstream and we specify a database name, the following sequence of operations fails:

CREATE TABLE t1 (x int);
INSERT INTO t1 (x) VALUES (1);

With an error that reads

thread 'test_db_name' panicked at 'called `Result::unwrap()` on an `Err` value: Server(ServerError { code: 1105, message: "Error during RPC (ReadySetHandle::table): Could not find table 't1'", state: "HY000" })'

If we fully qualify the table name with the database name in the INSERT INTO statement (e.g. INSERT INTO database_name.table_name …), the statement succeeds. It seems that ReadySet is mishandling the database name when it is connected to a MySQL upstream.

Expected behavior

Connecting to ReadySet using a database name, creating a table, and inserting a row without qualifying the table name with the database name succeeds.

Actual behavior

Explained above

Steps to reproduce

A minimal failing test can be found here: https://rs-paste.tail422d.ts.net/?6cf501bf73ab6147#EQdRXjUy4vg8fcEtz6mBmR1nSNCuHMbcT3r2KRKKiXLm

This test should be placed in the readyset-mysql/tests/integration.rs file

ReadySet version

commit hash 211fc79d877b695149d49c29f9251daef1ce0e46

Upstream DB type and version

N/A

Instance Details

N/A

Logs

Relevant logs listed above

[ethan-readyset]

I'm going to backlog this ticket since we determined that it's only affecting tests. Does that sound good to you @tamas Juhasz? Or were you planning to fix it?

[ethan-readyset]

When the database name is not provided in the query the table schema is empty and this causes the query address lookup to fail. The message that the table does not exist is somewhat misleading as it does exist but if the database is not provided it can't be found. In case an upstream has been defined the failed query will be sent to the upstream where it would succeed and no error message would be returned.  

One possible fix for this would be to set the Schema with the one that ReadySet is serving in case it is empty. However, since this only happens in the unit tests there is no urgency in fixing it.

[ethan-readyset]

Lower priority for now, as we are focusing on Postgres.

[ethan-readyset]

Nice find!