Added test coverage for DatabaseOperations.sql_flush().

[jdufresne]

No description provided.

[charettes]

I wonder if it would be better to write a backend agnostic test that executes the provided SQL and ensures tables are flushed and sequences are reset. I'm not sure there's a lot of values in asserting against the exact SQL generated since we offer no stability guarantees in this regard.

[jdufresne]

I wonder if it would be better to write a backend agnostic test that executes the provided SQL and ensures tables are flushed and sequences are reset.

I can work towards that.

I'm not sure there's a lot of values in asserting against the exact SQL generated since we offer no stability guarantees in this regard.

I think it provides value. It shows exactly (modulo some ordering) the output of the function. We can inspect this output directly to ensure the commands that we expect to render, are returned. A change that unexpectedly changes these statements -- perhaps causing a regression -- will be noticed.

One of the motivations of this PR is that I would like to change PostgreSQL sql_flush() to change the sequence resetting logic from multiple SQL statements to use the RESTART IDENTITY keyword of the TRUNCATE statement. My project uses flush operation internally and some profiling shows that it can consume a large percentage of the runtime. As these functions have minimal test coverage so far, this test adds a minimal base level of assurance before any changes take place. Asserting the function's exact return value will show me that my upcoming change to use RESTART IDENTITY will be implemented.

I do agree there is additional value in executing the SQL as well. I think they would complement one another.

[charettes]

change the sequence resetting logic from multiple SQL statements to use the RESTART IDENTITY keyword of the TRUNCATE statement.

Oh I looked at it as well while helping with the MySQL adjustments that took place in this area but I assumed it was only working with identity columns and thus was blocked by #30511. I must have misread the PostgreSQL documentation since it clearly state it will reset sequences as well.

I don't see much harm in testing the exact SQL as well in this case.

[jdufresne]

write a backend agnostic test that executes the provided SQL and ensures tables are flushed and sequences are reset.

👍 I added ExecuteSqlFlushTests for this.