Multi statement, Attach, Detach, Use statements support

[hmeriann]

• Added multi-statement generation to the fuzzyduck.cpp;
• StatementGenerator::RandomValue() and StatementGenerator::RandomPercentage() became public to be used in fuzzyduck.cpp

[Tmonster]

Awesome! Are you maybe missing a couple of files? Or maybe forgot to add some changes?
The PR here has the following files changed https://github.com/hmeriann/duckdb/pull/7/files,

extension/sqlsmith/fuzzyduck.cpp
extension/sqlsmith/include/fuzzyduck.hpp
extension/sqlsmith/include/statement_generator.hpp
extension/sqlsmith/statement_generator.cpp
extension/sqlsmith/statement_simplifier.cpp

while this PR only has the following 2 files changed

src/fuzzyduck.cpp
src/include/statement_generator.hpp

[Tmonster]

Edit: Nevermind, this is just for multi statements, and the PR to add attach & use will come later

[hmeriann]

@Mytherin could you please have a look at this PR?

[Tmonster]

Just the one comment. Otherwise Looks good to me from here

[Tmonster]

Now that duckdb_sqlsmith is an extension and has duckdb as a submodule is it possible to use the TESTING_DIRECTORY_NAME declaration from test/helpers/test_helpers.cpp?

[Tmonster]

you might have to include duckdb/test/include/test_helpers.hpp

[hmeriann]

Defined new TESTING_DIRECTORY_NAME, also put it in the .gitignore

[Tmonster]

Oh shoot, I'm just realizing now that maybe the duckdb_sqlsmith_unittest_tempdir might not actually be a good idea. If people build DUCKDB with the sqlsmith extension, then the duckdb_qlsmith_unittest_tempdir is not in the duckdb .gitignore. This means there will be a bunch of unittest files that github is automatically tracking. I'm going to test this first to make sure I'm right

[Tmonster]

So I think we should stick with "duckdb_unittest_tempdir". This directory is already in the .gitignore. Also, if someone builds sqlsmith in the duckdb directory or in the sqlsmith directory, the databases attached by the attach and use will automatically be in the .gitignore

[Tmonster]

remove the "../" these cause the testing directory to end up in a weird place

[hmeriann]

Removed - now it puts files generated during testing to the proper duckdb_unittest_tempdir

[Mytherin]

Looks good! Perhaps in the future we want to, instead of using USE, directly refer to attached databases in queries? That way we can also fuzz queries that access multiple databases, e.g.:

ATTACH ':memory:' AS db1;
CREATE TABLE db1.t1(i INT);
ATTACH ':memory:' AS db2;
CREATE TABLE db2.t2(i INT);
SELECT * FROM db1.t1, db2.t2;

[Tmonster]

lgtm 👍 . Also agree with Mark that we can just directly refer to persistent databases in the future instead of using USE

[Tmonster]

Have created an issue for fuzzing across databases #2.
Re-running the windows job to confirm failure is due to github-wide issue of replacing windows machines

[Tmonster]

Thanks!