Intercell seeding breaks cell magics

[chrispyles]

For cells that use cell magics, intercell seeding may cause errors that prevent those cells from executing. For example, a cell like

%%sql
-- just run this cell --
SELECT * FROM sqlite_master WHERE type='table';

becomes

np_V8HY0L.random.seed(42)
random.seed(42)
%%sql
-- just run this cell --
SELECT * FROM sqlite_master WHERE type='table';

I think the solution here should be just to not insert intercell seeding code for cells that start with %%.

[rohanjha123]

If there is no intercell seeding for magic cells, would we be able to autograde things like the SQL RAND() functions without having to specify a seed in every RAND() call across multiple cells?

[chrispyles]

If there is no intercell seeding for magic cells, would we be able to autograde things like the SQL RAND() functions without having to specify a seed in every RAND() call across multiple cells?

The current implementation of intercell seeding does not support seeding SQL; it only seeds numpy and the built-in random library. So yes, to seed RAND in SQL, you would need to add the seeds yourself.