Was there anything that you found to be surprisingly challenging, or was it mostly straightforward?
I think most of the challenge was self-inflicted.. I just sort of programmed as I went, with no planning at all, and so i wound up reimplementing the same concept a few different ways, which made smooshing it all together into the mega SELECT function difficult.
I also was surprised by some of the aspects of writing the macros. It wasn't really clear to me when pieces of the expression get passed in as separate arguments, vs when they come in as a nested structure.
That is, does `@SELECT foo @FROM bar @WHERE a>2` parse as a call to `SELECT(:foo, :@FROM, :bar, :@WHERE, :(a>2))`, or as `SELECT(:foo, :(@FROM(:bar, :@WHERE(:(a>2))))`? I think it was the second one, but I still don't think I really have a good grasp on the pattern.
> Cause it turns out, hey, a join is really simple! All it's doing is basically computing the Cartesian Product of your two tables, and then filtering out rows based on how you specified the query. No big deal.
Was this a simplification or that's how you think JOINs work? In either case, it's an extremely misleading statement.
I think what I meant was something more like "You can think of it as basically just the Cartesian Product of your two tables, and then filtering out rows based on how you specified the query." I'll make that change now.
I recognize that you would want to do the filtering first, of course, and only materialize the final table at the very end after everything else finishes.
Other than that, though, am I missing anything else? It's quite possible due to, as I said, how new I am to relational databases.
Do you use the new iterators to implement the joins? If so, it seems reasonable that you could implement the join/filter as logical separate components but have the filtering happen in-line with the joins during iteration. Then you'd get the best of both worlds. It'd be interesting to see how well the Julia jit can inline that type of code!
Also, if you're just getting into relational databases, definitely take some time to read up on the relational calculus, and relational algebra.
I think parsing wouldn't be hard either (since that's essentially what the macros are doing), I just didn't do it.
In fact, I think probably all the macro parsing code could be reused for a parser!
Do you think it's likely someone will eventually write a full database system in Julia?
It'd be kind of crazy do do it I'm such a high level language, but I wonder... It seems like Julia is pretty fast and capable.
(though it appears to be in need of an update to 1.0 still which is in progress: https://github.com/JuliaComputing/IndexedTables.jl/pull/182)