🔎Julia Evans 🔍Verified account

Does `SELECT` really come before `ORDER BY` (either in practice or even theoretically)? Then you need to evaluate the _complete result set_ before sorting and serving it?

definitely in practice the order of all these things is different -- i think it might be more correct to think of it as order by -> select because you can order by something that isn't in the select

This is really great! Just wondering about one thing: given the query doesn’t seem to use the ‘name’ column at all, I’m wondering if the names could/would be optimised out earlier in the sequence? It’s great to look at it like this to try different mental models!

yeah, what actually happens when the database runs the query is often really different from this (If you want to actually know, you can get the db to tell you its query plan!)

I am wondering if HAVING comes before the SELECT where the collapsing for the GROUP BY happens in your viz. If the aggregation is complex, it should calculate it (therefore collapse it) before sub setting, no?

no, that was on purpose! You can filter by something in HAVING that isn't in your SELECT (like select count(*) from x group by z HAVING max(thing) = y)