I tend to favor @Ankhers first approach as opposed to extracting the hour out in the query. That’s simply because the database must now run that operation on every row. If you have an index on the time column, you won’t be able to use it, unless the index is for the specific operation you’re performing.

In short, favor doing the work once, rather than many times.

The second approach could work fine if you had an expression index for it. But it would have to be a separate index from an index directly on the time column. So, you could end up with two potentially.

All of this is completely dependent on the situation though. If you have some other column you also always query on, it may reduce the result set, such that either solution is perfectly fine. I tend to avoid most indexes until it’s obvious they’re necessary. Before, I would try to pre-empt what indexes would be useful and later find most of them are unused because another one gave better performance, so the query planner always picked it. Having unused indexes affects write performance without any gain. Just watch your slow query log and you’ll be able to see what needs optimizing.

The efficiency of that query depends on your database, not on your code. If you are so concerned, add a DB index to the column holding the timestamps.

Never do premature optimization.