On Sat, 10 Feb 2018 12:18:56 +0100
l...@gnu.org (Ludovic Courtès) wrote:
> Thinking about it, it wouldn’t matter that HTTP requests are processed
> sequentially if database queries run really fast. I’m not sure if we
> can achieve it. WDYT?
That depends on how fast. But it should be possible to optimize the actual
query (using indices, lookups are O(log N)). Also, if it's the same
query as before, it usually will be really fast as most of the pointers
are still where they were before.
Sqlite3 already automatically created indices for all the primary keys.
There's also https://www.sqlite.org/pragma.html#pragma_optimize if we need it.
We can always try it with serialized database access and use a connection
pool should it get too slow later.