From 8eeda4aa133e6aa3e068b1a9f7d31cd9001475d4 Mon Sep 17 00:00:00 2001 From: Ben Sima Date: Wed, 24 Jan 2018 18:31:41 -0800 Subject: Fix the slow bootstrapping with `groupUpdates` Fixes #1 --- readme.md | 21 +++------------------ 1 file changed, 3 insertions(+), 18 deletions(-) (limited to 'readme.md') diff --git a/readme.md b/readme.md index 31e0edf..b7b9c4e 100644 --- a/readme.md +++ b/readme.md @@ -2,22 +2,7 @@ 2. Optional: have [nix](https://nixos.org/nix/) installed. If you *don't* use nix, delete line 3 of `main.hs`. 3. Run `./main.hs` and the server will startup -To bootstrap the db, do `curl -XPOST "localhost:3000/bootstrap"`. +To bootstrap the db, do `curl -XPOST "localhost:3000/bootstrap"`. This takes +like 3 seconds or so. -This takes a while; on my machine it averages 181 records per second. It's IO -bound, and in an un-optimized program GHC on Linux uses a single, blocking IO -manager thread (on Windows it's non-blocking, [apparently][1]). This can be -improved with the [Control.Concurrent][1] module, it which case we could launch -as many IO threads as we want, and do probably 10k records per second. There's -definitely an optimal amount of threads here, we'd have to test to find that. - -*However*, you can watch it bootstrap. In a separate terminal, do `curl -"localhost:3000/count"` to see it updating. New POSTs will also work and update -the database, even while it is bootstrapping, which is kinda cool. - -Try this in bash: - - while sleep 1; do curl -s "localhost:3000/count" | jq '.count'; done - - -[1]: https://www.stackage.org/haddock/lts-10.3/base-4.10.1.0/Control-Concurrent.html#g:10 +See `test.http` for some other examples. -- cgit v1.2.3