* One pass at materialized views, only about 30% faster, not good.
* Before merging master to test out bans.
* DB Rework working, still need more testing.
* Fixing accidental addadmin bug from asonix async merge.
* Fixing the comment delete trigger
* Some more DB additions.
- Adding a hot_rank desc, published desc index to post_aggregates_fast.
- Removed WITH CTE queries in favor of direct selects (since CTEs cant
use indexes)
* Removing some unecessary indexes.
* Some more DB optimizings
- Changing the fast_id pkeys to just ids on the fast tables.
- Removing the private_message_fast, since the view contains no aggregates.
- Comment and post voting now no longer pull from the views, they update the counts directly.
* Adding community_agg_view and post_agg_views Credit: eiknat.
* Adding user and comment_view migrations. (comment_view still broken)
* Adding more views. Credit Eiknat.
* Asyncify more
* I guess these changed
* Clean PR a bit
* Convert more away from failure error
* config changes for testing federation
* It was DNS
So actix-web's client relies on TRust DNS Resolver to figure out
where to send data, but TRust DNS Resolver seems to not play nice
with docker, which expressed itself as not resolving the name to
an IP address _the first time_ when making a request. The fix was
literally to make the request again (which I limited to 3 times
total, and not exceeding the request timeout in total)
* Only retry for connecterror
Since TRust DNS Resolver was causing ConnectError::Timeout,
this change limits the retry to only this error, returning
immediately for any other error
* Use http sig norm 0.4.0-alpha for actix-web 3.0 support
* Blocking function, retry http requests
* cargo +nightly fmt
* Only create one pictrs dir
* Don't yarn build
* cargo +nightly fmt
- Advanced code migrations now disable then re-enable triggers.
Brings run time down to < 15 seconds, no need to thread them.
- Changing ap_ids and actor_ids in migrations to a fake url,
so it doesn't break XsdAnyUri in activitystreams.
- HTTP and APUB clients can now send live updating messages to websocket
clients
- Rate limiting now affects both HTTP and websockets
- Rate limiting / Websocket logic is now moved into the API Perform
functions.
- TODO This broke getting current online users, but that will have to
wait for the perform trait to be made async.
- Fixes#446
