Happens to address part of https://github.com/matrix-org/matrix-public-archive/issues/271
but made primarily as a follow-up to https://github.com/matrix-org/matrix-public-archive/pull/239
---
Only 42% rooms on the `matrix.org` room directory are `world_readable` which means we will get pages of rooms that are half-empty most of the time if we just naively fetch 9 rooms at a time.
Ideally, we would be able to just add a filter directly to `/publicRooms` in order to only grab the `world_readable` rooms and still get full pages but the filter option doesn't allow us to slice by `world_readable` history visibility.
Instead, we have to paginate until we get a full grid of 9 rooms, then make a final `/publicRooms` request to backtrack to the exact continuation point so next page won't skip any rooms in between.
---
We had empty spaces in the grid before because some rooms in the room directory are private which we filtered out before. But that was a much more rare experience since only 2% of rooms were private .
The history visibility in Libera rooms is set to `join` which means it's not
accessible in the archive at all. Instead of leading a bunch of people to
`403 Forbidden`, we can just remove it from the default list.
The default list was mostly just copied from the Element list of defaults.
Room cards will now sort by room members descending (highest to lowest) as expected.
Fix https://github.com/matrix-org/matrix-public-archive/issues/218
The `/publicRooms` (room directory) endpoint already returns rooms in the correct order which is why we didn't care about the order before but the different `[].sort(...)` implementations in browsers necessitates we be explicit about it. Ideally, we wouldn't have to use the `ObservableMap.sortValues()` method at all but it seems like one of the only ways to get the values out. In any case, maybe it's more clear what order things are in now.
This bug stems from the fact that `[1, 2, 3, 4, 5].sort((a, b) => 1)` returns different results in Chrome vs Firefox (found from https://stackoverflow.com/questions/55039157/array-sort-behaves-differently-in-firefox-and-chrome-edge)
- Chrome: `[1, 2, 3, 4, 5].sort((a, b) => 1)` -> `[1, 2, 3, 4, 5]` ✅
- Firefox: `[1, 2, 3, 4, 5].sort((a, b) => 1)` -> `[5, 4, 3, 2, 1]` ❌
This helps when someone just pastes a room alias on the end of the domain,
- `/#room-alias:server` -> `/r/room-alias:server`
- `/r/#room-alias:server/date/2022/10/27` -> `/r/room-alias:server/date/2022/10/27`
Since these redirects happen on the client, we can't write any e2e tests. Those e2e tests do everything but run client-side JavaScript.
Follow-up to https://github.com/matrix-org/matrix-public-archive/pull/107
Part of https://github.com/matrix-org/matrix-public-archive/issues/25
Also does friendly redirects if you don't exactly use the right URL pattern.
For example, if you paste the full room ID with the `!` like `/roomid/!foo:bar`,
it will properly redirect you to `/roomid/foo:bar`. It also does this sort of
thing for URL encoded room ID's and aliases.
Fix https://github.com/matrix-org/matrix-public-archive/issues/25
Page-load with the correct homeserver selected (according to `?homeserver`).
Fix https://github.com/matrix-org/matrix-public-archive/issues/92
Also makes sure that the `?homeserver` is always available somewhere in the list; whether that be in the available homeserver list or the added homeserver list depending on it someone cleared it out or never had it because they visited from someone else's link.
Eric Eastwood