finalize tournament lobby + add variable event cards #908
Conversation
sgfost
commented
Oct 31, 2023
•
sgfost
commented
- resolves virtualcommons/planning#51
- resolves virtualcommons/planning#44
- resolves virtualcommons/planning#52
- resolves verify isLobbyOpen logic #909
keeping track of a queue (all clients not yet put into a group) in addition to pending groups and a simplified client->group mapping makes some of the logic a bit more straightforward. A mutex should be safer than the simpler lock - add timeout to clear a group and try again if all clients havent accepted in a timely manner - misc. minor refactoring/renaming ref virtualcommons/planning#51
| formGroup(): string { | ||
| const group = []; | ||
| const groupId = uuidv4(); | ||
| _.shuffle(this.state.queue); |
There was a problem hiding this comment.
would it be more straightforward to shuffle once and drain all valid groups at once and then leave the unlucky leftovers in the queue?
There was a problem hiding this comment.
Yeah, the queue isn't actually managed correctly, I'll look at this when doing so
| // groups indexed by a unique group id | ||
| pendingGroups: Map<string, LobbyClient[]> = new Map(); | ||
| // map client username to group id | ||
| clientToGroup: Map<string, string> = new Map(); |
There was a problem hiding this comment.
Would a Group class with a fixed size, LobbyClient[] and id simplify things? Then clientToGroup = Map<string, Group>
There was a problem hiding this comment.
Almost certainly, could also just store the group reference on the LobbyClient which I think was how the old tournament lobby did it but that'll either be a case of a really messy extension of the LobbyClient schema class or putting it in the base class to be a dead property elsewhere. The map is probably fine.. only needed for one operation
- remove all logic/private group-forming-related data from the state and put into GroupManager class which holds the waiting queue and set of pending groups - remove group disbanding in favor of sending the group off anyways and assuming that any disconnected player will attempt to join back. timeout was repurposed to do this 'forceful' moving after 15s - further renaming
add optional ?type param to /has-active/ to be able to check whether a player has a specific type of game active and avoid putting them into a freeplay game (that may have been accidentally started) when they are attempting to join a tournament
[no ci]
client/server communication and invitation flow could potentially be checked with the colyseus testing package if we can mock some of the services relied upon: https://docs.colyseus.io/tools/unit-testing - fix queue shuffling
+ some minor lobby room cleanup
sgfost
changed the title
|
Unit tests for the grouping logic + a fair bit of mostly manual smoke-testing have me feeling fairly confident about the lobby working correctly. I'll move onto the other few requirements in order to finish up this week and leave the start of next for further testing |
alee
force-pushed
the
tournament-lobby
branch
from
0e06463
to
c323388
Compare
- fixed an issue where the exit survey was being given on the gameover screen of a freeplay game due to some helper functions not distinguishing between tournament/game types when searching for invites
if treatments for a game's tournament are found, it will cycle through them evenly based on the last game's treatment for that specific tournament - add command for creating treatments
- includes some small fixes based on results
when the dynamic settings gets called in tests it causes everything to hang, as well as not working correctly. A better solution might be to mock the module for the test env with static defaults
bug led to a tournament only knowing about the most recently created treatment
Co-authored-by: Scott Foster <[email protected]>
|
the problem with treatments during the pretest was that I somehow botched applying the changes in 14f8bfd which fixes the issue. Fix for the lobby not staying open long enough is correct as well. Better tests added for both |
Co-authored-by: Scott Foster <[email protected]>
closes virtualcommons#912 closes virtualcommons#913 Co-authored-by: Scott Foster <[email protected]>
alee
force-pushed
the
tournament-lobby
branch
from
2da7b01
to
5002c74
Compare
