SL21W41: Difference between revisions

From ETM
Jump to navigation Jump to search
No edit summary
No edit summary
 
(13 intermediate revisions by the same user not shown)
Line 7: Line 7:
***Got rid of most of the special caching / update flags for desktop this way
***Got rid of most of the special caching / update flags for desktop this way
***Code reduction as the general purpose cache now works nicely
***Code reduction as the general purpose cache now works nicely
***'''Add a quick board comparison/equality check by comparing screen-size first (+ tile element count second)? Needed? The BoardHash should do this already right?'''
***Add a quick board comparison/equality check by comparing screen-size first (+ tile element count second)? Needed? The BoardHash should do this already right?
**Fix: Screen not refreshing on first open (Land manager)
**Removed '''current money''' from screen as it makes caching nearly impossible.
***'''=> Always changing information should not be part of Header/Footer'''
***=> "Adspace" on top could now be a seperate screen !
***=> "Adspace" on right could now be a separate screen !
***Separation of these spaces would make caching much more effective (board / pre render caching)
**Fix: Plant interaction
***Fix hitbox positioning of plants
***Fix hitbox count for cubic plants
***Fix screen position interaction on cubic plants
***Added pixel-perfect interaction for plants (transparent interactions are ignored)
***Hitbox shadows/Plant shadows are annoying/random? can we disable those?
****! We can only show F3 Hitboxes + Shadows OR NO F3 hitbox and no shadows!
*Nils intro to Kadi / Chunkli design
*Nils intro to Kadi / Chunkli design
**'''General setup of Stasi-Plugin, based on Chunkli Api'''
**General setup of Stasi-Plugin, based on Chunkli Api
***'''Think/Test persistence layer'''
***Think/Test persistence layer
***'''Make sure we respect shulker boxes in some way?'''
***Make sure we respect shulker boxes in some way?
***'''95% of data is archival data mostly? Important is mostly just worldchanges ~7 days + chat ~3 month; rest can be slow / compressed?'''
***95% of data is archival data mostly? Important is mostly just worldchanges ~7 days + chat ~3 month; rest can be slow / compressed?
***'''Most likely PostgreSQL + time-series => Timescale db? on an external logging server with batched, async, streamed inserts?'''
***Most likely PostgreSQL + time-series => Timescale db? on an external logging server with batched, async, streamed inserts?
****Specifically the data compression would be handy here for non-confirmant itemstacks where all data needs to be logged and not just a LibId
****"Small datasets typically see no performance advantage on TimescaleDB vs PostgreSQL (in fact, they might be slightly slower)"
****"The queries run (they should typically be time-oriented restriction clause)"
***We need to test couchbase for this
****Assuming the search/query times are 'ok' and the datasize is 'ok' as well we should probably opt for having only one DB as long as possible to reduce complexity
****Even "optimized" documents like (https://blog.couchbase.com/the-series-on-time-series-in-couchbase-episode-1/) are probably not worth it until far later in the project.
**Test Geospatial as a save/search format
***https://docs.couchbase.com/server/current/fts/fts-supported-queries-geo-spatial.html
***https://blog.couchbase.com/how-to-geospatial-polygon-search/
***https://blog.couchbase.com/geospatial-search-how-do-i-use-thee-let-me-count-the-ways/
***We can use 0.001 as one block in lat/lon
***1.000 would be the block at X 1000
***Y would be altitude, not directly searchable in geospatial data typically
***A blockchange at X -54_321 and Y 150 and Z 765 would be saved as { lat: -54.321, lon: 0.765, alt: 150 }
 
 
Stasi couchbase result => Works super well, Geospatial does not really help.

Latest revision as of 23:23, 3 November 2022

◄ SL21W40 | SL21W42 ►


  • Cleanup Week 2
    • Completely remove Desktop-special case as a screen, use dynamically created ScreenModel internally to reduce complexity
      • Got rid of most of the special caching / update flags for desktop this way
      • Code reduction as the general purpose cache now works nicely
      • Add a quick board comparison/equality check by comparing screen-size first (+ tile element count second)? Needed? The BoardHash should do this already right?
    • Fix: Screen not refreshing on first open (Land manager)
    • Removed current money from screen as it makes caching nearly impossible.
      • => Always changing information should not be part of Header/Footer
      • => "Adspace" on top could now be a seperate screen !
      • => "Adspace" on right could now be a separate screen !
      • Separation of these spaces would make caching much more effective (board / pre render caching)
    • Fix: Plant interaction
      • Fix hitbox positioning of plants
      • Fix hitbox count for cubic plants
      • Fix screen position interaction on cubic plants
      • Added pixel-perfect interaction for plants (transparent interactions are ignored)
      • Hitbox shadows/Plant shadows are annoying/random? can we disable those?
        • ! We can only show F3 Hitboxes + Shadows OR NO F3 hitbox and no shadows!
  • Nils intro to Kadi / Chunkli design
    • General setup of Stasi-Plugin, based on Chunkli Api
      • Think/Test persistence layer
      • Make sure we respect shulker boxes in some way?
      • 95% of data is archival data mostly? Important is mostly just worldchanges ~7 days + chat ~3 month; rest can be slow / compressed?
      • Most likely PostgreSQL + time-series => Timescale db? on an external logging server with batched, async, streamed inserts?
        • Specifically the data compression would be handy here for non-confirmant itemstacks where all data needs to be logged and not just a LibId
        • "Small datasets typically see no performance advantage on TimescaleDB vs PostgreSQL (in fact, they might be slightly slower)"
        • "The queries run (they should typically be time-oriented restriction clause)"
      • We need to test couchbase for this
    • Test Geospatial as a save/search format


Stasi couchbase result => Works super well, Geospatial does not really help.