Time  Nick      Message
09:27 sfan5     merging #16253, #16247, #16281 in 10m or so
09:27 ShadowBot https://github.com/luanti-org/luanti/issues/16253 -- Add documentation index for `doc/` as `README.md`, and small docs fixes by Xeno333
09:27 ShadowBot https://github.com/luanti-org/luanti/issues/16247 -- ContentDB: Allow hitting "Install" before package info finishes loading by grorp
09:27 ShadowBot https://github.com/luanti-org/luanti/issues/16281 -- Revert "Detect mouse moving out of inventory slot (#16101)" by Desour
11:26 sfan5     can someone review #16265? the non-benchmarking changes are literally less than 10 LOC
11:26 ShadowBot https://github.com/luanti-org/luanti/issues/16265 -- Make `core.get_node_raw` a public API by sfan5
11:29 [MatrxMT] <grorp> hmm, would it make sense to include the benchmark with the 50% improvement in the PR, if that case is the justification for it?
11:29 [MatrxMT] <grorp> the paste is also dead
11:33 sfan5     dunno
11:39 sfan5     IIRC the benchmark was generating random positions and counting the number of grass and stone nodes, by comparing directly with the content ids
11:39 sfan5     it's logical that it would be faster
11:40 MTDiscord <jordan4ibanez> 👀
11:40 MTDiscord <jordan4ibanez> This will be extremely useful when 5.13.0 is released
11:41 sfan5     hm but actually it's not that different from what devtest does
11:41 sfan5     well whatever
11:59 sfan5     planning to push https://github.com/sfan5/luanti/commit/ca94ee909b7d5ad28e04787349b80446bec7ef28 in 10m
11:59 sfan5     (with typo in commit msg fixed)
12:01 MTDiscord <exe_virus> Is there an issue that exists to make get_vmanip_raw() ?  I would assume similar or greater speedups in the high performance cases. I.e. I'd rather expose the full param0, param1 and param2 (32 bit number) to Lua as a single value for each node. Instead of our current access and representation pattern that uses a Lua value for each of these separately.   Obviously would need benchmarking but intuitively it must be faster
12:05 sfan5     you mean an alternative to VManip:get_data()? how much faster is that going to be if Lua has to fiddle with bitops to actually make any use of the returned values?
12:07 sfan5     how often do mods actually want param1 and param2? returning more data that's not needed is the opposite of faster
12:15 MTDiscord <exe_virus> Understand, but in this case, it would be available.   My thinking is the ability to set both param0 and param2 in one shot. Agree param1 is less commonly changed.   This is for mapgen speed optimization for pure Lua mapgens.   Most people are doing pure node mapgens so directionality matters a lot less (param2). But my core goal is to make slopes and rocks and other non-node shapes in a game specific mapgen, and raw writing in those
12:15 MTDiscord cases (cache the values), would be much faster.