2周前

Sure, if you put it like that...

Absurdity @lemmy.cafe

FundMECFS @lemmy.blahaj.zone

1周前

Dramatic Database Manipulation

33 1

您正在查看单个对话串。

78 评论

The opposite actually - rows are dramatically added to a database. In most games save files grow the longer you play.
- and even if some idiot put every zombie npc in a database (or if you want to think of it that way), you wouldn't just delete the rows! the bodies would disappear, so instead you would update that row like (npcState = KIL, bodyLocation =
  <some coords>
  ) or something. Especially if you wanted to keep player stats
  
  Isnt there this graveyard off map somewhere in Skyrim, where all the bodies get teleported?
  
  Why would that even be necessary? Sounds like one of those “make a guy with a train for a hat and run up and down this hall” moves they like to do
  
  Their code is literally spaghetti
  
  l-.. literally?
  
  I wouldn't be surprised if it is true though. Bethesda games are not exactly winning awards for coding elegance.
  
  Technically, the train was done by the Obsidian studio, not Bethesda per se, because of strict deadlines.
  
  If they literally don't have an object delete option then relying on render distance to make it go away is a ridiculous but simple solution
  
  i can't speak to that, but it sounds plausible. in that case the body location would be updated to those coordinates
  
  Grab the source here before Bethesda DMCAs it.
  
  npcState = KIL
  no
  
  Where was you when
  
  This is how it works in many game engines.
  You set up the monsters and just hide them/disable them. They're already allocated to memory.
  And it's a performance cost to create/delete versus just moving a dead enemy out of view, then respawning that enemy later in the level.
  
  im not very versed in game engine design, but dont they dynamically stream them into memory before they will be needed, and discard them when they wont nowadays?
  
  Dynamic streaming is common nowadays, as games have gotten large enough that not everything in a level can fit into memory.
  I don't know about what is actually done in industry but I feel like most of the time you wouldn't bother with keeping dead instances unless instancing is shown to actually be a performance problem, which will probably not happen all that often
  Godot for example doesn't have built in dynamic level streaming yet or a built in way to cycle through dead instances as far as I can tell, although I'm sure that wouldn't be hard to do with code
  
  If it's a type of enemy you see just one of at a time but see it often, sure. If there's many, cost of copy/delete is definitely not that high relatively speaking.
  (random sidenote: in the first Mirror's Edge game, you can sometimes hear enemies you passed scream as they fall when you pass from one part of a map to another, as the ground in the map is unloaded before the enemies unload)
  
  I want the rows deleted. I'm going to market it as the first game with true AI/enemy permadeath. Dibs on the idea!
  
  Maybe you would have an array of active enemies in RAM, and when enemies are killed they are removed from that array for example?
  In a game like Minecraft for example, you definitely wouldn't want to store every single dead entity and its location when there can easily be thousands created and destroyed in a single second
  It obviously depends on the game though.
  
  Definitely depends on the type of game, but it's more likely the game stores data about which areas you cleared and then infer that the bodies of any permanently remaining enemy (like bosses) is to be displayed.
  Can vary even more for procedurally generated levels. If the set of enemies is fixed and stay in calculated positions in a map generated randomly, then it might store an array or something tracking the enemies.
  
  Procedurelly generated stuff is all about storing the differences from the procedural generation.
  So for example minecraft saves don't store the terrain, they store how it differs (due to player interaction) from the procedurally generated baseline.
  (After all, all you need to recreate an untouched procedurally generate world are the bytes of the seed and nothing else)
- This is why Breathe of the Wild did the blood moon thing, periodically they'd just bring all the dead enemies back so file size didn't get too large.
- Also, it's an unreasonably fast database. That makes lots of trade-offs that normal ones aren't willing to do.
- Noita file save on the 7th parallel world intensifies
- Me in the matrix (so irl basically), holding a gun: "Don't worry, I'm not deleting you!"
- There for a minute when Dyson Sphere Program first went into open pre-release, something was wrong with their save file compression, and very quickly people were reporting multiple GB saves.
- I was looking at the savegames from the game control recently, it's kinda funny because you open them in notepad, you see a bunch of random gibberish from bad decoding (the game uses a proprietary save format) with the words "collected" "Collected" "unlocked" "available" "VariableRestoreHack" (??) "STATE_B_PUZZLE_SOLVED" "Powercore_Not_Attached" randomly interspersed
  Like, surely there is a better way to store 2 state data other than an english word?
  It does generally get longer as you play, but also "locked" just switches to "unlocked" for example when you unlock something
  
  Eh, really depends
  They are likely just serializing a bunch of data objects. And set states and flags with humans readable enums
  Enums make code a lot easier to read, especially if you use it to check stuff all over the place
  Using to a couple bytes more storage is worth it
  
  I thought that enums were supposed to compile down to aa, ab, ac when you actually build the game.
  
  Depending on the language, they are.
  Lots of games also use data structures derived from tables/csv at runtime to configure things like stats. So they would also need these (human readable) values in the save files
  Hard to tell from speculation and not having the data
  
  Depends on optimization levels, data types, and whatnot. If it's a string to be fed into the API of a different binary then the compiler will often not optimize down that representation. Internal function names are likely to be optimized that way, with lookup tables holding original function names (at least for any externally exposed function).

78 评论