00:02 -!- hf_guest_27094 has joined #hardfought 00:03 -!- hf_guest_27094 has quit [Client Quit] 00:04 [hdf-us] [xnh] Luxidream (Pri Hum Mal Law) killed the Wizard of Yendor, on T:49406 00:09 [hdf-us] [xnh] Luxidream (Pri Hum Mal Law) entered the Planes, on T:51361 00:09 [hdf-us] [xnh] Luxidream (Pri Hum Mal Law) killed the Wizard of Yendor, on T:51362 00:13 [hdf-us] [xnh] Luxidream (Pri Hum Mal Law) killed the invisible Wizard of Yendor, on T:51714 00:13 [hdf-us] [xnh] Luxidream (Pri Hum Mal Law) killed the Wizard of Yendor, on T:51714 00:13 go @luxidream 00:14 [hdf-us] [xnh] Luxidream (Pri Hum Mal Law) killed Pestilence, on T:51767 00:15 [hdf-us] [xnh] Luxidream (Pri Hum Mal Law) killed Pestilence, on T:51797 00:15 [hdf-us] [xnh] Luxidream (Pri Hum Mal Law) killed Death, on T:51814 00:15 [hdf-us] [xnh] Luxidream (Pri Hum Mal Law) killed Death, on T:51833 00:16 [hdf-us] [xnh] Luxidream (Pri Hum Mal Law), 2237232 points, T:51836, ascended https://www.hardfought.org/userdata/L/Luxidream/xnethack/dumplog/1528491687.xnh.txt 00:16 grats luxidream! 00:16 <@luxidream> thanks! 00:16 <@luxidream> not a bad variant 00:20 [hdf-us] [gh] aosdictj (Val Hum Fem Law) killed the invisible Demogorgon, on T:34743 00:20 yesssssssss 00:20 Junethack: Luxidream ascended a game of xNetHack on www.hardfought.org! 00:21 epic battle to the death on the causeway, spells versus repeatedly thrown mjollnir 00:29 Junethack: Luxidream ascended a game of xNetHack on www.hardfought.org! 00:45 [hdf-us] [dnh] Silver2195 (Val Dwa Fem Law), 188 points, T:568, killed by a falling rock 00:59 [hdf-us] [slex] Silver2195 (Mur Lis Mal Cha), 114 points, T:470, killed by a fatal weakness 01:27 -!- greqrg_ has quit [Quit: Leaving] 01:33 Wiped out all worm that walkses. 01:35 !tell antigulp small pluralisation issue - "worm that walkses" (this is in 0.3.0, so you might've fixed it already). 01:35 Will do, Tangles! 01:41 Is avalanche in slex the spell version of ?oEarth? 01:41 (Not that I'll ever get to cast it, but) 01:48 [hdf-us] [spl] Silver2195 (Car Mer Fem Neu), 218 points, T:835, killed by a death maggot 02:01 <@luxidream> !tell bhaak my xnethack ascension went in twice 02:01 Will do, @luxidream! 02:02 <@luxidream> !tell bhaak contributes nothing extra to the clan but is a bug nevertheless. it shows up correctly on scoreboard 02:02 Will do, @luxidream! 02:33 -!- ais523 has joined #hardfought 02:38 -!- raisse\splat has joined #hardfought 02:38 -!- mode/#hardfought [+v raisse\splat] by ChanServ 02:43 -!- Grassy has quit [Remote host closed the connection] 02:49 -!- Grasshopper has joined #hardfought 02:49 -!- mode/#hardfought [+v Grasshopper] by ChanServ 02:50 -!- Grasshopper has quit [Client Quit] 02:51 -!- Grasshopper has joined #hardfought 02:51 -!- mode/#hardfought [+v Grasshopper] by ChanServ 02:53 -!- Grasshopper has quit [Remote host closed the connection] 02:53 -!- Grasshopper has joined #hardfought 02:53 -!- mode/#hardfought [+v Grasshopper] by ChanServ 02:56 -!- ais523 has quit [Quit: sorry for my connection] 02:56 -!- ais523 has joined #hardfought 03:00 -!- Chris_ANG has quit [Ping timeout: 260 seconds] 03:07 -!- stenno\ito has joined #hardfought 03:12 -!- stenno\ito has quit [Ping timeout: 256 seconds] 03:44 [hdf-us] [spl] winny (Val Dwa Fem Law) completed Sokoban, on T:12523 03:50 -!- tungtn has quit [Ping timeout: 260 seconds] 03:50 -!- tungtn has joined #hardfought 03:53 [hdf-us] [nh343] oh6 (Ran Elf Mal Cha) wished for "blessed fixed greased +3 gray dragon scale mail", on T:27261 03:59 silly ole gremlin stole my fire resist lol 04:05 -!- oh6 has quit [Quit: Leaving] 04:30 -!- raisse\splat has quit [Ping timeout: 264 seconds] 05:00 -!- Mandevil\splat has quit [Quit: leaving] 05:11 -!- rebatela has joined #hardfought 05:20 -!- noty has quit [Quit: WeeChat 2.0.1] 05:23 -!- stenno\ito has joined #hardfought 05:29 -!- stenno\ito has quit [Ping timeout: 276 seconds] 05:32 [hdf-us] [un] richards (Val Hum Fem Law) killed Lerignac on T:7977 05:47 !tell AmyBSOD\srn wtf is this https://i.imgur.com/2M51NlP.png 05:47 Will do, Lymia! 05:47 !tell AmyBSOD\srn I've been getting variants of the vanilla sokoban with hidden holes and requiring all boulders to clear, unmapped sokobans with invisible boulders and... this. 05:47 Will do, Lymia! 05:48 <@luxidream> + 05:51 Lymia: oh good, for a moment I was horrified it was one of mine 05:51 AFAICT AmyBSOD decided to put all known Sokoban levels into her game, including the ones which were originally designed as jokes 05:52 that looks vaguely like one of jonadab's, but the ones that look like that were designed as "bonus levels" that weren't part of the normal route through the game and were just there for people who liked Sokoban 05:52 (invisible boulders, though? :-D) 05:52 Not auto-revealed. 05:53 https://i.imgur.com/Wzp3qfn.png 05:53 I'm stuck 05:53 I can't continue without mapping or finding the map 05:54 # evil patch idea by jonadab - full-screen-sized sokoban level 05:54 [hdf-us] [un] richards (Val Hum Fem Law), 18372 points, T:9076, killed by a bolt of lightning 05:54 It's a legitimate sokoban level 05:54 Except 05:54 It's not automapped 05:54 :c 05:55 I think it's Metropolitan Gallery of Boulder Art: http://jonadab.jumpingcrab.com/nethack-stuff/spoilers-fourk/sokoban/nonfourk-levels.html 05:55 having the actual layout may help (assuming that slex hasn't changed it) 05:57 "The Evil Variant has never been publicly distributed"? 05:57 Curious what *that* is 05:57 there's a lot of history, and it's rather morphed over time 05:58 it started out as a joke made during a Let's Play, basically of making an intentionally bad NetHack variant as a parody of how variants of other games often end up (i.e. ridiculously difficult, missing the original point of the game, poorly tested and coded, etc.) 05:59 then it turned into a place where bad ideas that people really wanted to implement for some reason could be sent to keep them out of more viable variants (particularly AceHack) 05:59 then people started making YANIs specifically for it 05:59 "Prize: three chests full of random chest contents" 05:59 it originally got abandoned after GruntHack started outdoing it 05:59 oh 05:59 great 05:59 *chests* 05:59 The bane of wisps 05:59 :( 06:00 Lymia: that may have changed for slex, the prize was originally intentionally bad so that you'd only do it if you really wanted to 06:00 anyway, slex not only started outdoing both the evil variant and grunthack in arbitrary evilness 06:00 it also started taking patches from it directly 06:00 thus kind-of negating its purpose 06:01 if you're interested in learning more, http://jonadab.jumpingcrab.com/nethack-stuff/evil-patch-ideas/ is a compilation of the YANIs in question 06:02 (although that also includes many false positives, also "YASIs" which are more silly/stupid than arbitrarily evil) 06:03 anyway, the original vision of the Evil Variant was quite specific, whereas the modern version is pretty much "anything goes2 06:03 s/2/"/ 06:04 aaah :D 06:05 one of the big problems is that slex is essentially the sort of game we were trying to parody 06:06 it's not exactly the same: not quite as bad in some ways, worse in others 06:06 but close enough to really make the original idea need rethinking 06:07 EPI: player-character monsters can #quit, ending your game. ← in response to the comment who said this should be in FIQHack: it should end /their/ game and they should just vanish for no apparent reason 06:07 -!- AmyBSOD\srn has joined #hardfought 06:07 AmyBSOD\srn: what changes did you make to Metropolitan Gallery of Boulder Art when putting it in slex? 06:08 It appears to have an upstairs 06:08 -!- Mandevil has joined #hardfought 06:08 -!- mode/#hardfought [+o Mandevil] by ChanServ 06:08 hi ais523! I guess you mean that full-screen soko level? well I had to make adjustments to allow >127 boulders and holes 06:08 AmyBSOD\srn: Message from Lymia at 2018-06-10 05:47 EDT: wtf is this https://i.imgur.com/2M51NlP.png 06:08 AmyBSOD\srn: Message from Lymia at 2018-06-10 05:47 EDT: I've been getting variants of the vanilla sokoban with hidden holes and requiring all boulders to clear, unmapped sokobans with invisible boulders and... this. 06:08 I didn't know that was a problem :-D 06:09 ais523, it's okey 06:09 Lymia welcome to slex :) it's all working as intended. But I believe in you. You can clear those levels! 06:10 if I were to attempt that level, I'd be doing it in a simulator with an undo button :-P 06:10 Some day, I'll get started on a dumb variant when I either decide the C isn't so horrifying as it seems, or decide to be crazy and reimplement Nethack in $GOOD_LANGUAGE 06:10 :D 06:10 Lymia: serious answer: the boulder in the south-east corner of the room you're in appears to safely be movable down three 06:10 but you can't know that without mapping (or looking up the original source) 06:11 Yeah, that was my thought too 06:11 err, down four, not three 06:11 three is safe too but doesn't accomplish much 06:11 r - a scroll called earth {0} 06:11 u - a scroll called earth {0} 06:11 ^^^^^ a - force bolt 1 attack 0% 700% 06:11 probably ultimately insufficient 06:12 the best way is probably to look at sokoban.des which tells you where all the boulders, holes and walls are ;) 06:12 * using coordinates 06:13 I wonder how hard it'd be to write a Sokoban level generator that's capable of generating puzzles solvable without auto-map. 06:13 (i.e. fog of war sokoban) 06:15 https://i.imgur.com/wKqsvEE.png 06:15 well 06:24 ais523: you should actually play slex, then you'd see that despite all the EPIs I implemented, it's not nearly as bad as you apparently think it is ;) 06:24 <@Winsalot> Hey Amy 06:24 @Winsalot: Message from AmyBSOD\srn at 2018-06-08 18:23 EDT: now I have more slex trophies than you :D 06:24 <@Winsalot> noooooooooooo! 06:25 hi Winsalot, going mystic now? are you getting good implants? ;) 06:25 <@Winsalot> Btw, just now I got generated as invisible ak thief is dead! which means I cant even farlook into myself 06:25 hahaha 06:25 <@Winsalot> implants are cool 06:25 yeah, because it's basically there's a 1/10000 or similarly small chance that you actually get bitten by the nastiest EPIs, but that's not really an excuse. eventually you lose a promising character for what amounts to no reason at all 06:25 <@Winsalot> got one which made me burned 😃 06:26 either lose the character, or have the game rendered unplayable by something or other 06:26 <@Winsalot> mystic seems fun, but so far I cant seem to get a decent roll on my normal and monster forms 06:27 ais523, you can cheat the fifth level designs by jumping down a hole, it seems :) 06:28 I wonder if someone can actually design that 06:28 Evil sokoban level that requires jumping down a hole to return to the upstairs to solve. 06:28 downstairs* 06:29 So if you throw a solver at it, it gives you "nop, unsolvable, idk" 06:29 some of slex's sokoban levels either become vastly easier if you jump down a hole at certain points, or actively require it 06:29 It occurs to me it works on the vanilla level 2 one with the big loop too. 06:31 Lymia: I believe AmyBSOD\srn's solution to Evil 2b requires a hole jump 06:31 AmyBSOD\srn: fog of war sokoban has the problem that it's easy, and quite common, for a level to be solvable but require guessing 06:31 you'd have to explicitly design the level so that no guesses were required 06:31 on another note 06:32 <@Winsalot> AmyBSOD\srn: you said implants are supposted to help races that have no hands or are deformed in other ways. But I'm wearing my implant and cant even open doors 06:32 this is slex though :D 06:32 Winsalot: well that's not one of the things implants help you to do... 06:32 AmyBSOD\srn: slex YANI: trap that duplicates the effect of the first trap you triggered; the second time you trigger one of those traps it duplicates the effect of the second trap you triggered, and so on 06:32 use kicking or #force to remove the doors 06:32 probably requires changing save data, sadly 06:32 ais523, lists in C 06:32 pls no 06:33 linked lists aren't too bad 06:33 <@Winsalot> Arent implants like a second pair of hands or something? How do you imagine them? 06:33 What's the equivalent of an EPI, except it's evil to the programmer, not the player? 06:33 Lymia: stash search 06:34 (a currently-implemented FIQHack feature that lots of people had wanted for ages but nobody could figure out how to implement) 06:34 the implant is supposed to be a device that goes into your head and thereby magically grants you properties, e.g. fire resistance and stuff 06:34 (until FIQ finally found a method that works) 06:34 (depending on what the implant does) 06:34 I should take a look at the FIQhack codebase some day :D 06:34 (Is it still a Nethack variant if the crazy Lymia goes and translates it to another language?) 06:35 <@Winsalot> Oh 06:35 a direct translation, yes 06:35 but that'd take years 06:35 <@Winsalot> well, magical arms that get implanted into your spine and help you open doors sound way cooler 06:35 if you're reimplementing it from spoilers, what normally happens is that you get 1-2% of the way in, realise "I have a really solid and flexible game engine so I might as well make my own game on it", and you end up writing your own roguelike that isn't NetHack but can be quite good in other ways 06:35 I think several roguelikes started like that 06:36 Heh :D 06:36 that said, I'm not sure a full rewrite is quite as ridiculous as people make it sound 06:36 more than half the lines in NH4 are either different from their counterpart in vanilla, or have no counterpart in vanilla 06:36 and that's despite it not even really playing any differently 06:37 It is to people who can't imagine writing Nethack to begin with :P 06:37 you could maybe translate NetHack into another language piecewise 06:38 maybe start by abstracting away the internal gamestate to use a consistent API rather than direct field accesses 06:38 then move it into a database or something 06:38 Not if you want to majorly rework the whole engine :D 06:38 most of the game is still written in C but you can now change parts of it into another language and have them work 06:38 I'm not sold on going all the way to a DB, though 06:38 the actual engine is only a tiny part of NetHack 06:38 Unless you mean a nosql DB 06:38 why a db? making the nethack core engine into a lib has been done before 06:38 bhaak: Message from @luxidream at 2018-06-10 02:01 EDT: my xnethack ascension went in twice 06:38 bhaak: Message from @luxidream at 2018-06-10 02:02 EDT: contributes nothing extra to the clan but is a bug nevertheless. it shows up correctly on scoreboard 06:38 no, I mean relational 06:38 Why? 06:38 like half the things that are difficult to handle in NetHack are due to the fact that the game is effectively trying to store relational data in an ad-hoc way 06:39 then you can just call into the engine and start replacing parts. 06:39 counter argument: the map doesn't fit the relational paradigma well 06:39 so you have stale caches, memory allocation trouble, and the like all over the place 06:40 Examples of these things? 06:40 bhaak: it fits it excellently, "item is on map square" is one of the things that's hardest to do in NetHack's codebase (although it has to, obviously) and very easy in a dB 06:40 I've never found a big reason to use relational DBs in modern languages, personally, unless I actually need ACID guarantees. 06:41 Rust can do, for example: current_level.items.iter().any(|item| item.pos = target_square) 06:41 let's see how it works in vanilla: you have the levl array that contains all the map squares, but it doesn't contain the items; it does however contain the most recently remembered item that the player saw there 06:41 because, it turns out, FP paradigm operators already do exactly that, with native structs. 06:41 but only its tile number, not the item object 06:42 then you have a second 2D array of the starts of item chains 06:42 those are intrusive lists that chain via nexthere 06:42 There's plenty of Scala/etc libraries, and a whole language feature in C# (LINQ) that map between the relational DB paradigm and FP list operators. 06:42 you also have a separate chain of all the items on the level (but on the floor), via nobj 06:42 <‎Lymia‎>‎ Rust can do, for example: current_level.items.iter().any(|item| item.pos = target_square) ← this is O(n) 06:42 If you're only using a relational DB for relational operators, and not actually ACID guarantees... uh... you have FP operators. 06:43 /that/'s the reason you use a relational DB 06:43 to get caches that you know will always act correctly without having to write and updae them yourself 06:43 ais523, is the overhead larger than the overhead for calling into sqlite? 06:43 how would you model an item and the map in a relational db 06:43 n is generally <1000, not an issue unless performed in a loop. 06:43 clearly there's a number of entries where the answer is yes; I don't know how large it is offhand 06:44 bhaak: at a quick guess: each item has a primary key which is its identity; each possible location in the game (map square, player inventory, monster identity, etc.) also has a key; each item knows where it is 06:45 Lymia: the issue with O(n) algorithms is that you can't normally nest them, as now you're quadratic 06:45 *monster inventory 06:45 IMO 06:45 probably "monster inventory" as a location would use the same keys as the monsters themselves 06:46 but the point is that all this means that you need to make very few assumptions about what can be where 06:46 so you would have to find out the id of any x/y location when you have to update it 06:46 A relational DB is the perfect way to take a simple implementation that works well and turn it into a nightmare to debug and understand. 06:46 Unless you have an ORM or something. 06:46 isn't that worse than how it is now. where each obj knows where it is 06:46 bhaak: yes, and? you'd be using those for /everything/ 06:47 what we have now is "each obj knows where it is, each map square knows which objects are there" 06:47 meaning two pieces of information you have to update every time an object moves 06:47 there's no normalisation, so it's very easy to introduce bugs 06:48 "an object has a timer and falls through hole" is a very large source of bugs 06:48 *falls through a hole 06:48 come to think of it, the current system is identical 06:48 Isn't the conventional answer to such issues ECS? 06:49 When an item moves, the system that keeps track of timeouts still has a handle on it through the monster. 06:49 in order to update an item's x/y location you need to find the pointer to its entry in level.locations 06:49 Lymia: ECS is solving a different problem I think 06:49 the only reason this is a problem in NetHack in particular is that the game makes a lot of assumptions that things are on the current level 06:50 (which bhaak has implicitly acknowledged by saying "x/y" out of habit and omitting the z coordinate) 06:51 What are the examples of these bugs? 06:51 That involve a timer like that? 06:51 !tell AntiGulp mystic scarabs are too easy, I can kill them with no trouble with a +0 Excalibur 06:51 Will do, AmyBSOD\srn! 06:52 there's also the additional 8 nobj states that declare where the item is currently 06:52 I'll point out that a DB doesn't automatically prevent making current floor assumptions 06:52 Unless your DB approach involves *each floor always simulating* 06:52 this is essentially a compound primary key 06:53 bhaak: yes, that's the biggest problem here; I think what you do is have globally unique primary keys (as in, different /tables/ don't share keys) 06:54 then to find out which nobj state you're in you use an indexing table that lets you know what the key refers to 06:54 this might actually use less memory than the current method, I'm not sure 06:54 but it seems correct because you need to know what state the item is in to be able to do anything with its location anyway 06:56 a globally unique primary key, like a pointer? :) 06:56 but I see where you're going 06:57 -!- Tariru has joined #hardfought 06:57 bhaak: except that the save code is 100% for free, as opposed to being ridiculously convoluted 06:57 NetHack has huge problems saving poitnerse 06:57 hi Tariru! glad to see you got out of that hellhole! 06:57 Hi Amy 06:57 Yeah, the game is looking quite ascendable now. 06:57 :) 06:58 ais523: in a more dynamic language, instead of linked lists they might have used dynamic arrays 06:58 No amulets of life saving though, and few wishes. 06:58 there's no reason to abandon things that work, it's just that doing them via a database rather than via ad hoc structures gets you a lot of otherwise difficult/painful-to-write code without costing much 06:58 I guess the way I think about it is, the problem with a list or array is that it only works in one direction, it forces you to think about a problem a certain way 06:58 or possibly a tree structure for easier access. 06:59 and if you have two things that are related both ways (an item is at a location and a location holds an item), you either need an index both ways, or else a complex data structure supporting the operations you need 06:59 with a relational database you get pretty much every operation you could need in an efficient way, so it doesn't place constraints on how you use the data 06:59 but yeah, in a more dynamic language the access problem blurs quite a bit. it doesn't matter code-wise with a proper abstraction if you have an ad hoc data structure or db tables 07:00 AmyBSOD\srn: Are there any other ways of getting rid of traps other than the spell/scroll/wands of trap disarming? 07:01 Tariru: some traps can be removed by being on them and using #monster in spider form, but that's limited to few trap types 07:01 also, you can dig down on most of them, but that also requires being on the trap square 07:01 any specific trap you want to remove? 07:01 Yeah, that's the problem. It's nasty traps spawning on upstairs that I'm concerned about. 07:02 I had to write a scroll to get rid of an MC trap that did that, but I don't have the resources to remove any more unless I use a wish. 07:02 theoretically you could cast air current or be repeatedly hit by a monster with "pushes you away" attacks and hope you end up on the stair, which doesn't trigger the trap 07:03 That would work, I do have the spell, not sure if it's still remembered though 07:04 Yep, 30% remaining on it still. That makes things a little easier. Might be able to spare a wish for life saving now then. 07:04 yay! 07:04 cast it a couple of times to ensure it won't be forgotten any time soon :) 07:05 I'm playing splicehack right now, and I'm a bit concerned about the attacks of thought/memory hounds... 07:05 I think 3000 turns ought to get me to the planes anyway. 07:05 any splicehack players around who could tell me how dangerous they are? 07:05 ais523, I'd argue that an even better solution is a library that generates automatically coherent indexes on various data structures. (In languages that support it, at least) 07:05 I've not played splicehack myself, so can't help there. 07:05 Lymia: that's basically what a database is 07:05 people seem to keep assuming that they're some ridiculously complex/heavy structure 07:06 It has no reason to provide ACID guarantees on anything 07:06 hmm, according to the source code, the memory hound's bite will erase item knowledge 07:06 I guess you'd need to use prepared statements so that they're precompiled into code rather than needing to be parsed from SQL 07:06 you can turn ACID off in many databases (for performance) 07:06 although in NH4 I took some trouble implementing many of the ACID properties myself 07:06 as they're valuable to players 07:07 It operates on a static level, without an interface layer that's *literally strings that are dynamically compiled* 07:07 I think the way most databases work is that the strings are compiled once into code, and then you execute the code, isn't it? 07:07 that's not much different from writing a C program 07:08 we had philosophical differences 07:09 A Nethack engine has no reason to contain a compiler, though. 07:10 NetHack contains two compilers 07:10 already 07:10 three if you count makedefs, although you probably shouldn't 07:10 Not at runtime 07:10 Does it? 07:11 if you're using one of the big professional databases, they don't compile the prepared statements at runtime either 07:11 the compiled prepared statements are stored in the database along with the data 07:12 SQLite appears to require statements to be prepared at runtime because it can't store them, but it's fairly primitive as databases go (whilst still being a lot more advanced than just aobut everything else) 07:12 but even then, it's not so much runtime as load time 07:12 because you only do it once when the program starts up 07:14 They are fundamentally not static, in the sense that compile polymorphism is. 07:14 I'm still having problems seeing any distinction between compiled SQL and compiled C 07:15 The big professional databases still do not bake queries into code that manipulates predefined structures in memory with code accessible to the same optimizer. 07:15 oh, I see, the problem you're drawing is not the fact that both languages exist, but the fact that they're two different languages 07:15 A binary built with such a technique contains no compiler, can DCE away unnecessary features, etc. 07:15 and thus you're paying FFI overhead 07:15 but the SQL compiler wouldn't be in the binary in any case 07:16 it'd be in a library, probably a shared one, and quite possibly one that ships with the system 07:16 (you'd probably be upset to learn how dynamic loaders work…) 07:16 yesssss someone in the castle barracks has a wand of death! I need that! 07:17 AmyBSOD\srn: you're getting excited over vanilla items? :-D 07:17 I'd have thought there'd be loads of better items that were newly added 07:18 ais523: well, samurai is the only role that I know how to play in non-slex, and the quest nemesis requires a wand of death unless you want to take chances 07:18 (playing splicehack right now) 07:18 hmm, that's a weird choice of primary role, but fair enough 07:18 I only know how to play Mah (somewhat) because I know how to play Wiz :D 07:18 samurai is literally the best role 07:19 <@Winsalot> samurai is weeb role 07:19 <@Winsalot> just saying 07:20 it's basically valk but harder 07:20 *easier 07:20 but you get a katana 07:20 valk needs to reach XL7 to get speed, samurai has it right from the start 07:20 samurai also has a good bow for those annoying floating eyes and blue jellies 07:20 speed matters before XL7? also, /oSM is common 07:20 daggers are a better ranged weapon than bow+arrow 07:20 Non-spellcasters are weird 07:21 Am I supposed to clear rooms 07:21 by 07:21 walking enemies one-by-one? 07:21 and valks can just melee blue jellies :-P 07:21 walking into* 07:21 Lymia: you kind-of just walk through the enemies 07:21 leaving a trail of corpses in your wake 07:21 -!- raisse\splat has joined #hardfought 07:21 -!- mode/#hardfought [+v raisse\splat] by ChanServ 07:21 at least, that's how a powerful melee role wants to end up :-D 07:22 and they can often do it even early-game, whereas casters have to wait until late game for their room-clearing attacks 07:22 wow, splicehack dragons are apparently capable of reading scrolls 07:22 ais523, really, I have two major issues with a database in Nethack, I suppose. 07:23 It's not particularly heavyweight, but it isn't lightweight either, compared to... using normal structures. 07:23 I guess I see normal structures as heavyweight because you need so many of them 07:24 Which... may be a tipping point, considering Nethack runs literally everywhere, and is used for public servers that run who knows how many instances 07:24 I don't want to be known as the variant that uses 100MB 07:24 well, databases swap out to disk much better than structs do 07:25 if you have ACID they can literally just drop unused data from memory as they know they can reload it from durable storage 07:25 so the concern is more about disk than memory 07:25 And more fundamentally, databases introduce a split between the DB representation and the client-code representation, and the DB side does not have access to any of the abstraction features that would otherwise be accessible to the outer language 07:26 This is no major loss if you're working in C, it is very annoying when you're working in Rust, and even the best ORM available still has code that is 5x longer than the FP operator version. 07:26 I guess that's valid, but the database is a powerful abstraction tool in its own right, so I guess the question is as to which set of abstractions is worth more 07:27 FWIW, I dislike ORMs, not really because of the concept but because most actual implementations seem to have no idea what they're doing 07:27 I dropped the ORM and stuck to raw queries, personally 07:27 same 07:27 One of my current projects is a Discord bot with a Sqlite DB, and the language feature split was what made modules that interface with the DB very annoying. 07:28 I think there's probably a market for "the DB compiles into your program" 07:28 And... as for dumping to disk in Rust... `#[derive(Serialize, Deserialize)]` 07:28 serde is a great library. 07:28 actually I'd be surprised if there's no product that currently does that 07:28 Lymia: be really careful wrt deep and shallow copies and linked references and the like 07:28 A lot of this is a way larger issue in less modern languages. 07:28 it is a very complicated task 07:28 (this is a major part of my day job) 07:29 in general there are way too many different ways to serialise that preserve different aspects of the memory state you're serialising 07:29 and any stock serialiser will have to choose one, which may not be what you wanted 07:30 It's probably easier than in, say, Java, because Rust's type system encodes a lot of information about programmer intent wrt that. 07:30 the model used by most stock serialisers is to create a deep copy with identical internal structure to the thing you're serialising, which is good if you're planning to kill and restart the process, but bad for cooperating with anything that's currently in memory 07:30 -!- Borto has joined #hardfought 07:31 I'm probably the only player in existence who picks disenchanters as the first genocide target. 07:32 [hdf-us] [nh343] rebatela (Pri Hum Mal Neu), 1797 points, T:2062, killed by a rothe 07:32 i really hate disenchanters 07:32 but even more, I hate alarms that don't go off 07:33 still waiting to bless my scrolls so I can wipe out L and c though 07:34 ais523, it seems that Serde doesn't currently handle this, but Rust likely conveys enough programmer intent to actually do a reasonable default. 07:35 Lymia: Rust's types explain the programmer intent for how the data relate to each other, also about what /can/ be copied 07:35 Rc (reference counted pointer) is the primary type that actually would allow shared state outside out of linear ownership graphs with no self-pointers. 07:35 however, a serialise+deserialise is essentially a copy operation 07:35 and it's reasonable to want to copy different amounts at different times 07:36 Rc can be assumed to be intended as a shared pointer (for the most part), and everything else a deep copy 07:36 IMO 07:37 All Nethack really needs is a straight dump/load of a single main game state struct to/from disk 07:37 For which use case, this would be sufficient 07:37 well, if something can't be shallow copied (e.g. Box), then obviously you're meant to copy it deeply or not at all 07:38 but yes, Rust serde would probably work for NetHack saving as it kills the process 07:38 so what you want is a deep copy of everything, which is one of the easiest cases 07:38 A deep copy preserving shared references, at least. 07:38 it would not work nearly as well in a multiplayer version, though 07:38 It seems proper serialization of Rc is only in forks of serde so far. 07:38 that's implicit in the idea of a deep copy 07:38 otherwise it isn't a copy at all 07:39 the interesting part of serialisation is what parts of the copy share with what parts of the original 07:40 wrt multiplayer, what kind of data would you share? 07:40 the way it's done in AceHack, you share most of the gamestate, except for player-specific things 07:40 most notably, struct you is not shared 07:40 that's between the players 07:41 in terms of the copies, you want to be able to serialise and deserialise and keep the links to the existing struct you (which isn't copied) intact 07:41 the better solution is probably the FIQHack-like approach of having each player's entire stats visible in the other players' games 07:41 (this also makes PvP possible) 07:42 If I was designing it from scratch, I'd probably try a synchronous approach, I suppose. 07:43 Where both sides know the entire game state, and only communicate actions, relying on a shared RNG. 07:43 At least.... if (AFAIK like Acehack) expect the communication to be local and not over the network 07:43 huh, that's a weird "at least", the most common reason to do that is to make the performance reasonable when a network is involved 07:44 Doesn't the fast way involve rollbacks, and such? 07:44 And, generally, complicated resolution methods. 07:44 each player is still copying their actions to everyone else's games; the rollbacks/resolutions are purely to handle what happens when you think the other person's taken an action but haven't seen what it is yet 07:45 NetHack is not very timing-sensitive so you could probably just wait for it without any real issue 07:45 Nethack is pretty lousy to play with lag, though 07:45 :( 07:46 To maintain game state, it would have to wait for the other player, even when they aren't on the same floor. 07:46 (In the sense of a Mutex) 07:46 multiplayer AceHack allows movement on different floors to work in parallel 07:46 and simply isolates them from each other completely 07:47 (e.g. if you drop an item down a hole, it'll only arrive on that level when you do) 07:47 oh, huh 07:47