I just think "auto" could be spelled correctly, other than that it seems fine

Unless, of course, the NPC's name is "Otto", then we could call it that.

Yay for a typo and copy and paste kicking my ass... commiting the typo fix now...

lol well that sure brightened my otherwise crappy Hump Day

Thank you.

lol glad I could help

Thank Jab it's perfect

If spell cast is 50 time a sec, it useless, to mutch check and ressource, and not allow a really customisation.
And maybe it can be very usefull to add some value in npc data, like attack speed.

First actual spawn_npc_spells doesnt contain spell percent field better to be in this table instead of spawn_npc.cast_percentage. A thing like that:
I Think we should handle AI with LUA script but we need also some more function with ARRAY (Jab you hate me right ?) list retrive spell list, percent chance...
This allow to set two things, 1 the probability of NPC casting a spell (handle in LUA) 2 The probability of spell choose when casting (handle with DB data in LUA). And 3 :p this allow to script all spell probability ONLY in LUA for complex Raid Boss script.

Assuming NPC Spell list can be call in LUA and Combat AI call LUA script.

In this lua file at Auto_Attack_Tick you do your process for AI (like I have start to do) simple example Generic Warrior AI.

CASE DEFFENSIVE
-- 35 % to caste combat art [ Kick (50 %) or Taunt (50 %) ] (In my example there spell list and percent value are in npc spells table but we can set in script a generic list if there is no spell list in db)
-- 2 % to switch aggro to heal on attackers group, if there is no heal switch on a random member of attackers group
-- 12 % to check if attackers group members 's target is heal of Spawn group, if there is no heal switch on a random member of attackers group who not focus Spawn if all member focus Spawn do nothing
-- 1 % to full wipe aggro (not mean end combat)
-- 50 % to simply auto attack

CASE OFFENSIVE
-- 35 % to caste combat art [ Kick (90 %) or Taunt (10 %) ] (In my example there spell list and percent value are in npc spells table but we can set in script a generic list if there is no spell list in db)
-- 5 % to switch aggro to heal on attackers group, if there is no heal switch on a random member of attackers group
-- 2 % to check if attackers group members 's target is heal of Spawn group, if there is no heal switch on a random member of attackers group who not focus Spawn if all member focus Spawn do nothing
-- 4 % to full wipe aggro (not mean end combat)
-- 50 % to simply auto attack

CASE BALANCED
-- 35 % to caste combat art [ Kick (75 %) or Taunt (25 %) ] (In my example there spell list and percent value are in npc spells table but we can set in script a generic list if there is no spell list in db)
-- 2 % to switch aggro to heal on attackers group, if there is no heal switch on a random member of attackers group
-- 5 % to check if attackers group members 's target is heal of Spawn group, if there is no heal switch on a random member of attackers group who not focus Spawn if all member focus Spawn do nothing
-- 1 % to full wipe aggro (not mean end combat)
-- 57 % to simply auto attack

It for generic, but maybe we need some more like Generic Gnoll Wario AI, this allow to set so combat text specific to gnolls and maybe some other stuff.

And at spawn event do "buff event", here we need to handle NPC buffing, simply store avariables buffs in NPC Buff List table look like NPC spell list (spell and percent chance for each mode Offensive, def..).

-- Enum buff list
-- if < 100 % do rand and cast if needed
-- if >= 100 % cast it

And we need one more stuff for recall a "buff check" this can be done in lua (rand spell probability each time) or in code (rand on time in lua, store NPC buff list and refresh it when needed). Assuming code can call NPC animation, on live, you can see npc cast buff when they are down.

Again that is just my idea, just purpose it But think there things can be help us to get a living world quickly

alfa wrote:First actual spawn_npc_spells doesnt contain spell percent field better to be in this table instead of spawn_npc.cast_percentage. A thing like that:

Code: Select all

ALTER TABLE `spawn_npc_spells`
ADD COLUMN `spell_chance_balanced`  int NOT NULL AFTER `spell_tier`,
ADD COLUMN `spell_chance_offensive`  int NOT NULL AFTER `spell_chance_balanced`,
ADD COLUMN `spell_chance_defensive`  int NOT NULL AFTER `spell_chance_offensive`;

I do not believe we need to change this. I see no reason each individual spell in a spell's list needs to have even a % chance to cast for AI. The NPC record itself is EITHER balanced, offensive, OR defensive, not possibly all 3 at once.


As for your other examples (cases), I agree with the configurable percentages of various things happening, but I feel those can be governed by rules (50% chance the cast will be offensive, 25% chance the cast will be defensive, 25% chance it will be both/either/neither).


I am very opposed to doing all AI functions in text-based LUA scripts, period. I just do not like LUA, a run-time, plain text file on some disk somewhere, being the main core of our engine. I am a DB guy, and feel all data should be in a database. I refuse to believe reading flat script files is faster than loading data from a database, or more efficient with memory or performance.

That said, you bring up very interesting points - and I want you to continue to do so. I may not like, or approve, of the ideas but they definitely lead to discussions which can only improve how we do things. When Jabantiz and/or Scatman chime in on this, they have the ability to sway my opinion... but for now, I'd prefer to simply make the C++ code "smarter" by adding options to the NPC's spawn records and/or LUA functions as you suggest.

AI in LUA is not my favorite topic. But keep the suggestions coming.

John Adams wrote:As for your other examples (cases), I agree with the configurable percentages of various things happening, but I feel those can be governed by rules (50% chance the cast will be offensive, 25% chance the cast will be defensive, 25% chance it will be both/either/neither).

Let me correct what seems my inconsistency

I mean to build Rules that say, if an NPC is BALANCED AI, then these are the percentages the spawn should use. But if they are OFFENSIVE only, then only case the spells in their OFFENSIVE list based on cast_percentage, and similar with DEFENSIVE.

For instance, if a NPC group has a Templar, that guy is not likely to be tuned for Balanced or Offensive, but will be Defensive. That does not mean he stands there doing nothing while no one needs healing. The Templar NPC's AI will cast offensive and melee combat UNTIL the smart AI sees someone needs Defensive help. That is my example of how it currently works.

We can always add more than those 3, as well. AGGRESSIVE, TIMID, COWARDLY that could get interesting heh.

John Adams wrote: I do not believe we need to change this. I see no reason each individual spell in a spell's list needs to have even a % chance to cast for AI. The NPC record itself is EITHER balanced, offensive, OR defensive, not possibly all 3 at once.

A NPC can be switch mode (ex def to off) when his life come under a percentage (it is that on live) also try to fight a group of mob on live, kill them multiple time in a different order, you will see comportement of mob change, so switching mob can be easely handle with some check.

John Adams wrote: As for your other examples (cases), I agree with the configurable percentages of various things happening, but I feel those can be governed by rules (50% chance the cast will be offensive, 25% chance the cast will be defensive, 25% chance it will be both/either/neither).

You cannot set rule for every Spawn, a berserk NPC in Offensive should be have chance to cast spell like 25 % cause in offensive mod it should buff for attack speed and spell should be attack spell in priority. In defensive mob, npc should cast spell like 35 %, cause in defense attack speed nerfed, and should be defensive spell like temp mitigation buff or other like that in priority.

So I really think add percent field for each mod in spell list can be really a good thing.

John Adams wrote: I refuse to believe reading flat script files is faster than loading data from a database, or more efficient with memory or performance.

In this two case there is a disk / memory access, ok maybe lua use more ressource BUT more realistic engine, and allow more customisation. DB request You do request, you read return, LUA, you open file and read it, so basicly the same things no ? Or why the we don't use DB for quest ?? :p (And for SQL you use memory / cpu for doing request)

But I agree it would be consume a little more ressource, but you can grab it without check 50 times a sec for do a rand and 20% of time (most I have see in db), so 10 time a sec read npc spell list and process.

But let dev say opinions about it

John Adams wrote:For instance, if a NPC group has a Templar, that guy is not likely to be tuned for Balanced or Offensive, but will be Defensive. That does not mean he stands there doing nothing while no one needs healing. The Templar NPC's AI will cast offensive and melee combat UNTIL the smart AI sees someone needs Defensive help. That is my example of how it currently works.

That is the same things I purpose, but in lua, more possibility than rules and 3 or 4 options I trust. In lua you can handle very large amount of situations (I not saying you cannot do it harcoded).

Ex Group of 3 NPCs 1 Guardian, 1 Berserk, 1 Templar

Start of fight Guardian Def, Berserk Off, Templar Blanced.
If Guard die first, Berserk switch to Def, Templar too
If Berserk die first, 2 other don't switch cause a Guardian do not need a Def Templar
If Templar die first, Guardian don't switch, Berserk switch to Def for Tank assist
If Templare is last alive, switch to Def
If Guardian is last alive keep in Def
if Berserk is last alive, switch in Off (fury)

My idea is to build generic script for each class, and after just use data in spell / buff lists for running it.
And an admin who want to change generic AI have just to edit file.
Also this allow to use a good lua API that needed for combat scripting.

alfa wrote:In this two case there is a disk / memory access, ok maybe lua use more ressource BUT more realistic engine, and allow more customisation. DB request You do request, you read return, LUA, you open file and read it, so basicly the same things no ? Or why the we don't use DB for quest ?? :p

Because LethalEncounter implemented LUA for all scripting, and refused my suggestion to leave Quest data in the database. So when he quit the project in 2009, the first thing Scatman and I did was put quest data in the database. So, I win

Scripting "Hello" and "Go do this" (SpawnScripts) is a helluva lot different than the system to load a 5,000 line combat AI script from a flat text file on the same physical hard disk as your EQ2World.exe is running on -- multiplied by 1,000,000 spawns in the database (conservatively).

Lastly, my MySQL instance is not on the same machine as my EQ2World.exe. Thus, hitting the MySQL server CPU and Disk is irrelevant.


I'm certainly open to discussing the other points, because while I do see your advantage to finely tune each and every encounter to the millisecond of accuracy, I am not sure I care that much. No one (but Core Dev) will ever give a crap that much about EQ2Emulator to begin with. Asking us to stop what we're doing and go this granular seems unreasonable at this time. Maybe 3 years ago. But, we'll see.

I want to hear Jabantiz tell it it will take 5 mins to implement, and will not affect one other thing in Emu, then I may become convinced.

John I not mean reading all script all time, generic ai lua script load at start, and if and only if spawn have be set to not use generic ai read specific script.
But again just an idea, not sure you understand all things I mean, but I do the best I can, it would be eaysier if you speak French :p

alfa wrote:And an admin who want to change generic AI have just to edit file

And that, right there, is my biggest problem with it. You have to be local, on the box, to make a change to the AI. Even slightest. Then reload. What if a future admin has a remote GM/Dev? They have to grant access to their machine to make a simple LUA script change? When changing a field value in a database is simple. Done. Reload. Carry on.

Then, I want to go find where my bug is. I have 10 fields in a database, oh, there it is. fixed. Done. moving on.

It's in a LUA? Where? /eq2/world/scripts/spawnscripts/somefolder/anotherfolder/this.lua which has 5,000 lines in it. Which line? #4,234? ok, scroll down, type something, save it. Maybe I fat finger and added an extra \, entire LUA engine shits the bed because it encountered a bad character and entire encounter is now bugged.

There are 10,000 reasons why I dislike text file scripting. If it were up to me, LUA would not even be a part of this emulator. Everything would be compiled code. However, I was convinced years ago that LUA was "simpler", and coding everything (C#, Java, Perl, or C++/DLL) would make the Emulator unusable to non-dev users.


Hint: You will never convince me Text File is better than DB. Don't even try. Instead, convince me that better quality encounters can be created using more advanced LUA functions.

John Adams wrote:Hint: You will never convince me Text File is better than DB. Don't even try. Instead, convince me that better quality encounters can be created using more advanced LUA functions.

Ok, Encounter will have a better quality if we using more advance LUA functions, and scripts

I talk about LUA cause that is what we have, not need to change all system, or maybe you think about thing like some Wow emu, that need to compile a dll file for each AI / Combat scripting

I care about 2 things, in this order:

• Stability
  Performance

That's it. Our little emulator engine runs fantastic when there is ZERO content loading into any zones.

Adding content, and the server performance is dropping like a brick from an airplane. And we've barely gotten started, adding content. When this database is full of 1,000,000+ spawns and all 800 zones+instances, etc... our Core server needs to be as efficient as possible BEFORE we get there.

So adding anything that I know for a fact is going to slow the processing down even more (like processing giant LUA scripts every 20ms) scares me. I really wish people would start listening to me and understanding my concerns.

Auto attack is not each 20 ms John, and it only concern mob that are fighting not all spawn. I know you look at performances, I did too, and be sure and Generic AI script can be smaller than 5000 lines, for sure, and I agree it need to be really accurate to not use to mutch if, rand, other stuff like that, script need to be smarter it can be, but from a generic view I don't see difference between load a script at server startup and use it in memory and code same things directly in code, routine / timer would be the same in this two case no ?. And again it allow customisation, and not need to handle generic combat AI a way, and specific combat AI another way.
But like you wish to heard point of view of dev about perf / stability risk.