Mister_Bison

Oh, that. But people can also do their own pimp_malloc() and pimp_free() in C, it's the same again. It's not because it was made popular by a language that it doesn't exist anywhere. There is *no* language that doesn't allow the user to fool up. It may be easier or harder, but you would be surprised.

Heh, true. But C++ still lacks the amount of standard (but poorly implemented as you said) things that Java has, allowing fast prototyping.

You could cut up on the fancy names/acronyms already >_>

Except if Objective-C has found a way to code an Oracle, it's just the same algorithm as in boost::shared_ptr, as these are also reference counting, and it's still vulnerable to circle-referencing. In boost you have weak pointers for that at least. But I never used that yet, never needed to. I prefer to see the ownerships of things and clear creation such as deep copies and shallow copies.

On other news, I think I've nailed the attribmod library, I'm tackling the Powers & Effects (restricted to attribmod effects, discarding the summon, setmode etc...) one, but that one is hairy, especially the "Ignore buffs&debuffs" flag on powers >_> I need that to "handle how Combat Training: Offensive buffs itself" (which it does at the moment) and resistance toggles. and a whole other bunch of things, with real entities. I could do it quick & dirty but it wouldn't be nice.

Dang, you're right. I feel stupid ^^

It's *exactly* the same problem in C, and yet C is used for large projects. new() and delete() are just encapsulation of malloc and free with init clean functions. Plus you have STL containers that do help you to properly implement standard features such as Maps, Lists and Dynamic arrays.

What's ARC by the way ?

same as me, and I did throw myself in this little show of wit because I needed to get that off my head anyway, even if I fail miserably it was driving me insane, and it could even provide an example of what to do or not do.

And generally that's ten times shorter to reimplement in another language than code the first time. And you still believe the 3000 thousands of hours would be necessary to redo the combat loop only ? I would understand if it was accounting for the total amount of effort required with just the game and the keenest eye and none of your posts or the Wiki, but the game logic has been hashed though a great deal already. We were just discussing the specifics of which loop did what, we were not discussing how Resistance was implemented !

Oh yeah, I'll can what I do... in a zip file ! (nice typo ^^ produce a few more and you'll end up the subject of a Zero Wing video. As a tribute, of course.)

Yeah, of course, if you start from scratch. Nobody should ever do that anyway, comparing to something is a better stability than dangling in the void, at least if you have enough critical judgement to stay away from bad code.

you would be surprised what a trained and flexible mind can do with C++. It's becoming more and more a no-paradigm programming language. Also, if you ever happen to see both beyond the code what's done in the background, and in the big picture what gets done from back to front, it's not a matter of language anymore. Church's thesis essentially demonstrated all computer languages are equal. Now I wouldn't program City on a Turing Machine, I'm not such a masochist

*sob* and again another scripting language.

*cough* cough* You may like C++'s templates.

I'm asking because coding the week-end is already off-work working so...

Thanks for the cheering, because I'm not trying to accomplish something, I'm doing something. I'm assessing how much can be done in a week-end day by a moderately knowledgeable programmer/engineer.

Salt or sugar popcorn ?

EDIT: Okay, first throw is done, I think. Now I need data to test it. Approx 4 hours (had a social meal). Then again, I was just writing what I had already pictured in my head.
I think I'll need to add some things though, but now the time is on pause until I get some creating and coding done to feed data to the calculator. Then I'll need some more time to make a UI to test it. I just need bulk import from some City of Data tables.

EDIT2: heh, forgot to take level into account, brb.

I'm sorry Codewalker, I should have answered specifically earlier. I have heard you so it changed my meaning of "take data is in City of Data" to "begin where City of Data left off, which is still a good start", and any programmer would be interested, if he has to, look at the work done to take apart what's still good in it and correct the wrongs that could still be sitting in the code as "//OMG this is ulgy fix it later". But I continued to use "take city of data" because it was more near than far of what is needed, I think. Of course that takes a computer engineer, not a programmer, to do that.

Did I mention I graduated as engineer, not from programmer background ?

So, off to implement the attribmod calculator !

that, but the server isn't real-time... at least not hard realtime... because the event loop can overflow.

You mean why you need a third person to be arbiters on the snow events ?

If I don't know how you name these things, I'm still as puzzled. I knew all these all along, these are on City Of Data, as well as attributes modifiers and scale. That doesn't make me budge.

Wait, you said you're asking me to also create powers and evaluate powers with this loop, which will call the attribmod calculator, not the other way around.

I said combination because that's one of the variability an engine has to deal with, the other being the variability of the effects, and your for(;{evaluate(power);} hides it even more. The complexity is in the evaluate, and I'm aware of that.

Just pointing I'm feeling the same. But I'm also kicking into touch here and going around the goal pillars because you don't necessarily need to be real time, after certain points, you *can't* be real time. So make the game "timeout" while you get things straight and send updates again afterward, the loop is on a sleep if it ends early, but continues on the contrary.

And you already spent all this time, it's now very easy to implement the details.

True, but as Codewalker said, it's easier to make the engine that takes each instruction laid out and carries them at normal pace than make a simulator.

Because you have an in-game tool to do exactly that, it's the power analyzer (MK3 to hit any mob). Once it has hit you have the hit entity's real numbers (attributes), and so this data would have come to nullify the need for this power.

Recoded and implemented a Learning (Artificial Intelligence) framework. In a team I'm currently working on... something... I guess I could say multi-sensors calibration and in particular, Kinect handling in an existing framework, the thing is enormous but I'm still green in this project. I didn't do real-time game development, I've had real-time courses though, and the server is NOT real time (else it would never "lag" (I'm talking about cimerora/hami lag))EDIT: forgot to tackel the networking code. I did do the master part of an I2C communication system as part of amateur robotics system back in school. I also pretty well remember the practical sessions with linux virtual machines.

Sure I'm not a multi-decade code veteran here, I didn't code in Basic or Fortran, nor on Amiga or Amstrad, but that doesn't mean I don't have a brain and can't learn or work, or that what I was taught was crap. Doesn't mean I know it all either. I actually want to appear stupid and get told things harshly than people saying nothing and letting me in the wrong. Actually, I've come to see this little explanation as tutoring on the game's mechanisms, making notes here and there that could see better use than in the dark corners of a single brain. Also, I'd like some people to come to term with their assumptions that there is only the harsh way to an implementation.

Wait, the client shouldn't do anything, it's a client, it sends request and gets updates, everything happens server-side, well, nearly everything, at least everything combat(and animation)-wise are decided on the server-side and just user input is forwarded by the client (things like respec or real powers for instance use data on the client side plus eventually some constant data received from the server ages ago, like the chosen powersets and currently slotted enhancements). Since you would know some lower bounds to some effects you can program the client to actually don't forward things the player can't do anyway, but ultimately the server has the hand and say 'it's ok'. If it was like you said, if the server kept in sync with the client, how could you have powers showing ready when they are not ? How could I animate Haste (though only the excitation/shrugging, not the pompoms) when I activated it when it wasn't server-ready ? I believe the client does not sync the "animation", the rooting for sure (because it defines what the player is able to do) but not the "play flare animation" part. Animation is fluff, feedback, rooting is an effect, even if it's "animation rooting", isn't this why some powers don't root anymore (non-offensive PBAoEs anyone ?) Or am I missing something

This kind of delay always happens when you have numerical systems (instead of analogical), time sampling tends to create delays if not resonnance, at least if you don't compensate for it, which is computationally expensive.

What's actually the most unnerving in network play ? movement lag. Power lag, everybody can do with it, but jumping a fraction of a second too late and your down the hole instead of on the other side. So movement has a special treatment in network games, everytime. Since you have the network lag, it's no use doing it 120 times a second, that's why the Entity placement engine is on this 1/8 timer, and power launching is the duty of the placement engine because line of sight is one of the requirements. But the combat/effect timer can be as fast as you want, basically, because you have events based on time, and in a second it doesn't matter if you loop 2 or 6 times because you'll have to do as much processing (well, not to maintain every timers) but you'll have as many messages to send.

Sorry, didn't mean that you were an idiot, I meant that reading what you said made me think you were an idiot, and I can misread, or read too literally. You could have seen from the answer I took your "data" was the power data, not the entity data, the memory that represented the actual state of the entity. But I did have the first misjudgment so I do understand why you just passed over that =) Also, I'm handicapped, because English is not my mother tongue.

[/quote]On top of that, many fields are themselves indirectly processed, like the Expression fields which themselves require an entire scripting engine just to process. And parsing the grammar of the Expression fields is exactly zero percent of the problem. The real problem is writing the code to properly evaluate all of the indirect references to the game. Which *itself* requires entity tracking code to allow for those indirect references to go somewhere. AttributeRequires as a single field could end up taking ten times more code than fully processing all of the damage-type attribute fields because of that.[/quote]What's this "expression field" you're waving around ? never heard of that. That's the list of power effects ? Codewalker already said it was done, and I'm not one to let done work spoil.

That's still only 2 things to do differently, and since you've got functions, with parameters you can code pretty much any number of cases in your functions. It doesn't reduce the complexity though, I agree. It has to be somewhere. But you don't have to code a function for every existing combination of effect. What's the parser for then ?

EDIT: As I lied down to sleep, I understood your point about damagetype attributes and health now. That does shake my already-laid coding plan, but I'll make up a workaround this night I think, else tomorrw, else, with an ugly hack.

That could sort things out. But do we agree I just have to produce a library that handles entities and all their attributes (inspired from the attribute window plus every hidden attribute/fields I need for the implementation) and "example" functions that apply all power effects that I can think of and translate it into calls to that library ? Then the next stage would be executing said function when parsing the matching power's effects part.

Okay so the action loop is filling the event loop's data ? Action's loop is processing the "launch power"s and the event loop the "Do X points of damage"s ? (notwithstanding all the other things in the event)

[quote]Nope. Every bit of data implies hundreds of lines of code to do something with that data.[quote]NO NO NO NO and NO for god's sake ! What kind of project manager believes that ?
Once you've implemented something able to decode the data behind, let's say, the Nemesis Staff, it's far easier to do the Force Field.Force Bolt ! But you can't directly do Smite then, except if you've done Shadow Punch ! You have a baseline of minimal data to process and then all the data can be processed by the engine ! It's not because you've got 1000 powers and that it takes 1000 lines to implement Brawl at first that it's going to require 1000x1000 to implement everything ! Gosh !

Of course. In my mind I assumed the person has access to the Mids representation of the data, or the way the Titan Network produce this data from the most obscure representation. Anyway it's the same data.

True, true. Benchmarking But actually, we have been writing the tech specs nonetheless, at least it's table of content.

I actually miswrote that. I meant: What makes you think the 1/30 sec loop is the animation one ? Is it exclusively animation ? I think it's the combat/vitals loop, and the 1/8 sec loop the positionning loop, and animation is just a by-product of the power part of combat.

Ok, so the only hard part is thoroughness, you say ? My thoughts exactly. But most of the thoroughness is already made by datamining efforts done by the Titan Network with City of Data or the Mids. Essentially, my estimation is that when you get to implement every player power, you've got all the mechanics, and then you just need the bulk of the data, that is not the combat engine, actually, it's the NPC data, the powers each have and what each power does. But I also think what remaining out of our reach is the "which powers has every NPC" database, because we can already have the real numbers of NPC powers.

It's an example of what, actually ? When does this get involved, and what does it do client-side ? I did throw grammar at it because I didn't know what that was. I do want to know when and what because that could help me understand the mechanics, not reinvent an emulating mechanism. I know you don't invent things, but I was as puzzled by this as much as if you did. I could have called that an "*** pull". If it's power-related and not in the City of Data database, it's useful. If you just applied that fancy representation just for the pleasure of finding me dumb-founded... well... I don't see the point.

ho ho ho, now you're getting interesting ! Okay, I take the bait. Though I don't have access to the database, so I'm going to make a dozen sample powers each from a different set, and that doesn't count, okay ? Also, I may cheat a little because I did code it a bit in my head so I have a headstart (heh). Language is C++ but no inheritance (so C code-efficiency but easier time writing the code).

I don't know the functions you implemented nor how you implemented, so I cannot judge. Even if I saw the code I couldn't even fathom the number of times you rewrote this or that.

You're thinking the seconds2frames is made while the game runs, of course that's inefficient. You've got to pretreat the data. Is it still inefficient ? I think it's the least-size representation, since the seconds get rounded to frames anyway. And least-size is always the most efficient (given everything else being equal). In truth I can't see what you mean by making the power activation period explicitly supported in the primary event loop, except what I'm doing: expressing the delays & time in general expressed in event loops, that is, frames (which is the end-product of a loop).

It *will* have all of these, in some form or another of course. The values each power has of these "power attributes" is supposed to be part of the data used by the engine. True, the Mids database doesn't have all of these. (stacking and near ground at least). I didn't think of suppression up to now, but that would take registering the last supressing effect's occurence.

Well, even if it's not the same engline but the same results, it has the same "complexity", but As Codewalker said, there is no need simulating when making the engine, implementing existing designs is far easier than making up a new engine with new designs or a one-shot simulator of each power.

Ranging from 2 minutes to a week-end, depends on what that humanly means :S Looks like a coma-separated... thing, each thing between comas is a token meaning... function names, operators and parameters ? You would need to set up a grammar in Yacc or Bison (heh !) and make your hearts desire come true.

First implementation: A "check toggle" clause for effects (next to the "onHit", etc, a "if X power toggled on"). Or, given this could be used for other powersets, just make a new "attribute" that's set by the toggle (it's just not a "vital" attribute of entities, it's here only for certain powersets).

But there is something you can use already, the "mode" variable every entity has. Like "disableAll" etc. These are just bits in a bitfield right ? It can be set/deset as part of effects too.

Obviously, you pass messages to the UI. And this is done as part of the powers. So be it, also set these "flags" or attribute. Use the decay (inverse to regen, used for Fury & domination) mechanism already implemented to reset this attribute to 0 (no combo) (with infinite decay every T seconds) and use specific values to test for equality with subsequent powers.

Or you can use modes or a mode-like system. A special bitfield for use by powersets, like the dual blade, dual pistols, or street fighting, Titan Weapon, Staff Fighting (notice these are all offensive powersets, so mutually exclusive on one single toon). Fiery Embrace is an exception, but it's a single power, it could manage with only one bit of the new bitfield. Each new power could use specific bitmasks to apply to the new bitfield, given the effect. This has to be done for these 5 or 6 power sets, and you have to code the method to parse these commands, but it's not goint to take 10 hours.

The second, obviously it's the simplest to implement and not quite inefficient. Why reinvent the wheel ? "Fury" is built as part of the melee powersets (all of them), just that the fury stat is not used (nor shown) to other ATs. You forgot Domination, that works the same way.


You already did that. And listing them would be technically beginning to implement the engine.

What makes you think there is a 30/sec animation loop ? I think there is no animation loop at all, it's just animation notices sent to the client, and eventually rooting notices to the the movement enging. If there was animation "synchronisation", it would never desync the Flares animation from floating text (which are also noticies to the graphics engine to render. Maybe a max of 3 of them / entity / frame + a total in case of overflow, in order to keep vitals in sync) or at least it would try to reanimate when at point blank. Which I've not seen the game doing even in very bad server lag situations. (like Nova'ing the hell out of towers in the last cimerora mish. I did see reanimating powers in traffic lag when duplicate packets were received.) If there is synchronisation, it's because the animation notifications is part of the combat engine's responsibility, because it knows the power launching, and thus, their animations.

That should be done only when it's both more efficient server-wise and if it's compliant with the present behaviour.

So what's difficult in doing that ? Here you *specifically* need script1 calling script2 and each needs to be different for it not to loop. Of course the better way would be designing a "chain" target zone, next to single-target, PBSphere, and targeted sphere, and that's more difficult because you have to change the data. But that's only a handful of powers. And I did specify that you could take twice as much time to do something better instead of just reimplementing the game.

That requires either a new mechanic, or pretreating the powers. I wouln't like the additional complexity of adding the possibilty of a single effect affecting only a select numbers of target, I would do the 0-range sphere pseudopet technique, even if its displeases me. But hey, I'm not perfect.

[/quote]Do you directly implement exactly how our game engine deals with buffs that buff themselves? Or do you fudge, again by hand, Combat Training: Offense?[/quote]You mean things like resistible resistance debuffs ? Hell, I would straightly implement it without trying to fix it. It's part of the compliance. Buffs are always Irresistible and/or "Ignores Enhancement & buffs" to try to diminish these effects, even for resistance (doesn't take +dam into account)

I'd pass the hot potatoe here, I only signed for a combat/vitals engine, the state of the enhancements is modified by another engine and taken to combat engine as input when the entity is a player (Enhancements could be carried to the pseudopets by way of buffs, because an enhancement is a power-restricted buffs, power that spawns the pseudopets).

Yes but you can't change what's an engine and what's a data. When you had issue 7 (or the one when city of villain came out) you could have inserted the Demon Summoning Powerset without changing the engine (well, maybe the graphical one), but you could not do the same for the Dual Blade or the Bio Armor/Nature Affinity one because it's using all new mechanisms. Once you have put the Nature Affinity in the game (engine + data) you don't need to modify the engine to make Bio armor work, just list a set of powers, that is data, that the engine interprets.

Conclusion:Hey hey hey ! we nearly went over all the mechanics of the combat in the game in a single post ! And you still think this would take 1000 hours ? Okay, I'll give you the time we are spending discussing this

I knew you were too exhaustive. The combat engine I was thinking about only encompassed the vitals, not the positionning. Making an engine where you can run/jump/fly is basics in any game and engine, and *it* should handle the line of sight, just pass the command to launch the power through it and *it* will queue the power until it can get launched, where it will notify the combat engine that will continue to queue it for (eventual) animation reasons, and pass it back to the positioning engine when ok etc... as long as they don't agree back-to-back (in the same frame), in which case the power is launched. So I'm only talking about the power/animation/vitals engine.

The key here is not to think of arcanaTime as a thing to implement, put it in the engine, like fighting game do. Time (as in wall time) doesn't really exist, only frames do. And every player has seen evidence of this in CoH, because when it lagged in Cimerora, the time was lengthened because the game needed more time to compute every event that happened at the same time in a "frame". It doesn't need to be "time-accurate", because it gives the "tic" and everything waits on it.

Graphics and game engine were tightly intertwined back in the old days, basically because you had no need of processing events inbetween longer-to-render frames, or at least so the game devs thought. Basically, the whole server (I think that's how you discovered ArcanaTime) is frame-based, it's just a receiver that gets notified of entities (including user) input (including placements) and executes scripts - the powers. Given a programmer that had the full specifics to leverage the Mids database (which I suppose he could ask the Titan to give), he already has the power data, he just needs to translate it in frames and effects and so it's just a request-based server that takes requests from the entities and produces everything you talked about: the recharge, the effects, etc... It's just translating each and every power effects into a modification, formulas to apply to values and delays in frames (these delay are calculated in frames at launch, either because the distance between the to entities and the velocity of the projectile is know, or the effect just states "after Xseconds"), and recharge is translated into "points" that, if full, the power is recharged and at each frame, the recharge rate is calculated to produce the number of points given to each power on cooldown.

In the case of CancelOnMiss of DoTs/HoTs, for instance, these are just chained scripts/effect, where the first application (what's actually listed in the power effects) is queue script2), where scriptN contains (OnHitapply damage and queue scriptN+1)). and ends at scriptM for a DoT with M ticks. In the case of pseudopets, spawn the correct entity.

I may have jacked up the calculus because I supposed the power effect lists are a given, but if you consider power effect as data, and not the engine, it is *not* upward of 1000 hours to make. Maybe not 3 week-ends as I said, but still, 1000 hours ? That's taking far too many feature into account, not just the *power/vitals* engine. Plus, I didn't include the AI, just the thing that would make everything "live" and tick based on AI and player input.

And also, I want to say that it's how *I* would implement it, and it should be compliant to how the game plays, if it's not *the* way it's implemented. If I'm wrong about this idea, I'd like to be challenged with cases where it doesn't work.

You're thinking one programmer has to do it alone, but the right analysis can break down the problem into thousands of pieces and then make is combine gracefully like the devs never imagined before because they use old tools. To simulate the combat dynamics only, one wouldn't need more than 2 or 3 week-ends, double that to make it optimized. I mean with the present data.

One of the biggest stops right now is not being legally able to use the assets of the game nor being able at all to use the assets on the servers. The assets that really give life to the game; spawn points, scripts and conversations.

To program all that from scratch won't take long, it's the content that will take time.

Everything is possible in the software world. Even making something exactly the same, it's just going to cost time and time for adjustments, but it's not infeasible to at least have something similar that just needs to be tweaked around the edges.

Do bear in mind that if we recreate something today, it won't be with yesterday's tools

Fanks for the splanation.