Task System (preliminary work)
 

I've been looking at whether it would be possible to get the Task System working. The OPcodes and
structs are pretty much squared away (a lot of this was already mapped out in the code, so I just
filled in the gaps with some info from a packet collect by Solid11 over at PEQs repository).

At the moment, I have written code to send the relevant OPcodes to the client and I'd thought I'd
share.

Bring up the Task Selector:

http://www.rama.demon.co.uk/taskpics/task1.jpg

Send Task Description and Activity packets:

http://www.rama.demon.co.uk/taskpics/task2.jpg

Send Task Stage Completed packet:

http://www.rama.demon.co.uk/taskpics/task3.jpg

Send Task History:

http://www.rama.demon.co.uk/taskpics/task4.jpg

I don't know what should go in the 'Quest Progression' part of that window ... maybe an Opcode I
am missing. I'll have to hop on live and see.

I haven't seen any packets relating to the 'Shared Tasks', but that is not a particular priority for me.

At the moment, all I have are some #commands to bring up these windows etc, as a test. There is still
a lot of work to do to integrate it into a working Task system. If anyone has any thoughts on how to
tie it in to, for example, the quest system, feel free to post your ideas.

Hahaha! Talented and funny!

Isn't this one of the primary roadblocks on any sort of LDoN functionality? I know getting any form of instancing working is also a pretty big hold-up, but I believe that the original form of LDoN didn't really have any -- It was simply 10 different versions of the same zone that could handle a group each. The Rujarkian Hills 'instance' was ruja-rujj, for example. Though I'm sure this is already common knowledge around here.

Just wondering, since that's the first I remember the task window ever coming into play. I know it was used for quests after that, but I didn't play much at all during that period and am thus unfamiliar. Was it retroactive at all? Did it handle some older quests by chance? My gut feeling is that it didn't.

Dude!

I've been thinking about this for a while, but haven't had the time to get into it much. I think it would be great if we could tie it into the quest system. I haven't come up with a good solution, because I just don't know enough about how it works within the client. Basically, how it's passed to the client. Is all the quest data passed into the client at the time the quest is assigned, and then how do the updates affect it?

I was thinking about doing something on the order of quest::taskassign(blah, blah,blah) or however it worked out to assign tasks. Then quest::update(blah,blah), or what ever as far as how the data is then passed to the client. That is if it could even be integrated into the quest system.

Anyway, that's an awesome start. Thanks for taking it on...

Having no idea whatsoever how it's actually handled, I feel perfectly qualified to pitch an idea for how it should work.

Initiate a quest with an NPC. That should pass the quest completion criteria to the client's task window.

quest::task(type,quantity,id);

Where type is the type of quest (collect or kill), quantity is how many must be collected/killed, and id is either the item ID or NPC ID.

Is there an op_code triggered by looting a certain item ID or killing an NPC of a certain type?

Feel free to shoo me away from this thread. I know I'm in far over my head. Just pitching that out there, though, as a rough example.

Nah. The older quests were never "retrofitted" for the task system, which is a failing in Live, IMO. Would have been handy for those long epic quests. Or even the smaller ones. I think most of the "older" quests eventually yielded mostly crappy, in comparison, stuff so was thought to be more work than it is worth. It would be really nice to work it into the emu quest system though so as to rectify it if possible. We still have a lot of older quests that haven't been completed that could be fitted to make it work if it was possible to have it work that way.

No, the way I see it working is, whenever you kill a mob, or loot something, the server will need to check if you have any active tasks, and if killing that mob, or looting that item, was an objective of the task, then incrementing the kill/loot count for the task and sending the Task Complete packet when you have killed/looted the required number.

I welcome anyones ideas. I never actually did any of these Tasks when I played on live.

That sounds about right. I'm trying to remember the different quest types. I remember kill and collection, as those were the easiest. My friends and I pretty much kept requesting a new task until we got one of those, heh. Again, though, this only applies to LDoN. I suppose it could apply to many different quests that came later and fit the format correctly. Might need someone with a little more insight as to how it was used on Live, but your understanding of it sounds pretty spot-on thus far. It's the application of the system that's going to take some understanding.

There is an 'Activity Type' field, which determines, whether the activity is loot, deliver, kill etc. I've not tried all the possible values yet, but I know there is an 'Explore' type, which I imagine could be tied into the quest proximity detection in some way.

Most of the task quests involved kill and collect. There were locate explore type quests as well, where you just ran around looking for stuff. These were mostly to get questers familiar with the zones and such. I don't have the latest expansion so I'm not sure if there are other types in the later ones. I've done a lot in the DoN and a several in the Depths of Darkhollow. Most of those were kill and collect.

Yeah, set up an invisible mob/object and set a proximity around it. Have the quest for that object send a task update. If you have the task your task updates, if not the client will likely ignore it.

Awsomeness Derision, did you used to work for Sony? lol j/k but nonetheless great work.

Damn edit button

The main things i remember from the task system was

EQlive must of kept a table for what quests you completed. Becuase some of the quests where long 12 mission arcs.
DoN used it heavely for its raids/group content.
GoD/OoW used the alt+z window for it's missions. i think they are alot different but on the other hand might be alot alike /shrug.
Ldon used alt+v i think

Basically, we can already handle this kind of stuff via player quests too, proximity mobs for explore etc.

Would add a internal task tracking system for time, completion/failure and maybe item amounts since that would be cumbersome then could probably do the rest via quests.

quest::updatetaskvalue(taskfield,value)
quest::enabletaskforplayer()
quest::disabletaskforplayer()
quest::opentaskwindow()
quest::istaskcompleted(taskid)

etc. probably more or different.

Biggest issue if you've gotten most the packets worked out will probably be trying to with some finesse track the quests offered. Since quests offered to a player will vary based on: who the player is(race,class), what they've done(tasks/quests complete), which npc is giving it.

That said if you need help and I get time I'd be most excited to help you hammer out a system to do it.

Maybe quest globals could be used for tracking purposes for tasks? They sounds to be similar systems and I think you would just need to allow the task window to work with the quest globals system. If so, it seems like it would minimize the work needed to get this finished. Or maybe use some of the code from the globals system to implement something similar to be used only for tasks. I am sure you would need a table for tracking them and similar fields.

Actually it would probably be harder to integrate it into an old system than to create a new one as the system would probably be fairly straight forward really. For performance issues I think also it's probably not a good idea; think every time you have to do a quest action you need to re-export globals, integrating it into qglobals would probably inc the number of globals in use by a lot and make it just that much more work every time the server processed quests.

Yes, I was looking at the quest system this morning, and that was my thought.

While I am at work, I thought I would start sketching out how to implement this,
and welcome any thoughts you have on my tentative design choices.

First the tables. Global tables for the stuff that doesn't change (once the task is written),
and character tables to record their progress.


The active and completed task information is not retained by the client, certainly not across
logins, and probably not across zoning, so needs to be resent after zoning.

As I see it, zone will need to load the global task and activity tables in their entirety on
zone bootup. This may become an issue later on if their are lots of tasks written. As this is
global static data, it could be a candidate for shared memory, but I will leave that for the
future.

The next issue is how to represent the Tasks in memory. Each Task has a unique ID. I am leaning
toward #defining a MAXTASKID (say 10000 ?), and using a fixed size array ...

TaskInformation* Tasks[MAXTASKID]

The benefit I see of doing this is fast cross referencing to the Task/Activity information from
the per-character data. I think this would also make it easier to #reloadtask a single task
during development.

Onto the per-character info. Should I restrict a character to a maximum number of active tasks, e.g. 20 ?
The benefit I see to this is it constrains the amount of checking that needs to be done on killing/looting
to see if there is a matching goal in the characters active activities.

Pseudo Code:

On NPC Kill or Loot:
for each active task the character has:
for each open activity in that Task:
if ActivityGoal == Kill This NPCType or Loot this item.id Increment Goal Count

It did occur to me that an extra flag could be added to the npc_types and item tables, 'involvedintask',
so that the code could check this flag and if it is set to No, not bother checking the characters
active tasks.

Task initiation will be done through the existing Perl quest system, so the usual checks
for class, race, faction can be done before the NPC initiates the Task Chooser. (Also provide
a function so that a check can be made in Perl to see if the player has already completed
a particular task, so as not to offer it again).

I also see Task Completion/Reward as being handled by the Perl quest system, i.e. the last
activity in a Task would be go speak to an NPC who would then do something along the lines
of:

I would be interested in any links to quests on Alla that use the Task system. I've
been using http://everquest.allakhazam.com/db/q...tml?quest=3237 as a guide, as
that is the one in the packet collect I was looking at.

Looks like you have a good handle on it, and your initial approach seems viable to me. I would think that limiting the number of active quests a toon can have is wise at this point. Once you get started implementing it, I'm sure you'll come across other challenges. As for tasks, I'm mostly familiar with these...

This is an example of a collect task:

http://everquest.allakhazam.com/db/q...tml?quest=4320

This is a combo kill/collect:

http://everquest.allakhazam.com/db/q...tml?quest=3065

I know we're not up to DoN yet, but these tasks are examples.

See this to me presents a unique problem in that it would be cumbersome if not impossible to write quests to include all the tasks a person has from a given npc. Keep in mind some task givers are giving out lots and lots of tasks, especially the ones for shards and such give out literal boatloads of tasks.

Personally I'd do something like:

Would allow you to group tasks together into tasksets and the quest code could be simplified to quest::BringUpTaskWindow(taskset). Different npcs offer different tasks and more than 1 at a time if you qualify for them on live.

Would also need some way of enabling the tasks if one were to go with such an implementation.
Could then offer quest::EnableTask(taskid), quest::DisableTask(taskid), etc to make it much simpler to write quests that use the task system.

Putting it in shared memory is also probably a better option as the structures are probably going to need to be in some cases quite large and with the assumed 10000 entries it would really add up for a server that loads quite a few zones. A couple MB per zone doesn't seem like a big deal until you load up 50.

Ok I have to say this is awesome! Ive been waiting for tasks to work since i came to eqemu! Awesome work derision

I already have the feature coded to call the task chooser from Perl with up to 10 tasks (could be increased by changing a #define):

So the initiating quest can build up the $args string with the list of Tasks the player is eligible for. When I've got the whole system working end-to-end, I'll revisit this to see about implementing it as you suggest.

Shared memory I will leave til last, as it's not something I've ever worked with.

First, allow me to say that your findings are very promising. A good Quest Journal was really a missing part of EQEmu (and in fact of EQ for some quests).

I did not play with the task system very much on Live, it really came when I was taking a break from the game. There is one thing you should be careful of : may be some Live tasks have simultaneaous activities. Something like "Make 7 flame Posts" and "Get banned for one week", that you need to complete before you can progress to the next activity. As you see I am sticking to your examples spirit :)

Regarding the design, you could make the Quest Journal have even more reach by separating the low-level Journal handling from the Task system. In short, it could allow quest designers to track quests in the Journal even for quests that are not formatted like tasks. Let me take an example, this time from a quest idea I actually started implementing through the normal Perl system..

"The young gnome had been playing a few pranks to several personalities in Ak'anon, following the directions of the enchanter guild mistress. The last one had put him in hot waters though : passing for a half-elf had been easy with the illusion spell the enchantress had given him, but he was still shaking about the thought of what would have happened had he answered to the necromancer in the wrong way. But the real problem was : what to do with the letter the necromancer asked him to deliver to the dark elf necromancer guild master in Neriak ! Clearly he had gotten more than the enchantress had hoped for, surely she sent him there on purpose, not just to play one more prank. He could certainly bring the letter to her, or may be actually delivering the letter to its actual recipient would make him rich. Neriak was a wealthy city."

At some point the character can choose one of two major paths in a quest. Having the Task Task system handle that could be tricky. Handling arbitrary quest progress triggers (like we do in Perl sometimes) would be overkill. It is made for simple tasks and let's face it, many quests are like that. On the other hand it would be very nice to allow regular Perl quests to benefit from the Journal, given the quest writer accepts the extra work naturally.

You could achieve that by untying the actual quest progress (task progress) from the Task system logic. The tables would be unchanged, except the comment for completed would not apply : for arbitrary quests donecount has no meaning. The activity would be complete when the completed field is filled.

What does actually change then ? Answer : the way the tables are managed. At the moment may be you are thinking of wiring the table management into the task system. Instead you could separate the table management, and expose all "atomic" operations on those tables as quest:: xxx calls in Perl scripts, on top of using them for the task system of course.

Quest writers could then decide when to :
* quest::start_task(taskid, activityid) : the quest could start at any activity, defined by the Perl script
* quest::complete_activity(taskid, activityid) : completed date set by the server
* quest::start_activity(taskid,activityid) : this activity is started. Several activities could be up at the same time.
* quest::complete_task(taskid) : completed date set by the server

Of course those "manual" activities would not track progress , the Perl script does. But the player would at least see the quests he is on and the text for the current activity.

It would be the quest writer responsibility to start the next step(s) after completing the previous one(s), to detect when an activity is complete etc. A few checks could be added to avoid inconsistencies, like tasks completed whereas some activities are not complete for this task. But the code need not ensure an open task has an ongoing activity. The Perl scripts pilot the Quest Journal, and try to reproduce faithfully the facts. The Task system in this case is not active. With an activity_type of "Unmanaged" or somesuch it would be easy to deactivate the task system for those "quest" tasks and activities.

Ultimately this could lead to a flexible, semi-automated quest/task system : quests/tasks and their steps would be defined in the DB, with the transitions from a set of steps to the next ones. Each activity/step could either be handled by Perl, disabling any automated counting of objectives, or be fully supported by the engine for easy activities (loot, kill, explore). This would just depend on the activity_type. But it can be built one step at a time :)

By doing that we could have an emu better than Live with regards to quest tracking : even old quests could be tracked by the Quest Journal, given they are reworked of course.

I suggest this because I doubt it would be much more work to structure the code to allow the Task system to be optional with regards to the Quest Journal. Most of the things to write are the same, you just need to let the Task System pilot the Quest Journal, not intermix them and make them dependent on one another.

I hope this did not read like gibberish :)

That's a good point. The task I have been using as my guide has a sequential set of activities, where one must be completed before the next is revealed.

There are two types of task activity packet. One for completed and 'in progress' tasks which sends the details of the goal, and progress toward completion.

The second type is really just a placeholder, which consists just of an Activity Number, but no details of what the task is. These show up as ??? in the Task Journal.

It would be easy to alter the table and code to add a predecessor activity field. I know, (and you can see from the example in my first post), that it is possible to have more than one activity in progress displayed at the same time.

As for your other points, I skimmed over them, but they sound interesting, and I will look them over again and see what's possible as I continue to implement the system :)

I've been beavering away at this since Friday and thought I would provide an update on what I have functional so far:

Database Load routine for global task data
Database Load/Save for character task data

The database save is synchronous, but it only writes out anything that has been changed. I'll have to see if this can be done asynchronously like the saving of the other character data is.

Active task information is sent to the client when it enters a zone and so is retained across logins and zoning.

Initial Perl interface to bring up the Task Chooser with a list of up to 10 tasks:

The way I have it working right now, is the NPC has a list of all the tasks he can offer ($tasklist). It then checks whether the player already has each task in the list already assigned, (quest::istaskactive) and thus builds up a list of tasks it will offer ($tasksoffered), not including any tasks already active. quest::taskchooser then brings up the task selector window.

I have also added a new sub EVENT_TASKACCEPTED. This tells the NPC what task you accepted, so it can respond appropriately, summon an item etc. (When you accept the Children of the Fay task, you are given an Elven Bottle of Wine that you have to deliver).

You can remove a task by clicking the Remove button in the Task Journal window.

Two more Perl quest functions, quest::istaskactivityactive(taskid, activityid) and quest::flagtaskactivitycomplete(taskid, activityid).

This code checks when you hand in the Elven Bottle of Wine to see if you are on activity 0 of the Children of the Fay task (task 1), and if so, accepts the item, and flags that activity complete, sending the 'Task Stage completed' message, and unlocking the next activity:

I'm still undecided about how to implement the predecessor activity requirements. E.g. say a task has 10 activities and when you completed activity 1, activities 2 and 3 open up. Task 4 then won't open up until tasks 2 and 3 are complete.

As I see it, I can either put a limit on how many predecessors an activity can have (let's say 4), and then have four columns in the table, predecessor1, predecessor2, etc.

Or, have a text field which could contain a comma separated list of predecessors, e.g. "2,3". I dislike this because it means I have to parse the string to get at the predecessor info.

I'm leaning toward option 1 (a fixed, small number of predecessors and having a column for each).

Still lots left to do!

I think something like the "_entries" or "_metadata" tables currently in the database would work best, similar to what KLS recommended above. Here's an example:
"step" could be used to group each stage, so using your example, activity 1 would be step 1, activities 2 & 3 would be step 2, and activity 4 would be step 3:
That way, it's more dynamic, and doesn't create a large table if you want something more than 4 prerequisites deep.

Hope this gives you some ideas, the whole task system seems REAL promising :-)

Thanks! I think just adding a 'step' column to the activity table, rather than creating another table should suffice.

Anyway, I've hit upon another snag. I've just implemented the code to update 'Kill' and 'Loot' activities, based on NPCTypeID or itemid respectively, which works great, however ...

I wanted to add a Kill activity to kill 5 Orc Centurions, but there are 40 entries for Orc Centurions in the npc_types table!

As they say, no plan survives contact with the enemy! I would welcome ideas on this one. Doing a lookup against a list of 40 NPC IDs per kill seems a bit OTT. Maybe some sort of sub-string match again the NPC name may be a less expensive way to go.

Hmm. Could be you just let any of them meet the requirements. Or you could just filter them out like you say. It probably makes more sense to filter them out and then let any of the resulting id's after you filter them to your satisfaction, meet the requirements for the task.

I think an easy way to do this would be to allow either a # (NPC ID for a specific mob) or a string. Then, the server can check to see which it is. If it's a #, use the specific NPC ID. If it's a string, allow multiples based on the string (SELECT name FROM npc_types WHERE name LIKE '%an_orc%'). I think it would be a lot easier to utilize overall (an_orc for example, instead of an_orc_warrior, an_orc_oracle, or an_orc_pawn). However, in the same token, I believe it could cause problems for mobs that have different names, but are different (lower) levels (read: exploit).

As much as I like the first option, I think the "better" way to do it would be to create a similar lookup table like I mentioned above. Yes, it is another table, but it makes it much more powerful overall. That way, you could also include named mobs that are rare spawns to include towards the kill target. Yeah, it seems like more work in the database, but that's also what relational databases are for: creating complex relationships between data sets :-)

Use quests for the kill component =p

*smacks forehead* I should have had a V8!

Thanks :)

I've now mapped out some more of the activity types:

So 1, 2, 4 and 5 can be handled through quests.

3 I already have handled by hooking into Corpse::LootItem, and I imagine 6, 7 and 8 should be done by hooking into the code as well.

I'm guessing that the 'Touch' activity means clicking on a clickable-object, like a door or something. I found a task on Alla that has 'touch' as an activity:

http://everquest.allakhazam.com/db/q...tml?quest=3151

But it's not clear what it means.

I'm guessing that the 'Touch' activity means clicking on a clickable-object, like a door or something. I found a task on Alla that has 'touch' as an activity:

http://everquest.allakhazam.com/db/q...tml?quest=3151

But it's not clear what it means.

It means zone into a certain instanced / zone basically ~.~

When you say : for kills, handling through quests, do you actually mean patching/creating many NPCs Perl quest files ? There can be an orc_centurion in several zones, some of them actually "dinging" the quest..

As the kill activity is very common, having a lookup table (ID of activity, ID of mob) seems worth doing. Especially as you can attach this list/set to the activity in a field once it is loaded from the DB. It probably is not in the scope of Live-like, but extending this to several activity types you could then have activities that accept looting "any rusty weapon". Not trying to make my shopping list for the first implementation, just trying to keep doors open here :)

Or may be you thought of another way through quests ?

As a professional programmer let me say one thing : filtering on the name (or a part of the name) will inevitably create unsolvable situations, or to say it bluntly, it will suck. It always does when you start doing such things.

I took it to mean calling the task activity update method from within sub EVENT_DEATH, which works.

My concern was, in the case of an Orc Centurion, having to check through a list of 40 NPCIDs every time the player killed something to see if he had killed an Orc Centurion. It occurs to me now that if the list is sorted, I could do a binary search on larger lists, so maybe it wouldn't be so bad :)

I could always include both methods and have a flag to indicate whether the activity count update was to come from a quest, or be checked from a list.

EVENT_KILLED_MERIT was designed with mob kill quests in mind. You could make 40 different quest files or you could just make a single an_orc_centurion.pl. It basically just simplifies the problem that was discussed above but is hardly the only solution.

You really have two ways of processing quests : the configured way, and the custom way.

At the moment EQEmu "shines" in the custom way. I put shine in quotes because I hate Perl, but other than that it is very much open-ended :) With the Perl scripts you can do a lot of things, it just takes time, knowledge and patience (did I mention "time" ?).

The other way is what wow-emulators provide : table-driven definitions to easily specify kill quests, loot quests, explore quests and a few others. The list is limited, the configuration tolerated is limited, but it is very efficient (in that the server handles it, and it avoids the silly Perl syntax errors).

Both can co-exist. The server can first check the "configured" quests, and then the fully-custom ones. At the moment EQEmu is lacking in the second option, which covers a big portion of the quest/task neeeds. If we had that, many "newcomers" could write their own basic quests, by tweaking tables instead of writing Perl. I am not only speaking of "Live"-like quests there, many custom quests would become very easy.

This is really what I was hinting at, that there are two parts in what you would like to implement, Derision :
* Exposing the functionality (opcodes) allowing the Quest Journal to work. Providing the functionality through functions in C++ and possibly Quest:: Perl function calls. This is the basis for task tracking.
* creating an "easy-quest" framework, using the requirements of the EQ Task system as a guideline.

You can have all those alternatives in a game :
* easy and limited quests with no tracking
* complex and open-ended quests with no tracking (as it is at the moment)
* easy and tracked quests (Quest Journal + Task system, what you are trying to achieve , and what basic wow emu supports)
* complex and tracked quests (what I am suggesting you enable by opening the Quest Journal implementation)

Not trying to scare you off there, but the task you have started encompasses all four, as adding the third one to the second one is close to enabling the four programming-wise. If your design is good you should get the other two as a nearly-free bonus. If not, well... You will still get the third one ! Just do not be puzzled when you realize the task you embarked on is tough. eh eh. The end-result will be awesome, but the price has to be paid !

At least that is my analysis of what is going on :cool: I wish I could offer my help on this project, but I will be quite busy in the coming month. Still if you would like to submit some interfaces (function or quest:: Perl prototypes...) for review, some of us could find the time to provide some feedback. It is up to you.

I think you can forget your worries about scanning 40 mob IDs in a list, as long as everything is already in memory : our regular CPUs do that in no time. Doing a binary search for that small a number is not worth it. From 500 IDs on OK may be... Or may be from 5000... You have to profile that to be sure.

What matters is that you read the ID list only very rarely from the database. DB queries cost. If it is all loaded in one shot at startup, then it is not a concern. Just make sure you do not issue one DB query per activity, as it would be a killer. For example, issue one query to load all activities at once, then one query to load all dependent mob IDs at once, then do any activity-dispatching (list creation and populating) in-memory. If you add proper "ORDER BY" sections to your queries you can even scan your activity list and your mob ID list in parallel, making it a very efficient process.

I've been busy with work this week and not had as much time as I would have liked to work on this, but today I figured out the Titanium OPcode to flag a task as complete (which is different from 6.2) and make it display the green 'Task Completed' message (I think I have found the struct values to make it display 'Task Failed', but I haven't got around to testing that yet.

http://www.rama.demon.co.uk/taskpics/taskdone.jpg

The more astute amongst you will note that the 'Children of the Fay' task does not finish in Kurn's tower, however I used a bit of 'artistic license' is setting the task goals during development.

There is still quite a bit left to do, but the system is now functional end-to-end for tasks with talk/kill and loot activities. Hopefully I will get a lot more done over the weekend.

That is so awesome! I am currently working on a custom zone that will use qglobals to track kills similar to tasks. But, it would certainly be cool to use the real task system for it instead!

I've started to document how the system works on the Wiki, the overview page links to the pages I've written so far:

http://www.eqemulator.net/wiki/wikka...SystemOverview

There are still some Perl interfaces to be written, and a few tweaks to the tables, code cleanup and testing (robustness against bad data in the tables, discovering the limits on the number of active tasks, task selector entries, completed tasks, etc, that I can send to the client without it crashing) before it will be ready for an alpha release.

Dude! Can't wait... I love seeing stuff like this getting done in the emu.