Bukkit

PLEASE find a way to integrate this plugin with FrameStore.

I love the dynamic economy of HyperConomy, and the visually appealing design of FrameStore, but alas they don't work together.

@RegalOwl

About the futures data.

What I envisoned was this:

When the server goes down for regular backups, I query the DB and figure out the number of ores mined in the last 24hrs.   And, using some heuristic determine how many items to offset the median by.  So I understand the data flow/timing problem you mention (thanks for reminding me).  But in the short term, I may be able to work around it until something allows for dynamic real-time  updates (think:  WalMart feeds Proctor and Gamble databases -- which is how most big-box retailers work these days.. When a customer buys X units of Y, the database at the manufacturer is updated not just the retailer).

So, for example if I see a spike in Iron Ore mining, I may adjust the median for say, Iron Ingots, or Iron Pick Axe etc..

I need to collect data over time to model it -- throw it in to a JMP model or something to look for any correlation between ore mined and iron-based items sold/purchased.

As far as the code environment, I just need to get the right MCP environment setup.  I've dabbled with mods (FML) but plugins may be just as simple to setup in the IDE.   I run a MCPC+ server so I need to make sure that I have the right MCP version so I can test my mods to your plugin before submitting the diffs for your review and approval.

One of the things I will probably try to do first is make the formula configurable in yml.

I prefer data driven operations, so the formula to compute prices should (IMHO) be something I can actually configure in files rather than in the code.  I may want to put a dampening function on the price so that I can smooth out the prices..  Eg., if I drop 64 items at once to sell, I should be able to get the same price for all 64 (I think you do this actually already because you can set N (Line 4) of a sign).

I also want to re-work some of the offsets.. LIke instead of just tax to offset prices, I want to setup a discount/markup % as well.  So for a server shop, I don't really want to use taxes, but I do want to markup any purchases so that it always costs more to BUY the raw material than it does to sell it.  the use of "tax" to do this seems contrived.  Not a complaint, just an idea.

I also think about accounting for demand when stock is 0.   I get some players who ask "why are there no X in the server shop?" and on my server (which is 99% player run economy based) the answer is invariably "because no player has sold any to the server."

I need to know this data to make sure we're balanced.  We run a MagicFarm server (blatent plug:  FooFarm.net)  and so the idea of resources sold/purchased is at the heart of the design of player-run-economies.

I remember fondly the mud we used to play 20 years ago where we had player run breweries.
There was a clan that made barrels, a clan that grew hops, and a clan that brewed beer.  It all worked beautifully.  Beer was required to heal.    

Minecraft has taken this idea up an order of magnitude, and it's great.  Your plugin provides a real interesting way to implement these player run economies.  You got a good thing here.

@RegalOwl

Nah, you're cool. You got more time to code these things than I do.

I gotta get my Eclipse setup to make modifications to plugins. One of these days.

thank you for this awesome plugin!

@RegalOwl

I had it set to 1hr, then 5hr, then 24hr.

Meanwhile I checked the database and there were multiple entries of prices per object (as I demonstrated with the SQL query).

It was unclear at first until your post-reply that the data in a History sign is delayed.  

So, let's say for example that the history table has many records per object.  Then creating a History sign, the data in the sign is still "(?)".

Why wouldn't a non "(?)" result be the initial string if there was data to compute upon making the sign?  That is what was confusing. I thought I had valid data to compute a History but.. as you hinted, the result in the History sign is delayed (still trying to see where that time out is counted).

I also see that there are some magic numbers in that class..

For example, in History.java

                                if (timeCounter >= 3600000) {
                                        // if (timeCounter >= 120000) {

3600000 looks suspiciously like a 24hr clock time out.. But it's still a magic number in the code.  (just what SW engineers call the insertion of hard coded numbers in code -- magic numbers).  Over time, magic numbers in code tend to bite you in your 6.  A simple const enumeration in the class would be great.

About this "(?)" thing:

Just a programming idiom -- the return value of "(?)" for distinctly different rationales is confusing.  

Eg:

        public synchronized String getPercentChange(HyperObject ho, int timevalue) {
                if (ho == null || sr == null) {
                        return "?";
                }
                CommonFunctions cf = hc.getCommonFunctions();
                double percentChange = 0.0;
                double historicvalue = getHistoricValue(ho.getName(), ho.getEconomy(), timevalue);
                if (historicvalue == -1.0) {
                        return "?";
                }

It's confusing because the User (who is staring at the sign) doesn't know really what the problem is.  

If the handle is bad, we return the same string as if the handle was good and the data was not available.  Completely different situations.  Return values from functions should have distinct values for different reasons.

It's sort of like  CheckBankAccount() returns -1 if the account is unavailable because the database handle is null or  -1 if the account has been closed.

I can't return 0 for a closed-account because 0 could be the balance of an open account.
I can return -2 for a closed account.  And then I've just made two distinct return values for two distinct causes.  Good.   But I could go further --  Or in fact I could return -X for a closed account if the account still had funds X in it.   But then I cannot return -1 for a bad handle.  So more feature I add, the more I box myself in a corner.

Instead, the return value can be a structure (class object) so that the meta data in the struct has members:

code
balance

where code is an enumerated type:  E_NOHANDLE, E_CLOSEDACCT, S_OK,
and balance is just a float.

Then I can tell easily if the query was successful or failure by testing  Result.code != S_OK
else treat Result.balance as legitimate.

It would be better, IMHO that the return value be relevant to the meaning.   

No data in _history table => "N/A"
No handle  => "Error"

etc..

One more super small nit to pick (sorry, the embedded software engineer in me can't resist )

percentChange = ((currentvalue - historicvalue) / historicvalue) * 100;

percentChange is a double.

So if it was me, the operand 100 should be 100.0 in the code.  Makes it more readable.

I know Java doesn't care and does "the right thing",  but human reading the code may be confused if the operands are integers or not.  And, unless I missed something you divide by historicValue without checking it's non-zero.  I forgot if you put in a try block, nope seems not.   Just an fault/bug waiting to happen.

Up to you;  it's your code.

Don't read me wrong -- I really love this plugin and use it heavily in my server.  I've got a lot of reading to do in pricing and market dynamics to pick the right median.

I am tempted to link the results of the CoreProtect plugin (which uses MySQL so I can query it easily in the shell) with yours (hc_objects table)  so .. heh (evil grin)  I can actually determine how many blocks of a certain ORE are being mined as insider information on setting the median dynamically via cron on an hourly basis.

Keep up the great work.  It's awesome.

@RegalOwl

I'll make a ticket and give more details.

Short answer tho:

It doesn't happen every time. I'm using MySQL and I don't have an errors.log.

This also happens on build #288.

In build #284 I periodically have long delays (up to 20 minutes) right after

[HyperConomy] Internal economy unhooked from Vault.

It seems like this task is the cause of the delay:

[Databukkit[HyperConomy]] Finishing the current SQL write task... Please wait.

I have 5 different economies currently. Even if I start up and then shut down right away, this still happens.

But only sometimes.

Any idea what might be causing this?