XLua, MMF2 Extension

I'm exploring making this object global by nature, with the ability to create different independent lua states in numbered slots. Not only would it be really convenient to access a common lua state in another frame, but it might mitigate some of the problems I've been trying to work out for MB with resetting state.

New major XLua release: http://hocuspocus.taloncrossing.com/rii/xlua-global.zip

I've rewritten significant portions of this object to "globalize" it. The lua object itself is now little more than a view, which you can bind to one of many possible independent states. Once bound, any operation is performed on that state. You can also unbind, or rebind the object to a new state. You can also bind multiple XLua objects to the same state, and they will all respond to events.

This object is fully backwards compatible, but you must bind it to a state before you start using it (you don't need to explicitly create the state first, binding to an empty state will create it). If you don't do this, your application will act like there is no XLua object at all -- any operation on an unbound XLua object is essentially ignored.

If you delete a state, any object previously bound to it will be unbound. You can now reset state by deleting the lua state, then rebinding the object to it (thus recreating the state).

Although I haven't really tested this yet, you should be able to access a common lua state from any frame, by binding an XLua object to the same state. As long as it's already been created, it won't be recreated. Delete the state at the end of the frame if you don't want it available elsewhere.

so basically

unbind state (<- source of crash)
detroy state 0 (<- also a source of crash)

bind state 0 (<- doesn't crash anything)


??

Because when I do that, it crashes.

This post has been edited by MechaBowser on Mar 21 2009, 06:42 AM

So just to elaborate on this, your application will crash if you try to destroy (or perhaps unbind) a state in an On Function event, if that function was triggered by the same state you're trying to destroy. Control from an On Function event always returns to lua execution, so if the state doesn't exist anymore, bad things happen.

http://hocuspocus.taloncrossing.com/rii/xlua-global.zip

Small update to the MMF Interface:

mmf.Keyboard.KeyDown(key)
mmf.Keyboard.KeyUp(key)

These events test current state, if you want to capture the keyup / keydown events themselves, use the WIN Interface.

The key parameter is the numerical value of Microsoft's Virtual-Key Codes, documented here: http://msdn.microsoft.com/en-us/library/dd375731(VS.85).aspx

I recommend creating a table of the keys you care about for convenience and readability.

The documentation is still a bit out of date and doesn't include these functions.

time to completely rewrite my keyboard code >:(

pretty soon you're going to pop out sound/music functions/variables, aren't you

having a load of issues

1) table values and entries aren't reset
2) if alot of functions are callen at once, some of them are politely skipped without reporting any errors

Chances are I won't catch you on MSN for a couple days so could you explain a little better these two issues? What is "a lot" of functions, and which functions are we talking about.

Well I've got like 5 particles that spawn at once. They're added to a table and then looped through. When their lifetime variable reaches a certain value they're destroyed. As they're all created at the same time, they're destroyed at the same time, too. However, even though they're all removed from the objects table, sometimes one of them aren't destroyed (which is managed by calling a function to MMF2).

I'd really need to see a code sample, including relevant events, for both creating and destroying these objects. How are you handling object destruction? Fixed ID?

If there were problems with calling lots of events at once, every time I pained a bunch of tiles in a level, they'd be mottled with "holes". But that doesn't happen.

Odd.

Yes I'm using fixed values as identifiers for the objects. It's wierd because if I print at the destroy function inside Lua, and add a counter in MMF2 for every destroy function called, they appear different.


Also, I'm having another wierd issue. I have this raycasting going where I check for objects between two points. It's all handled inside MMF2. However, when I tried to do a raycasting, it didn't work. I tried adding a print function in front of it and it suddenly worked. I then tried to comment out the print and it stopped working again. It seems as if some values are 'broken' unless printed. I also have this issue with my require function that I told you about earlier, so it's not a single-case issue.

Small bugfix: http://hocuspocus.taloncrossing.com/rii/xlua-global.zip

mmf.Object.SetLayer only changed the layer of the object's sprite, but not the object itself. This interfered with pasting routines, and probably other stuff as well.

Documentation has also been updated for prior changes.

http://hocuspocus.taloncrossing.com/rii/xlua-global.zip Another update

This fixes one of Mecha's bugs, which as a stack bug caused by invoking an MMF Function from a nested "Load String" action. Up until now, "Load File" and "Load String" improperly bypassed the call stack.

There is also a new action, "Invoke MMF Function", which allows calling one of your "On Function" events directly, and works with parameters and return values. It is morally equivilent to a lua function that just calls another MMF Function, and saves a few CPU cycles.

Would it be possible to get what line a "on function" is executed at? That would be really useful both for bug-tracking and error reports.

Also a "push Nil return value" would be neat


edit: after moving away from all dostring activities possible; I am still having troubles with tables not being erased when they should.

This post has been edited by MechaBowser on Mar 29 2009, 05:41 AM

hey RII

I don't think xlua likes it when I call a function 300 times at once

I also think that this might be the reason why it've had those errors in the past considering it returns the same error

but yeah, MSN and I can show you proof!

edit: I must warn you, it involves dostring

This post has been edited by MechaBowser on Apr 5 2009, 12:49 PM

I did a few tests on my issues with missing objects.

1) First, I did a test to see if any bugs occoured when not using dostring. I stopped the load map process and instead just plain restarted the level. It worked fine.

2) Second, I took all the lua scripts my map would call and dropped into a lua file. It resulted in a good 320 line lua file, consisting mainly of function calls. I then tried to require the file in my level and not load a map at all, and the same problem occoured as when using dostring. However, I could see that after 220 function calls, the +1/-1 fixed value pattern issue appeared. Everything was fine beforehand. Function # 220 was nothing out of the ordinary, nor were any functions before or after.

The issue could be that I call too many functions at once. All these functions have to communicate with MMF2 one way or another, so I'm probably killing the xlua object somehow. What does bother me, though, is that it doesn't happen before the 220th function.

UPDATE

3) I tried to call every function slowly with a 5 second break inbetween. No difference. Appearently it isn't because I'm calling these functions rapidly, but the issue seem to appear because I call that many by number.

This post has been edited by MechaBowser on Apr 7 2009, 09:21 AM

Yes~!

I have successfully isolated the bug.

RII COME AND GET IT I SAY
I'LL BE WAITING

http://hocuspocus.taloncrossing.com/rii/xlua-global.zip Important bug fix

Pushing or Getting large integers (> ~16.75 million) failed to work properly, because they were being passed by MMF as single-precision floating point. All numeric handling has been switched to integer. This means if you were passing an actual floating point value in either direction, that will no longer work. Push it as a string instead, since MMF and Lua will both handle the conversion with minimal effort or transparently.

A "Push NIL" action has also been added both for MMF and Lua functions.

guaranteed i'll forget about this and make angry posts later when i find i can't push decimals

EDIT: wait how would this work with interface functions that are supposed to take numbers, or does this not apply to interface functions

This post has been edited by Xgoff on Apr 7 2009, 10:54 PM

interface functions don't care. Pass what you would expect.