Add support for Lua or AngelScript scripting.

[juj]

Due to dozens of problems and missing features related to QtScript correctness and performance, add support for creating in-world scripts with either Lua or AngelScript. (preferably both [tell me which is better afterwards {I am serious}]).

[antont]

i made the boilerplate for a Lua scripting module using QtLua a while ago, is uncommitted on my laptop at home - i could check it in the evening and push somewhere. qtlua seemed pleasant to work with so far at least, i've no idea how it performs in the c++ <-> lua conversions

for performance and other things can be a good idea to test the PythonQt thing we already have in and working, it at least reuses the qobject instances cleverly afaik

[juj]

For the remainder of this issue item, you may post all odd snippets which crash inside Qt, or silently fail, or similar. (you may also create a separate issue for them)

Here is one:

var v = float3.zero; v.x = 1; print(v); // Prints (0,0,0) print(float3.zero); // Prints (0,0,0)

Above v is initialized to be a reference of float3.zero. Because float3.zero is read-only (you can't make the zero vector non-zero), assigning to v is not allowed. However, QtScript will silently ignore the statement 'v.x = 1;' without issuing an error (which I think is a bug in QtScript itself).

To allow assigning v.x = 1;, the constant float3.zero should be made to behave like a value type, but it is not possible with QtScript (or JavaScript) (afaik).

[erno]

Pypy claims to have a safe Python sandbox: http://doc.pypy.org/en/latest/sandbox.html

[juj]

There is also an internal QtScript bug related to printing variables, which can crash the server, but I will not disclose that until the bug has been fixed inside Qt.

[juj]

Using Python for inworld scripts would require a) The ability to create separate script contexts per EC_Script component. b) Implement the support to use the script class mechanism for Python as well. c) Safe automatic teardown of script contexts. (can use a OnScriptDestroyed mechanism for those features that require C-like free()s) d) Implement several bindings to Python (i.e. issues #34, #35, #36, and related ones)

[antont]

Making static QtScript bindings for parts of Tundra might be a good thing -- and the possible performance improvement of that is easy to test, posted details of that and some other ideas in http://groups.google.com/group/realxtend-dev/browse_thread/thread/380c061b2e3b0042

[juj]

There now exist a bindings generator and an interop mechanism for MathGeoLib to Angelscript, available for testing here: https://dl.dropbox.com/u/40949268/emcc/MathGeoLibTestAS.html This code runs Angelscript in a web browser and contains the MathGeoLib library to play with.

[juj]

There is a work-in-progress integration branch of angelscript to Tundra at https://github.com/LudoCraft/Tundra/tree/angelscript

[jonnenauha]

Nice. This is quite welcome if gives perf gains. If Angescript can be ran on browser clients (like .js can) too thats a bonus.