Re: linking tables to C++ objects?

[Sam Roberts <sroberts@...>]

On Mon, Feb 26, 2007 at 01:08:40PM -0800, Graham Wakefield wrote:
> Ah, now I read Rici's note.
> 
> So in pseudocode again:
> 
> // creation (synchronized) of table, userdata & c++ instance:
> 
> ptr = [C++ instance]
> udata = userdata[ptr]	// boxed pointer
> table = lua_newtable
> setfenv(udata, table)
> REGISTRY[weakvaluedtable][ptr] = udata

You can't literally use a ptr, you must use a lightuserdata as the key.

> REGISTRY[weakvaluedtable][table] = ptr

^-- probably unnecessary, see below

> // from C++ ptr to table:
> 
> getfenv(REGISTRY[weakvaluedtable][ptr])
> 
> // from table to C++ ptr:
> 
> REGISTRY[weakvaluedtable][table]

Why would you would you want to map the table to the ptr?

lua code will have only a udata, and will pass a udata to C code, your C
code will call lua_touserdata() to get the ptr. No table required.


> // garbage collection:
> 
> udata.__gc { // remove ptr & table entries from REGISTRY 
> [weakvaluedtable]) }

No, the only thing you need to do in your __gc is free any non-lua
resources.

This could be as easy delete on the C++ object boxed in the udata.

__gc isn't called until there are no longer ANY references to the udata,
INCLUDING the weak-valued table. Note that this allows your C code an
easy way to determine if lua has a udata that boxes some ptr -> look up
REGISTRY[weakvaluedtable][lightuserdata(ptr)], if its nil, the udata
never existed, or has been GCed.

> Surely there is a more efficient way to do this!?

Don't create table entries you don't need.


You need to go from ptr -> lightuserdata(ptr) -> udata, thats what the
table is for.

You want the udata to be garbage collected, thats why the table has weak
values, and you implement the __gc metamethod.

You need to go from udata -> ptr, thats what lua_touserdata() is for.

Optional: if you need to go from udata -> table, maybe to store some
number of lua values with your udata, thats what lua_getfenv() is for.



Sam