 |
lua-l archive |
|
|
||
|
Thanks for the reply Tomas. Tomas wrote:
> So we end up with an environment that has lots of userdatas that
> get used for subsequent Lua code.
This environment is reused by different executions of different
scripts?
Yes.
> if the code that I am executing does a require"somemodule" then
> the environment of the functions in the module are set at require()
> time, that is when the code is loaded. I have no obvious way
> of resetting the environment of the required()ed functions as my
> transient environment changes.
You are reusing the functions of the module but changing its
environment?
Yes, exactly.
> One option would be to use (in the require()ed code something like
> setting the environment to have __index and __newindex to
> functions that somehow dynamically find the environment of the
> caller.
>
> setfenv(1, setmetatable(env, {_index = function_that_finds_env()}))
This seems to be very difficult to implement, isn't it?
If I understand it correctly, this problem seems to be similar
to the one that we faced with VEnv in Lua 5.0. We could redefine module()
to use with Lua modules, but we couldn't redefine the behavior used by C
modules so we have an environment that doesn't work with required binary
libraries.
Your problem seems worse than mine.