Re: Function environment curiosity

[Jerome Vuarand <jerome.vuarand@...>]

2009/6/3 Rob Hoelz <rob@hoelzro.net>:
> I thought I understood function environments, but I wrote a small piece
> of code that has me guessing.  Here's the code:
>
> local p = print
>
> setfenv(0, setmetatable({}, {__index = _G, __newindex = function(_, k,
> v) p 'uh-oh'
> end}))
>
> print 'still here'
> print = nil
> io.write('print: ' .. tostring(print) .. '\n')
>
> I thought that this would print the following when run:
>
> still here
> uh-oh
> print: nil
>
> but instead, it printed:
>
> still here
> print: nil
>
> Also of note is that if I put this in a file and require that file,
> global assignments cause 'uh-oh' to be printed, as I intended.  Also,
> getfenv(0).print = nil triggers __newindex, as expected.
>
> I thought that GETGLOBAL/SETGLOBAL was equivalent to operating on
> getfenv(0); I checked the output of luac -l and it is using
> SETGLOBAL.  What do I have wrong?

This is because the newindex metamethod is not called if the index
already exists in the table.