Re: debug.getlocal losing information

[HyperHacker <hyperhacker@...>]

On Wed, Jul 7, 2010 at 21:46, David Manura <dm.lua@math2.org> wrote:
> On Sat, Jul 3, 2010 at 3:30 AM, HyperHacker <hyperhacker@gmail.com> wrote:
>> I've written a function that performs variable substitution in strings
>> (e.g. replacing "$foo" with value of foo). The only way I could find
>> for it to access the local variables of its caller is to use
>> debug.getlocal...
> On Wed, Jul 7, 2010 at 9:05 AM, GrayFace <sergroj@mail.ru> wrote:
>> Why? The old behavior seems more logical. Now we would have to worry about
>> tail calls when we pass a stack level to 'error' function?
>
> I've argued in the past that the `level` parameter on `error` was
> problematic in itself [1] and that there's no really good way to do
> string interpolation [2], with the best solution possibly being Lua
> supporting some lexical or preprocessor step that transforms printf
> "${x},${y},${x}" into printf("${x},${y},${x}", x, y, x), which would
> then allow functions like `printf` to complete the implementation of
> interpolation as they see fit.
>
> The point I'm making is that if better solutions to these problems
> were available that avoid stack levels (like the recent replacement of
> setfenv/getfenv with _ENV), then the question of whether stack level
> distances are preserved may become moot.
>
> [1] http://lua-users.org/lists/lua-l/2010-01/msg01255.html
> [2] http://lua-users.org/lists/lua-l/2010-07/msg00105.html
>

If I'm understanding correctly, being able to access the _ENV of your
caller, and its caller, etc would basically give you the same thing
you can do now with a debug.getlocal loop, but more conveniently -
it'd still have the tail call issue, no?

-- 
Sent from my toaster.