Re: Suggestion for 5.3: import() for creating a "slew" of locals

[Philipp Janda <siffiejoe@...>]

Am 19.11.2013 00:00 schröbte Sir Pogsalot:
> As far as I know, assigning to the allocated locals would involve a table
> lookup *once* at runtime -- the local access would still be faster?  Maybe
> I'm wrong? :>

The assignment is not the problem. *Using* the dynamic locals is: You 
would need to change the way locals are implemented (or patch the 
bytecode at runtime).

Now:

    local tconcat = table.concat  -- use register no. 4 for tconcat
    -- ...
    -- tconcat is register no. 4, so do a getlocal(4) or getupval(4):
    print( tconcat( sometable ) )

Then:

    import( table, { "concat" }, "t" )
    -- ...
    -- need to change the getglobal("tconcat")/getlocal(x)/getupval(x)
    -- to a getlocal(y) or getupval(y) when import is finished:
    print( tconcat( t ) )


An alternative would be an `import` compiler directive, so that register 
allocation happens at compile time again. You would need a different 
syntax though, because `{ "concat" }` can't be a table constructor 
anymore (and "t" a string), because you need the list and prefix at 
compile time. And now we are back to macros/token filters ...


Philipp