Re: static vs. dynamic scoping

[Jilles Tjoelker <jilles@stack.nl>]

On Wed, Nov 10, 2010 at 03:15:59PM +0100, ольга крыжановская wrote:
> Eric Blake wrote:
> > Or should dash switch entirely to static scoping (my gut feel is that
> > static scoping may be more efficient to implement, which fits in line
> > with dash's desire to be as lean as possible)?

> static scoping is much more efficient. Think about accessing a global
> variable in a deep function call stack, e.g. 20 functions deep. With
> dynamic scoping you have to look up the list of local variables for
> each function *first* before looking at the global variables. In my
> example this means 20 look ups. Each further function adds another
> look up.
> For static scoping you just look at the local variables of the
> *current* function and then look up the global ones. This is faster
> and much more efficient.

That's not how dash implements it, though. The local builtin saves the
value and attributes of the variable and they are restored upon return.
There is just one table of variables to search (two in older versions
and other ash versions, the additional one storing variable assignments
for regular builtins like IFS= read x; these were converted to the local
mechanism recently).

The real efficiency benefit of static scoping comes when disallowing
access to locals when the name is not known at compile time: at run
time, the names of the locals are not necessary.

-- 
Jilles Tjoelker