Re: Should let symbols be interned?

[Lennart Borgman (gmail)]

Juanma Barranquero wrote:
> On 1/22/07, Lennart Borgman (gmail) <address@hidden> wrote:
>
> > Thanks, but I did not mean on this level. On this level I would expect
> > it to be transparent to the user/programmer.
>
> Please, explain that.
>
>                    /L/e/k/t/u


I tried to, but it seems like I am not clear enough. I know very little 
about the internals of Emacs, how the big obarray 'obarray is handled 
etc. Maybe it is hard to understand me because of this.

I believe however that it takes more time to insert a symbol in big 
obarray than in a small. But that depends on how the internal structure 
of an obarray looks and how it is handled. I assume it is some kind of 
tree structure.

'obarray is big. For 'obarray I would expect the most common access to 
be a search. Then the tree probably should be handled to optimize that. 
Balancing the tree is however a time costly operation (or maybe not?). 
Then perhaps it would make sense to have a small obarray for let 
variables. A bigger part of the access to this would be insert. Because 
it is small I believe insert can be handled in a shorter time than for 
'obarray.

If this would outweigh the cost of doing a search in two obarrays for 
every variable symbol lookup is of course a question which can not be 
answered without knowledge of both the algorithms and typical use cases.

And thinking about it it looks to me that avoiding balancing the let 
symbols could achieve the same benefits without the cost of two lookups.

So this is all some rather dim thoughts about obarray lookup. Nothing 
very interesting if you do not know this part of Emacs and I don't.