Re: [TCLCORE] Unifying list and dict types

[Twylite <tw...@cr...>]

Hi all,
> The more I look at this, the more I'm concerned about the sheer
> complexity of what you propose. Building a new complex data structure is
> a non-trivial task and demonstrating that it is genuinely cheaper than
> existing ones (which are probably about as simple as they can be;
> there's very little waste) is quite hard. You've got some way to go to
> win me over as yet. (Call me conservative if you want.)
>
> Have you considered whether a different hash function would be better?
> (This is independent of the memory management details that you've been
> focussing on.) There's a long-standing issue with the fact that Tcl's
> hash functions are actually very weak; much better ones are available
> these days. Some performance data would be interesting.
I'm just wondering if anyone has thought to take a look at what Lua does 
with its "tables", since they are central to the language.  LuaJIT is 
currently leading the performance benchmarks for interpreted languages.  
And while I'm talking about it, has anyone considered the possibility of 
targeting LuaJIT for Tcl9?

Twylite