Re: Raising awareness about guile-pfds status

[Linus Björnstam]

If it is HAMTs or persistent vectors you want, I have a git repo of Andy's Fash 
and Fector (functional hashmaps and functional.vectors). Fash lacks some parts 
to become a fast implementation of (srfi 146 hash), like being able to properly 
delete keys. I would implement it myself if I had any friggin idea what the 
code did, but to my limited mind it is quite impenetrable.

Other than that, I have a decently fast implementation of a pairing heap in my 
Nietzsche repo, which I found is generally slightly faster than the height 
balanced leftist tree in the PFDS library (and, should be said, about an order 
of magnitude slower than a regular binary heap using vectors).

The Fector library: https://bitbucket.org/bjoli/guile-fector

The fash library: https://bitbucket.org/bjoli/guile-fash/

My pairing heap: 
https://bitbucket.org/bjoli/nietzsche/src/default/data/pairing-heap.scm

I don't know how using records compares to using cons pairs in guile, but there 
might be some speed gains. It also lacks a proper (as in fast) list->node 
because my 23 year old me didn't know how to implement it. It should be 
trivial. 

Fash and Fector are very fast. Andy likes speed :)


-- 
  Linus Björnstam

On Mon, 15 Jul 2019, at 18:06, Amirouche Boubekki wrote:
> The original maintainer of guile-pfds is sadly not responding to my mails.
> 
> Right now, guile-pfds is the goto solution for functional data structures
> in Guile,
> and prolly other Scheme implementations.
> 
> At least the hamt has a bug, see https://github.com/ijp/pfds/pull/6/files
> 
> I think this is a very important package especially for guile that doesn't
> have a functional hashmaps.
> 
> Anyone willing to take ownership of the project?
> 
> 
> ref: https://debbugs.gnu.org/cgi/bugreport.cgi?bug=35518
> 
> -- 
> Amirouche ~ amz3 ~ https://hyper.dev
>