[m-dev.] New datatype: rope

[Paul Bone]

On Thu, Nov 11, 2010 at 02:36:05PM +1100, Peter Schachte wrote:
> >> how about generalising it by allowing segments of the
> >> rope to be made of any instance of a sequence type class?  Then you could make
> >> a rope from separate bits of strings, arrays, lists, substrings, basically
> >> anything you have lying around that somehow implements a sequential collection
> >> of homogeneous things?  I'm not sure what the best set of methods for the type
> >> class are, but something equivalent to CAR, CDR, CONS and NIL.
> > 
> > This is probably okay, provided that you include elem and slice operations and
> > probably right-hand-side equivilents of CAR CDR and CONS.
> 
> What are "right-hand-side equivilents of CAR CDR and CONS"?

SNOC RAC and RDC.

Given that:
    [1, 2, 3] = cons(1, [2, 3])

SNOC is therefore:
    [1, 2, 3] = snoc([1, 2], 3)

see cord.snoc/2

> Maybe it would be worth including map, filter and fold in the type class?  Or
> maybe just mapreduce?

Probably.

Not just mapreduce, it's ill-defined, what does reduce mean?  fold left or fold
right?  Associativity is important.

> > I'd also like to see
> > more typeclasses in the standard library for collections.  Mutterings have also
> > been made about first class, and possibly parametric modules.  That is, rather
> > than use type classes use modules that can be parameterised by other modules.
> > 
> >     import cord(list) as cord_using_lists
> >     import cord(slice) as cord_using_slices
> 
> But that wouldn't achieve what I was after:  allowing mixing of various kinds
> of sequences in a single rope.

I don't think typeclasses will give you that either.

Perhaps typeclasses and existential types will.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 490 bytes
Desc: Digital signature
URL: <http://lists.mercurylang.org/archives/developers/attachments/20101111/94812406/attachment.sig>