FEEDBACK WANTED: Type/class unification

[Markus Schaber]

Hi,

Jeff Shannon <jeff at ccvcorp.com> schrub:

>> With the automatic turning to 0/maxint, they even are limited to
>> non-negative integers. So it is impossible to use any form of
>> negative indexes, which sometimes would be nice when dealing with
>> coordinates.
> 
> Um.... you *can* use negative values in slices:
> 
>>>> mylist = [1, 2, 3, 4]
>>>> mylist[-2:-1]
> [3]
>>>> mylist [:-2]
> [1, 2]

Yes - but this means "count from the other end" instead of "go 
backwards from index 0"
 
> What you can't have, is negative indices into a list.  

And this is what I sometimes would like to have.

> Neither of
> these, however, have anything to do with the specific case being
> discussed, which is the default value used for unspecified slice
> endpoints.

Thats correct, there I wrote something different to what I thought of.

My complain was meant about the fact, that the conversion is done 
before my method is called, instead of leavint this decision to my own 
implementation.

> A list of lists is really not the same thing as a 2d-array, and is
> probably
> not suitable for use as a coordinate system.  If you really *must* do
> so, though, and your coordinates *must* include negative values, then
> you could wrap the list of lists into a class and provide an automatic
> translation.

At least when dealing with something like a "Fungespace" (i. E. you 
have small integers as indices in an array, but the offsets can get 
negative) it would be useful.

I always have to use an own class for it (as there's no builtin list 
with negative indices), but I can't use slicing syntax with this own 
class as the negative values are automatically converted to end-value 
before calling my method.

There may be other scenarios when this conversion would be useful - 
imagine a double linked ring list. Then small negative values could 
mean "come from the end" which is much faster then coming from the 
start when the list is long.

markus
-- 
Defend democrathy and freedom!