Re: opinion poll on text widget iterators

[Guillaume Laurent <glaurent worldnet fr>]

Havoc Pennington <hp@redhat.com> writes:

> Right now I have an opaque reference-counted iterator object,
> FrooTkxtIndex. The annoying thing about this object is that 
> it sucks as an iterator.

Another annoying thing is the type name. :-)

> FrooTkxtIndex* iter;
> iter = froo_tkxt_buffer_get_char_index(buffer, 0);
> while (iter != NULL)
> {
>   /* do stuff with iter */
> 
>   FrooTkxtIndex* freeme = iter;
>   iter = froo_tkxt_index_next_char(iter);
>   froo_tkxt_index_unref(freeme);
> }
> 
> So, yuck.
> 
> Proposed alternative is:
> 
> FrooTkxtIndex iter;
> froo_tkxt_buffer_get_char_index(buffer, 0, &iter);
> while (froo_tkxt_index_next_char(&iter))
> {
>   /* do stuff with iter */
> }

Actually, for C++, the first form is much preferable. An STL iterator
is usually used as follows :

for(iter = begin(); iterm != end(); iter++) 
{
  /* do stuff with iter */
}

So we need to have seperate operations for moving it forward, and
checking if we've reached end or not. The unref part is in
iter::operator++(), so no problem if it takes 3 lines, the programmer
won't see it.

> Misgiving #2: these index objects are 90% of the time not actually
>  used for iteration; it will be both easier and more efficient to use
>  _foreach() API's that are also provided for iterating over the
>  buffer.

Not in C++, we already have for_each(), so having one to wrap looks
quite awkward. I wonder how Python or Perl manage this - my guess is
that the problem is similar.

What we need for an iterator is :

- a way to move it forward
- a way to test if it has reached past the last element
- a way to dereference it

All these operations have to be O(1) for an iterator to be
useable. Then depending on the feasability, goodies include :

- a way to move it backward
- a way to set it to a random position

> "Dereferencing" a FrooTkxtIndex is O(n) where n is the length
>  of the line containing the index.

We've got a problem here... Why is that ?

> So misgiving #2 is that FrooTkxtIndex isn't actually inconvenient in
> the common use case.

But these "common use cases" are actually quite uncommon if you
consider how an iterator is used in general.

-- 
					Guillaume.
					http://www.telegraph-road.org