> > I have made a lib containing some external iterators,
see:
> > http://aeditor.rubyforge.org/iterator/
> >
> > download it here:
> >
http://rubyforge.org/frs/download.php/704/iterator-0.8.zip
>
> Thanks Simon. I just looked through what you have. You
gave
> me a few ideas. But, here are a few things I don't like:
>
> * for efficiency I would like the ability to read/write
> multiple elements at a time. I want the API to have a
superset
> of the features in IO if possible.True, I didn't made mine for speed.
> * instead of calling has_next? I'm just having next return
nil
> or an empty string/array (for getting N elements). If the
> collection can have nil elements, you'll have to ask for N
> elements (could be 1) so that you get an empty array at the
> end.What if you are iterating through an array that
coincidentially
contains a nil.. like this [42, nil, 42] ?
For collections that can contain nil, you'll have to pass the
optional length argument (=1 if you want 1 element). So in the
above, consecutive calls to get(1) will return [42], [nil],
[42], and (or nil - haven't decided). Calls to get() will
return 42, nil (you'll think your're at the end), 42, nil. So
get(length) is like read(length) (returns a String) and get()
is like getc() (returns a character/Fixnum).
> * no standard insert/delete operations. I don't need them
now,
> but they might be useful to others.Yeah, I don't have neither insert nor delete.
I didn't needed it when I made the lib. I though about it,
but couldn't decide how to approach this problem.
So that the number of methods doesn't blow up, I decided to
have my "get" and "put" methods take a flags argument to
provide many functions:
forward/reverse
move/hold
read(return original value)/ignore(return length)
scan or replace vs. delete or insert
Last I counted, I have 12 get functions and 10 put functions
(all of which can operator on individual elements or a
string/array of them). The base class will base all of these
on 4 primitives (getdelete before/after and insert
before/after), but most classes will want to override most of
it.
> * like the C++ iterator, yours standardizes on
> allowing/creating/comparing multiple iterators on the same
data
> structure. I don't want to do this because many sequential
> data structures can't handle this. The primary example
would
> be an IO (especially a non-file) which only allows you to
> read/write at one location at a time. Another would be a
text
> editor buffer where you can only insert/delete at the
cursor
> position. Instead of allowing multiple iterators for a
data
> structure, you'll be able to get and set the position where
> "position" is an object that you might be able to compare
> and/or do arithmetic on (depends on the data structure).This sounds interesting. I would be interesting at looking
at how you would approach this.
The best analogy to this is simply the IO tell/pos and
seek/pos= methods. For this type of position (an integer),
you can compare and add/subtract, but other position objects
may not be so flexible. I don't think I'll specify what these
position objects look like. Some classes where the underlying
data-structure is unidirectional (i.e. IO pipes) will use this
to signal when to start/stop buffering so that you can go back
to the original position.
> I think I'll call the class I'm doing "Cursor" because it
most
> resembles what you can do at the cursor in a text editor
and
> I've also seen the term "cursor" used for external
iterator.
>
> I might take some of the ideas in your Iterator classes
also
> (some from C++ STL iterator it looks like), if you don't
mind.Feel free to use all of it as you like.
Is your project on rubyforge ?
not yet.
···
--- Simon Strandgaard <neoneye@gmail.com> wrote:
On 4/28/05, Eric Mahurin <eric_mahurin@yahoo.com> wrote:
> --- Simon Strandgaard <neoneye@gmail.com> wrote:
__________________________________________________
Do You Yahoo!?
Tired of spam? Yahoo! Mail has the best spam protection around
