buffer segments, uniform and automatic (was: Re: [Elisp: 8 out of 10 problems] I think the last one! (point))

[Emanuel Berg <incal@dataswamp.org>]

> Okay, I can make out 'begin' and 'length'. Without looking,
> I do not understand at all what 'reg' is supposed to mean.
> region? regexp? register?

I came up with an idea how to do this!

It is based on this policy:

1. Everything has a beginning and an end.

2. Everything belongs to something else.

3. The only standard worth its name isn't the one that comes
   from ISO, or the one in a manual, it isn't even the one the
   best programmers (almost) always stick to - it is the one
   created by a computer program.

So as for the segments in a text mode buffer, as for
point (2), it would look something like:

  [ buffer [ paragraphs [ sentences [words] ] ] ]

Definitions of each has to be setup manually and individually,
as for (1), so for example for word it could be

(list (progn (forward-word) (forward-word -1) (point))
      (progn (forward-word) (point)))

Then the program will define a set of functions for each
segment type, the same for each.

Note that so many common operations can be defined from the
definition, i.e. the (BEG END) pair:

-len is (- END BEG)
-reg is (BEG END)
-beg is BEG
-end is END
-cur is (buffer-substring-no-properties BEG END)
...

Everything else is then based on combinations of those:

-prev of -beg and -cur
-next of -end and -cur
...

And it goes on:

-all  of -cur and -next     (for some outer BEG END)
-count of `length' and -all (for some outer BEG and END)

I'm working on it now!

PS. Change of subject line as I think the previous one with
    [square brackets] hindered the thread from being
    archived online.

-- 
underground experts united
https://dataswamp.org/~incal