R help - Aug 2004 - AW: AW: built-in Sweave-like documentation in R-2.x

> > Is selecting and 'C-c C-r'-ing the 3 chunks separately that
bad?
> 
> Yes.  The UI should take care of it for him.
right.
 
> > Others may have better suggestions.
> 
> A bit more work on the chunk evaluation approach within Emacs is one;
> it almost does what is needed, but not quite.  
why almost, but not quite?

...without these "almost, but not quite" I would rather
confirm your statemnt :)

--
Valery

> An simpler alternative to code would be to allow one to
> eval-chunk-and-step, stepping through chunks, similar to C-c C-n for
> stepping through lines.  Would this solve the basic problem?  3 x (2
> or 3 keystrokes) for 3 chunks.
hm, not really ideal...

> I can't imagine an evaluation which would cross chunks but use only
> part of chunks (this suggests bad programming design to me), but
> perhaps you (or others) have an example of when this functionality
> would be useful?  (i.e. actual regions to eval which cross code-chunk
> boundaries but contain 1 or 2 incomplete code-chunks?).
hm, when?..
The very typical situation for me today is R-code 
with structure like this:

##############################
# Section 1. Loading data 
# sometime execution time is *very* long (up to 5 minuutes with read.table)
# This section have to be executed quite rarely
...

# Section 2. Some initial data preparations and preprocessing
# execution time might be long. This section have to be 
# executed a bit more often then Section 1.
...

# Section 3. Task-oriented part
# execution time might be very long,
# this section is executed very often.
...
##############################

actually the Section 3 motivated me to start this thread.
because it usually consists of many chunk alternations 
and should be reexecuted many times during development.

--
Valery