R devel - Oct 1997 - R-alpha: Using autoload and smaller 'base' package [was "Re: New Version"].

[This is a "spill-over"  from the  "R-core"  mailing list,
	 since several of you may be interested and have comments ... -MM]
>>>>> "Thomas" == Thomas Lumley
<thomas@biostat.washington.edu> writes:
    Thomas> On 29 Oct 1997, Peter Dalgaard BSA wrote:
    >> Perhaps we should revise the current contents of the base library
    >> and replace some noncrucial items with autoloads? (The ts stuff for
    >> instance, glm, maybe even lm).

    Thomas> It would be interesting to experiment along these lines.
    Thomas> Clearly if something is going to be loaded it saves time and
    Thomas> space to do it immediately, but if it isn't going to be used
    Thomas> it's better to keep it on disk.  Personally, I use glm nearly
    Thomas> every time I use R, but other people may have different needs.

    Thomas> The problem clearly isn't urgent, since the only library
slated
    Thomas> for inclusion is ctest, which is nice and compact. However, a
    Thomas> configurable, modular approach seems like a good idea, and
    Thomas> works for emacs.  Given the desire to run R on small machines
    Thomas> we certainly can't follow the S-PLUS path of including
    Thomas> everything useful in the default set-up. For example, survival4
    Thomas> takes up a quarter of the cons cells using the default
    Thomas> settings, and nlme() (if we ever do it) will presumably be
    Thomas> fairly large as well.

NOTE: These things (libraries) are called ``packages'' from 0.60 on ..
					    --------

    Thomas> Perhaps this is more of a r-devel type comment?

Yes ! [and you see, I took the liberty ...]


Actually, I have thought along similar lines:
I think it is a neat concept to use autoload and  more than just the 
'eda'
and 'mva'  libraries. 

BTW: Is it a good idea if I include  my  
	stepfun / ecdf     functions into a new library  'stepfun' ?
     ie. a library that belongs to the distribution instead of being contrib?
	
Martin
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-r-devel
mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To:
r-devel-request@stat.math.ethz.ch
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=

Perhaps someone could bring those of us not in "core" up to date a
bit.
What are advantages of autoload and the implications, especially for
memory and speed.  Currently I have to divide my package into several
pieces in order to load it (I'm not too  sure why, but it is fairly
big). This is what I do:

R  -v6 -n250000
library("dse/tframe")
library("dse/dse1")
library("dse/dse2")
library("dse/dsex1")
library("padi/padi")
library("dse/dsex2")

Some pieces are much more important and frequently used than others.
Also, within some pieces there are functions which are used a lot more
than others. Should I consider autoloading everything? Will I still need
to divide the package into pieces? Should I avoid autoloading anything
that is used really frequently?

Paul Gilbert

=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-r-devel
mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To:
r-devel-request@stat.math.ethz.ch
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=

>>>>> "Martin" == Martin Maechler
<maechler@stat.math.ethz.ch> writes:
>>>>> "Thomas" == Thomas Lumley
<thomas@biostat.washington.edu> writes:
    Thomas> On 29 Oct 1997, Peter Dalgaard BSA wrote:
    >>> Perhaps we should revise the current contents of the base
library
    >>> and replace some noncrucial items with autoloads? (The ts stuff
for
    >>> instance, glm, maybe even lm).

    Thomas> It would be interesting to experiment along these lines.
    Thomas> Clearly if something is going to be loaded it saves time and
    Thomas> space to do it immediately, but if it isn't going to be used
    Thomas> it's better to keep it on disk.  Personally, I use glm nearly
    Thomas> every time I use R, but other people may have different needs.

    Martin> Actually, I have thought along similar lines: I think it
    Martin> is a neat concept to use autoload and more than just the
    Martin> 'eda' and 'mva' libraries.

    Martin> BTW: Is it a good idea if I include  my  
    Martin> 	stepfun / ecdf     functions into a new library 
'stepfun' ?
    Martin>      ie. a library that belongs to the distribution
    Martin>      instead of being contrib?

Can we do like emacs, and have a means for "dumping" packages into R?
Even if it is just a matter of adding or removing comments in
appropriate places, this could be useful.  Just like emacs, this would
increase the initial size, but means that little time is wasted during
load.

[I assume that the discussion is with respect to replacing packages
with autoloads, and I for one would consider doing that with some of
the packages as a "Good Thing"(tm) (since I've got to the point
that
I'd rather wait to have things grow as I need them, rather than let
them start huge and grow incrementally at a small pace).]

best,

-- 
-tony (Anthony Rossini)	           Statistics Dept., U South Carolina
rossini@stat.sc.edu	           Columbia, SC 29208
http://www.stat.sc.edu/~rossini/   803-777-3578(O) 803-777-4048 (fax)  
        "'Bother,' said Pooh, as Cthulhu rose up and ate him."
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-r-devel
mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To:
r-devel-request@stat.math.ethz.ch
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=