[R] Does R have a "const object"?

[luke-tierney at uiowa.edu]

On Wed, 16 Mar 2011, Gabor Grothendieck wrote:

> On Wed, Mar 16, 2011 at 12:16 PM,  <luke-tierney at uiowa.edu> wrote:
>> On Wed, 16 Mar 2011, Gabor Grothendieck wrote:
>>
>>> On Wed, Mar 16, 2011 at 11:49 AM,  <luke-tierney at uiowa.edu> wrote:
>>>>
>>>> Just as a heads-up: it is likely that unlocking the bindings in base
>>>> for pi, T, F, probably all BULTIN and SPECIAL functions, and possibly
>>>> more, will start signaling warnings in the near future.  Doing this
>>>> may be useful at times for debugging but it can mess up assumptions
>>>> others make about how things in base work and so reduce code
>>>> reliability.
>>>
>>> That seems ok for pi, T and F but if its extended to everything in
>>> base then I would hope there is a nowarn= argument or other easy way
>>> to avoid the warning message.
>>>
>>
>> That would defeat the purpose.  Unlocking things in base may be useful
>> for experimenting or debugging but it is not a good idea otherwise.
>> [? assignInNamespace could be more explicit on htis and will be soon.]
>> There is a reason we lock bindings in the first place, and that is so
>> one can assume that these bindings have certain values and certain
>> properties and one can write reliable programs against these
>> assumptions.
>>
>
> Its useful for being able to set defaults for arguments that do not
> have defaults.  That cannot break existing programs.

Until the next program decides do co change those defaults and either
can't or does and you end up with incompatible assumptions.  It also
make the code with the added defaults inconsistent with the
documentation though, which is not a good idea.  It may seem
convenient but it isn't a good idea in production code that is
intended to play well with other production code.

> Note that if this feature is implemented in a heavy handed manner it
> could cause havoc as at least one package that is depended upon by
> literally dozens of other packages (and possibly hundreds if one takes
> into account dependencies of dependencies) cannot function.

The reason I sent my initial message is so those who do this sort of
thing can start thinking about other approaches.  I do not expect to
need this change for closures any time soon, but will need it for
constants and primitives.  It may be useful for some closures as well,
so that change may come farther down the line.

luke

-- 
Luke Tierney
Statistics and Actuarial Science
Ralph E. Wareham Professor of Mathematical Sciences
University of Iowa                  Phone:             319-335-3386
Department of Statistics and        Fax:               319-335-3017
    Actuarial Science
241 Schaeffer Hall                  email:      luke at stat.uiowa.edu
Iowa City, IA 52242                 WWW:  http://www.stat.uiowa.edu