R devel - Feb 2005 - formatC with illegal input crashes Rgui (PR#7686)

Full_Name: Peter Ehlers
Version: rw2001pat (2005-02-03)
OS: Win XP
Submission from: (NULL) (136.159.61.115)


formatC(1, flag="s") crashes Rgui.
Similarly for flag=[SnZ].
Stupid input, of course, but I'm error-prone.

Peter

We'll fix this, but you might like to note that sprintf is more
bulletproof and as from 2.1.0-to-be, a lot more tolerant, e.g.

sprintf("%s", 1)

works there and in 2.0.1 gives a useful error message.

On Fri, 11 Feb 2005 ehlers@math.ucalgary.ca wrote:
> Full_Name: Peter Ehlers
> Version: rw2001pat (2005-02-03)
> OS: Win XP
> Submission from: (NULL) (136.159.61.115)
>
>
> formatC(1, flag="s") crashes Rgui.
> Similarly for flag=[SnZ].
> Stupid input, of course, but I'm error-prone.
-- 
Brian D. Ripley,                  ripley@stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595

Prof Brian Ripley wrote:
> On Mon, 14 Feb 2005, Wolfgang Huber wrote:
> 
>> Dear Prof. Ripley,
>>
>> Would it be possible to make sprintf accept vector arguments?
>> I.e. allow expressions like
>>  sprintf("%04d", 1:3)
>> to produce something like what currently needs to be done via
>>  sapply(1:3, function(i) sprintf("%04d",i))
> 
> 
> I have thought about this, but it is not really clear what the right 
> thing would be here with multiple arguments.  What should
> 
> sprintf(("%04d %s", 1:3, "abc")
> 
> do?  Produce a character vector of length 1 or 3?  If of length 1,
> "0001 0002 0003 abc" or "0001 2 0003 abc"?
> 
> I guess the most R-like thing would be to recycle args to the length of 
> the longest and then use them in parallel, but that can be done fairly 
> easily by *apply.
Personally, I would prefer the recycling, but of course it can be also 
done this way:

 > mapply(sprintf, "%04d %s", 1:3, "abc")
    %04d %s       <NA>       <NA>
"0001 abc" "0002 abc" "0003 abc"

the only slightly unaesthetic thing being the names of the resulting vector.

-- 
Best regards
   Wolfgang

-------------------------------------
Wolfgang Huber
European Bioinformatics Institute
European Molecular Biology Laboratory
Cambridge CB10 1SD
England
Phone: +44 1223 494642
Fax:   +44 1223 494486
Http:  www.ebi.ac.uk/huber

sprintf() in R-devel is now vectorized, re-cycling its arguments, 
including fmt, as required.

On Tue, 15 Feb 2005, Wolfgang Huber wrote:
> Hi Peter,
>
> thanks. Yet my intention was not to discuss whether this could be done at
all
> (never had any doubts about that), but how it could be done nicely and 
> conveniently for the application programmer.
>
>  Best regards
>  Wolfgang
>
> Peter Dalgaard wrote:
>> Wolfgang Huber <huber@ebi.ac.uk> writes:
>> 
>> 
>>> Personally, I would prefer the recycling, but of course it can be
also
>>> done this way:
>>> 
>>> > mapply(sprintf, "%04d %s", 1:3, "abc")
>>>    %04d %s       <NA>       <NA>
>>> "0001 abc" "0002 abc" "0003 abc"
>>> 
>>> the only slightly unaesthetic thing being the names of the
resulting
>>> vector.
>> 
>> 
>> ...which is of course fixable with either of
>> 
>> 
>> 
>>> mapply(sprintf, MoreArgs=list(fmt="%04d %s"), 1:3,
"abc")
>> 
>> [1] "0001 abc" "0002 abc" "0003 abc"
>> 
>>> mapply(sprintf, 1:3, "abc", fmt="%04d %s")
>> 
>> [1] "0001 abc" "0002 abc" "0003 abc"
>> 
>>> mapply(sprintf, "%04d %s", 1:3, "abc",
USE.NAMES=FALSE)
>> 
>> [1] "0001 abc" "0002 abc" "0003 abc"
>> 
>> (Only the last one is completely failsafe since the first two relies
>> on 1:3 not being character:
>> 
>>     if (USE.NAMES && length(dots) &&
is.character(dots[[1]]) &&
>>         is.null(names(answer)))
>>         names(answer) <- dots[[1]]
>> )
>
>
-- 
Brian D. Ripley,                  ripley@stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595