R help - Dec 2008 - Finding the first value without warning in a loop

Dear R useRs,

with the following piece of code i try to find the first value which can 
be calculated without warnings

`test` <- function(a)
{
  repeat
  {
    ## hide warnings
    suppressWarnings(log(a))

    if (exists("last.warning", envir = .GlobalEnv))
    {
      a <- a + 0.1
      ## clear existing warnings
      rm("last.warning", envir = .GlobalEnv)
    }
 
  if(a > 5 || !exists("last.warning", envir = .GlobalEnv))
    break
  }

  return(a)
}

if i run this with test(-3), i would expect a=0 as return value.
Is it also possible to hide warnings during my function, i guess i use
suppressWarnings in a wrong way here?

Thanks and best regards

Andreas

Have you considered instead sending the warnings to someplace where it  
won't bother the user? If so then sink() might be useful.

My reading of the help page suggests that your call to suppressWarnigs  
is storing any results outside the Global environment. This is what I  
came up with. It may not be as general as you may want, since it only  
detects one kind of warned event, a NaN result.

testit <- function(a)
{ options(warn=-1)
repeat
    {z=log(a)
     a=a+ 0.1
     if(!is.nan(z) || a > 5 )
           {options(warn=0);   return(a); break }
}   }

 > testit(-3)
[1] 0.1

-- 
David Winsemius

On Dec 7, 2008, at 2:38 AM, Andreas Wittmann wrote:
> Dear R useRs,
>
> with the following piece of code i try to find the first value which  
> can be calculated without warnings
>
> `test` <- function(a)
> {
> repeat
> {
>   ## hide warnings
>   suppressWarnings(log(a))
>
>   if (exists("last.warning", envir = .GlobalEnv))
>   {
>     a <- a + 0.1
>     ## clear existing warnings
>     rm("last.warning", envir = .GlobalEnv)
>   }
> if(a > 5 || !exists("last.warning", envir = .GlobalEnv))
>   break
> }
>
> return(a)
> }
>
> if i run this with test(-3), i would expect a=0 as return value.
> Is it also possible to hide warnings during my function, i guess i use
> suppressWarnings in a wrong way here?
>
> Thanks and best regards
>
> Andreas
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.

Try this:

for (x in c(-2, 2, 4)) {
	tryCatch({
		y <- log(x)
		xx <- x
		break
	}, warning = function(w) {})
}
print(xx) # 2



On Sun, Dec 7, 2008 at 2:38 AM, Andreas Wittmann
<andreas_wittmann at gmx.de> wrote:
> Dear R useRs,
>
> with the following piece of code i try to find the first value which can be
> calculated without warnings
>
> `test` <- function(a)
> {
>  repeat
>  {
>   ## hide warnings
>   suppressWarnings(log(a))
>
>   if (exists("last.warning", envir = .GlobalEnv))
>   {
>     a <- a + 0.1
>     ## clear existing warnings
>     rm("last.warning", envir = .GlobalEnv)
>   }
>
>  if(a > 5 || !exists("last.warning", envir = .GlobalEnv))
>   break
>  }
>
>  return(a)
> }
>
> if i run this with test(-3), i would expect a=0 as return value.
> Is it also possible to hide warnings during my function, i guess i use
> suppressWarnings in a wrong way here?
>
> Thanks and best regards
>
> Andreas
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>

Thanks for that. I am having some difficulty figuring out how it  
works. The help page tells me (I think) that "y<-log(x)" gets
matched
to expr. Am I correct in thinking that if either an error or warning  
is produced, that tryCatch silently returns <something>  (or  
<nothing>?) to the for loop, and  only otherwise goes on to evaluate  
the rest of the second assignment and break call? So any warning would  
abort processing of the remaining expressions?

Are the assignment "xx<-x" and the break call considered
"handlers" as
described in the help page?

I see the purpose of the last argument. Taking out the warning =  
function(w{}) changes the behavior when none of the elements in the  
vector passed to x were valid.

 > for (x in c(-2, -1)) {
+ 	tryCatch({
+ 		y <- log(x)
+ 		xx <- x
+ 		break
+ 	})
+ }
Warning message:
In log(x) : NaNs produced
 > print(xx)
[1] -2


-- 

David Winsemius

On Dec 7, 2008, at 12:21 PM, Gabor Grothendieck wrote:
> Try this:
>
> for (x in c(-2, 2, 4)) {
> 	tryCatch({
> 		y <- log(x)
> 		xx <- x
> 		break
> 	}, warning = function(w) {})
> }
> print(xx) # 2
>
>
>
> On Sun, Dec 7, 2008 at 2:38 AM, Andreas Wittmann
> <andreas_wittmann at gmx.de> wrote:
>> Dear R useRs,
>>
>> with the following piece of code i try to find the first value  
>> which can be
>> calculated without warnings
>>
>> `test` <- function(a)
>> {
>> repeat
>> {
>>  ## hide warnings
>>  suppressWarnings(log(a))
>>
>>  if (exists("last.warning", envir = .GlobalEnv))
>>  {
>>    a <- a + 0.1
>>    ## clear existing warnings
>>    rm("last.warning", envir = .GlobalEnv)
>>  }
>>
>> if(a > 5 || !exists("last.warning", envir = .GlobalEnv))
>>  break
>> }
>>
>> return(a)
>> }
>>
>> if i run this with test(-3), i would expect a=0 as return value.
>> Is it also possible to hide warnings during my function, i guess i  
>> use
>> suppressWarnings in a wrong way here?
>>
>> Thanks and best regards
>>
>> Andreas
>>
>> ______________________________________________
>> R-help at r-project.org mailing list
>> https://stat.ethz.ch/mailman/listinfo/r-help
>> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
>> and provide commented, minimal, self-contained, reproducible code.
>>
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.

log(-1) is not an error in R bug it does generate a warning and a NaN result.

If there is a warning then the statement that causes the warning and all
remaining statements in the tryCatch are not executed.  Instead the
warning= function is invoked.

On Sun, Dec 7, 2008 at 1:09 PM, David Winsemius <dwinsemius at
comcast.net> wrote:
> Thanks for that. I am having some difficulty figuring out how it works. The
> help page tells me (I think) that "y<-log(x)" gets matched to
expr. Am I
> correct in thinking that if either an error or warning is produced, that
> tryCatch silently returns <something>  (or <nothing>?) to the
for loop, and
>  only otherwise goes on to evaluate the rest of the second assignment and
> break call? So any warning would abort processing of the remaining
> expressions?
>
> Are the assignment "xx<-x" and the break call considered
"handlers" as
> described in the help page?
>
> I see the purpose of the last argument. Taking out the warning >
function(w{}) changes the behavior when none of the elements in the vector
> passed to x were valid.
>
>> for (x in c(-2, -1)) {
> +       tryCatch({
> +               y <- log(x)
> +               xx <- x
> +               break
> +       })
> + }
> Warning message:
> In log(x) : NaNs produced
>> print(xx)
> [1] -2
>
>
> --
>
> David Winsemius
>
> On Dec 7, 2008, at 12:21 PM, Gabor Grothendieck wrote:
>
>> Try this:
>>
>> for (x in c(-2, 2, 4)) {
>>        tryCatch({
>>                y <- log(x)
>>                xx <- x
>>                break
>>        }, warning = function(w) {})
>> }
>> print(xx) # 2
>>
>>
>>
>> On Sun, Dec 7, 2008 at 2:38 AM, Andreas Wittmann
>> <andreas_wittmann at gmx.de> wrote:
>>>
>>> Dear R useRs,
>>>
>>> with the following piece of code i try to find the first value
which can
>>> be
>>> calculated without warnings
>>>
>>> `test` <- function(a)
>>> {
>>> repeat
>>> {
>>>  ## hide warnings
>>>  suppressWarnings(log(a))
>>>
>>>  if (exists("last.warning", envir = .GlobalEnv))
>>>  {
>>>   a <- a + 0.1
>>>   ## clear existing warnings
>>>   rm("last.warning", envir = .GlobalEnv)
>>>  }
>>>
>>> if(a > 5 || !exists("last.warning", envir =
.GlobalEnv))
>>>  break
>>> }
>>>
>>> return(a)
>>> }
>>>
>>> if i run this with test(-3), i would expect a=0 as return value.
>>> Is it also possible to hide warnings during my function, i guess i
use
>>> suppressWarnings in a wrong way here?
>>>
>>> Thanks and best regards
>>>
>>> Andreas
>>>
>>> ______________________________________________
>>> R-help at r-project.org mailing list
>>> https://stat.ethz.ch/mailman/listinfo/r-help
>>> PLEASE do read the posting guide
>>> http://www.R-project.org/posting-guide.html
>>> and provide commented, minimal, self-contained, reproducible code.
>>>
>>
>> ______________________________________________
>> R-help at r-project.org mailing list
>> https://stat.ethz.ch/mailman/listinfo/r-help
>> PLEASE do read the posting guide
>> http://www.R-project.org/posting-guide.html
>> and provide commented, minimal, self-contained, reproducible code.
>
>