[R] tryCatch - Continuing for/next loop after error

William Dunlap wdunlap at tibco.com
Thu Mar 10 17:29:59 CET 2011 

> -----Original Message-----
> From: r-help-bounces at r-project.org 
> [mailto:r-help-bounces at r-project.org] On Behalf Of Jonathan P Daily
> Sent: Thursday, March 10, 2011 7:49 AM
> To: Nipesh Bajaj
> Cc: r-help at r-project.org
> Subject: Re: [R] tryCatch - Continuing for/next loop after error
> 
> Wow, I had a major brain fart there. The function after error 
> was supposed 
> to accept an argument, and next won't do anything there 
> because inside the 
> function there is no loop. Also, the first argument of tryCatch is an 
> expression, so do the assignment there. Corrections inline below:

The stuff about the error handle requiring an argument and
not being able to call next() from outside of a loop is true,
but you do not need to do the assignment inside the main expression
of tryCatch().  I find it clearer if you do the assignment outside
of tryCatch() so the main expression can return one thing and the
error handler can return something else.  You can use this in an
*apply() function or a for loop, while the assignment-in-tryCatch()
won't work in a *apply() call.  E.g., suppose we have a function
that only works on positive odd integers:
  o <- function (x) {
      stopifnot(x%%2 == 1, x>0)
      (x - 1)/2
  }
Then you can use tryCatch to flag the bad inputs with a code
that tells you about the bad input:
  > sapply(1:5, function(i)tryCatch(o(i), error=function(e)-1000-i))
  [1]     0 -1002     1 -1004     2
or with a for loop where you don't have to prefill the output
vector with any particular values:
  > value <- numeric(5)
  > for(i in 1:5) value[i] <- tryCatch(o(i), error=function(e)-1000-i)
  > value
  [1]     0 -1002     1 -1004     2

To do that with with an assignment in the main expression of tryCatch
would require something like
  > value <- -1000 -(1:5)
  > for(i in 1:5) tryCatch(value[i] <- o(i), error=function(e)NULL)
  > value
  [1]     0 -1002     1 -1004     2
which seems to me uglier and harder to follow since value's entries
are filled in different parts of the code.

Bill Dunlap
Spotfire, TIBCO Software
wdunlap tibco.com 
> --------------------------------------
> Jonathan P. Daily
> Technician - USGS Leetown Science Center
> 11649 Leetown Road
> Kearneysville WV, 25430
> (304) 724-4480
> "Is the room still a room when its empty? Does the room,
>  the thing itself have purpose? Or do we, what's the word... 
> imbue it."
>      - Jubal Early, Firefly
> 
> Nipesh Bajaj <bajaj141003 at gmail.com> wrote on 03/10/2011 10:21:38 AM:
> 
> > [image removed] 
> > 
> > Re: [R] tryCatch - Continuing for/next loop after error
> > 
> > Nipesh Bajaj 
> > 
> > to:
> > 
> > Jonathan P Daily
> > 
> > 03/10/2011 10:21 AM
> > 
> > Cc:
> > 
> > r-help
> > 
> > Hi Jonathan, I was also trying to understand this tryCatch 
> function on
> > my own problem. Here is mine:
> > 
> > fn1 <- function(x) {
> >       if(as.integer(x) == 5) {
> >             stop("stop") 
> >          }
> >       return(x+5)
> >    }
> > res <- matrix(NA, 5, 7)
> > for(i in 1:5) {
> >       for(j in 1:7) {
> >             res[i,j] <- tryCatch(fn1(j), error = function() next)
> >          }
> >    }
> 
> for(i in 1:5) { for(j in 1:7) {
>         tryCatch(res[i,j] <- fn1(j), error = function(x) x)
> } }
> 
> > 
> > I was expecting while completion of these loops, the 5th column of
> > 'res' will have NA values. However I got some error:
> > 
> > Error in value[[3L]](cond) : unused argument(s) (cond)
> > 
> > What is the actual problem here?
> > 
> > Thanks,
> > 
> > 
> > On Thu, Mar 10, 2011 at 7:19 PM, Jonathan P Daily <jdaily at usgs.gov> 
> wrote:
> > > --------------------------------------
> > > Jonathan P. Daily
> > > Technician - USGS Leetown Science Center
> > > 11649 Leetown Road
> > > Kearneysville WV, 25430
> > > (304) 724-4480
> > > "Is the room still a room when its empty? Does the room,
> > >  the thing itself have purpose? Or do we, what's the 
> word... imbue 
> it."
> > >     - Jubal Early, Firefly
> > >
> > > r-help-bounces at r-project.org wrote on 03/10/2011 03:51:15 AM:
> > >
> > >> [image removed]
> > >>
> > >> [R] tryCatch - Continuing for/next loop after error
> > >>
> > >> Costas
> > >>
> > >> to:
> > >>
> > >> r-help
> > >>
> > >> 03/10/2011 03:53 AM
> > >>
> > >> Sent by:
> > >>
> > >> r-help-bounces at r-project.org
> > >>
> > >> Please respond to costas.vorlow
> > >>
> > >> Dear all,
> > >>
> > >> I am not sure I understand fully the functionality of 
> "tryCatch" and
> > >> "try" commands and how these are used to continue a 
> for/next loop if 
> an
> > >> error occurs within the loop.
> > >>
> > >> Can somebody point me to material (or share some code) with more
> > >> extensive examples than the ones in the help/FAQ pages?
> > >>
> > >> Do explain my problem in more detail:
> > >>
> > >> for (i in 100:1000)
> > >> {
> > >>       ## do some other stuff
> > >>
> > >>      dataset<-head(data,i)
> > >>      tryCatch(estimatemodel(dataset))
> > >
> > > if estimatemodel returns an error (i.e. via a call to 
> stop()), then 
> this
> > > should break out of the loop:
> > >
> > > tryCatch(estimatemodel(dataset), error = function() break)
> > >
> > > if you want to skip to the next iteration, use:
> > > tryCatch(estimatemodel(dataset), error = function() next)
> > >
> > >>
> > >> }
> > >>
> > >> My for/next loop reads in data (increasing the dataset 
> by one point 
> at
> > >> every loop run) and then estimates a model. When the problem is
> > >> computationally singular, the loop exits. I want to 
> continue the loop
> > >> and register an "error" estimation value for that step. 
> However when 
> I
> > >> use use the try tryCatch(estimatemodel(data)) (where 
> estimatemodel() 
> is
> > >> a wrapper function calling the model estimation and optimization
> > >> routines), the problem still persists.
> > >>
> > >> Is this the correct way to use tryCatch (or try) or 
> should these go
> > >> inside the actual code bits (i.e., in a more low level 
> fashion) that
> > >> conduct the optimization and model estimation?
> > >>
> > >> Apologies if this is not clear enough.
> > >>
> > >> Best,
> > >> Costas
> > >>
> > >> ______________________________________________
> > >> 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.
> > >
> 
> ______________________________________________
> 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.
>

More information about the R-help mailing list