I see that both which(condition) and subset(data,condition) both treat NA's in condition that same as FALSE's. This leads people to use those functions for their NA-treating properties instead of for their main functionality (which may not be the best way to get things done).
I wonder how much code would break if x[condition], for a logical condition, would break if it returned only the elements of x corresponding to TRUE's in the condition (instead of also returning NA's for NA's in condition). How much currently broken code would start working? I suspect omitting the NA's from the output might be better. I'm don't know about the case of NA's in integer subscripts. This would affect things in the logical case because x[NA] would use the logical case and x[c(1,NA)] the integer. As for your VisTRUE, the following is much faster but silently coerces non-logical arguments to logicals. VisTRUE2 <- function(x) !is.na(x) && x Bill Dunlap Spotfire, TIBCO Software wdunlap tibco.com > -----Original Message----- > From: Joshua Wiley [mailto:jwiley.ps...@gmail.com] > Sent: Sunday, October 17, 2010 6:27 PM > To: David Winsemius > Cc: William Dunlap; R-help@r-project.org > Subject: Re: [R] delete data row > > I used the -which() construct initially to try to show "deleting" > cases. I believe it hung around longer than it should have. That > said, I have also had David's experience with NAs. What about a > vectorized version of identical(TRUE, x)? This avoids the which() > problem Bill pointed out, and the NA issue David mentioned. Does it > introduce new problems? > > x <- 1:10 > y <- log(x-5) > VisTRUE <- Vectorize(isTRUE) > x[VisTRUE(y > -Inf)] > > Josh > > > On Sun, Oct 17, 2010 at 4:38 PM, David Winsemius > <dwinsem...@comcast.net> wrote: > > > > On Oct 17, 2010, at 3:56 PM, William Dunlap wrote: > > > >> > >>> I had been thinking of: > >>>> > >>>> x <- c(1, (2^(0.5))^2 , 3, 5, (2^(0.5))^2 , 3, 1) > >>>> y <- 2 > >>>> x[-which(zapsmall(x-y) == 0)] > >>> > >>> [1] 1 3 5 3 1 > >> > >> Using which() to convert logicals into integer > >> subscripts is almost always unnecessary and often wrong. > > > > At one time I believed that too. However, in the situation > where the test > > produces NA rather than a numeric value when one is > indexing in the first > > argument. I have had the unpleasant experience of pages if > useless and > > frustrating to understand output because of this "feature". > > > > I learned to either use which() in the first argument to > "[" or to use > > subset to avoid inadvertent "returns" from logical indexing. > > > >> x <- 1:10 > >> y <- log(x-5) > > Warning message: > > In log(x - 5) : NaNs produced > >> x[y>-Inf] > > [1] NA NA NA NA 6 7 8 9 10 > > > >> x[which(y>-Inf)] > > [1] 6 7 8 9 10 > > > > If that test were used in a dataframe indexing, the entire > line might come > > back as a "result". > > > > > > > >> In this case it fails when no x is close to y, > >> because integer(0) is the same thing as -integer(0): > >> > >>> x[-which(zapsmall(x-10) == 0)] > >> > >> numeric(0) > >> > >> The whichless version, using logical subscripts, > >> works (in this case we want all of x): > >> > >>> x[zapsmall(x-10)!=0] > >> > >> [1] 1 2 3 5 2 3 1 > > > > Maybe the rule should be don't use the -which construction: > > > >> x <- c(1, (2^(0.5))^2 , 3, 5, (2^(0.5))^2 , 3, 1) > >> y <- 2 > >> x[which(zapsmall(x-10) != 0)] > > [1] 1 2 3 5 2 3 1 > > > > -- > > David. > >> > >> When using logicals as subscripts, read the "[" > >> as "such that". > >> > >> Bill Dunlap > >> Spotfire, TIBCO Software > >> wdunlap tibco.com > > > > > > -- > Joshua Wiley > Ph.D. Student, Health Psychology > University of California, Los Angeles > http://www.joshuawiley.com/ > ______________________________________________ R-help@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.
