R: ifelse on string

Question

I am populating a new variable of a dataframe, based on string conditions from another variable. I receive the following error msg:

Error in Source == "httpWWW.BGDAILYNEWS.COM" | Source == : operations are possible only for numeric, logical or complex types

My code is as follows:

County <- ifelse(Source == 'httpWWW.BGDAILYNEWS.COM' | 'WWW.BGDAILYNEWS.COM', 'Warren', ifelse(Source == 'httpWWW.HCLOCAL.COM' | 'WWW.HCLOCAL.COM', 'Henry', ifelse(Source == 'httpWWW.KENTUCKY.COM' | 'WWW.KENTUCKY.COM', 'Fayette', ifelse(Source == 'httpWWW.KENTUCKYNEWERA.COM' | 'WWW.KENTUCKYNEWERA.COM', 'Christian') )))

Solution 1

I suggest you break down that deeply nested ifelse statement into more manageable chunks.

But the error is telling you that you cannot use | like that. 'a' | 'b' doesn't make sense since its a logical comparison. Instead use %in%:

Source %in% c('htpWWW.BGDAILYNEWS.com', 'WWW.BGDAILYNEWS.COM')

I think... If I understand what you're doing, you will be much better off using multiple assignments:

County = vector(mode='character', length=length(Source))
County[County %in% c('htpWWW.BGDAILYNEWS.com', 'WWW.BGDAILYNEWS.COM')] <- 'Warren'
etc.

You can also use a switch statement for this type of thing:

myfun <- function(x) {
  switch(x,
         'httpWWW.BGDAILYNEWS.COM'='Warren',
         'httpWWW.HCLOCAL.COM'='Henry',
         etc...)
}

Then you want to do a simple apply (sapply) passing each element in Source to myfun:

County = sapply(Source, myfun)

Or finally, you can use factors and levels, but I'll leave that as an exercise to the reader...

Solution 2

A different approach:

county <- c("Warren","Henry","Fayette","Christian")
sites <- c("WWW.BGDAILYNEWS.COM","WWW.HCLOCAL.COM","WWW.KENTUCKY.COM","WWW.KENTUCKYNEWERA.COM")
County <- county[match(gsub("^http","",Source), sites)]

This will return NA for strings that do no match any of the given inputs.

Using Hadley's suggestion (lookup-tables-character-subsetting):

lookup <- c(WWW.BGDAILYNEWS.COM="Warren", WWW.HCLOCAL.COM="Henry", WWW.KENTUCKY.COM="Fayette", WWW.KENTUCKYNEWERA.COM="Christian")
County <- unname(lookup[gsub("^http","",Source)])

Comments

[Señor O]

+1 for switch. If you find yourself using more than 2 ifelse's, you should probably be using switch or cut.

[hadley]

You could also eliminate the match and use character subsetting directly: github.com/hadley/devtools/wiki/…

[NiuBiBang]

@Justin, sorry to ask but can you post a more specific example for the switch function? Specifically, how do I create the County variable from the Source variable using switch? Much agreed that my ifelses are not parsimonious.