[R-Forge #1602] Can't merge multiple objects with character values

[joshuaulrich]

Submitted by: Ken Williams (@kenahoo)
Assigned to: Nobody
R-Forge link

It appears that the merge() function fails when fed more than 2 xts objects, and they contains character data:

> x <- xts(letters[1:5], Sys.Date() + cumsum(1:5)*60)
> merge(x)
           [,1]
2011-09-28 'a' 
2011-09-30 'b' 
2011-10-03 'c' 
2011-10-07 'd' 
2011-10-12 'e' 
> merge(x,x)
           x   x.1
2011-09-28 'a' 'a'
2011-09-30 'b' 'b'
2011-10-03 'c' 'c'
2011-10-07 'd' 'd'
2011-10-12 'e' 'e'
> merge(x, x, x)
Error in merge.xts(x, x, x) : unsupported data type

By contrast, it works fine when the data is numeric:

> merge(x)
           [,1]
2011-09-28    1
2011-09-30    2
2011-10-03    3
2011-10-07    4
2011-10-12    5
> merge(x,x)
           x x.1
2011-09-28 1   1
2011-09-30 2   2
2011-10-03 3   3
2011-10-07 4   4
2011-10-12 5   5
> merge(x,x,x)
           x x.1 x.2
2011-09-28 1   1   1
2011-09-30 2   2   2
2011-10-03 3   3   3
2011-10-07 4   4   4
2011-10-12 5   5   5

Perhaps something's being silently converted to a factor internally?

-Ken

[joshuaulrich]

The n-way merge case currently only works for logical, integer, and numeric types. Character, complex, and any other type will fail. It should be possible to add support for character and complex data.

[SteveBronder]

Wanted to note this also effects lag.xts(). A workaround for that function is to do an lapply() over your lags and then bind them all together with cbind.zoo()

library(xts)
set.seed(1234)
dat = xts(data.frame(char = rep(c("A","B"),5)), order.by =as.POSIXct(paste0("1992-01-",1:10)) )

xLagDiff = lapply(1:5,function(i) {
  lags = lag.xts(dat,k = i)
  colnames(lags) = paste0(colnames(lags),i)
  lags
  })

xLagDiff2 = do.call(cbind.zoo,xLagDiff)
xLagDiff2

[joshuaulrich]

@SteveBronder, your example doesn't run. lag is only defined as a function, so your lapply call fails. Could you please edit your example to be minimal (i.e. without requiring another package and running an ARIMA model) and reproducible (runs from a clean R session).

[SteveBronder]

@joshuaulrich apologies, amended the example

[joshuaulrich]

@SteveBronder awesome, thanks!

Note that this only affects lag because lag calls merge when length(k) is > 2. In fact, it works if length(k) <= 2. For example:

R> lag(dat,2:3)
           char char.1
1992-01-01 NA   NA    
1992-01-02 NA   NA    
1992-01-03 "A"  NA    
1992-01-04 "B"  "A"   
1992-01-05 "A"  "B"   
1992-01-06 "B"  "A"   
1992-01-07 "A"  "B"   
1992-01-08 "B"  "A"   
1992-01-09 "A"  "B"   
1992-01-10 "B"  "A"   

[joshuaulrich]

Yeesh, just looked at the R-Forge issue tracker and saw:

Date Submitted: 2011-09-27 12:46

I hope "better late than never" applies... Sorry about the massive delay, @kenahoo.