spread with missing values of a factor of key_value

[MCOtto]

I want to make tables from different data where I use spread to make tables with levels of a factor for columns. spread is a good tool to do this. I need all columns even if there are no observations of some levels of the factor. reshape::cast with add.missing accomplished this; spread has no option with the same ability. Here is a simple example:

require(tidyr)
 require(dplyr)
 Data<-expand.grid(
    row=paste("r",1:3,sep=""),
    col=paste("c",1:4,sep="")   
 )

 Data<-mutate(Data,
    value=as.integer(gsub("[rc]","",paste(row,col,sep="")))
 )

 filter(Data,col!="c3")%>%
 spread(col,value,fill=0)

 require(reshape)
 filter(Data,col!="c3")%>%
 cast(row~col,
  add.missing=TRUE,value="value",fill=0
 )

cast keeps c3 even though there are not values because levels(Data$col) has all the columns. Spread does not:

row c1 c2 c3 c4
1 r1 11 12 0 14
2 r2 21 22 0 24
3 r3 31 32 0 34

This is the facility from cast I would like retained or returned. May be I am missing something in the spread options or some other function I should be using in tid- or dpl-yr.

[desmondpatel]

Is tidyr available for R version 3.03?

[MCOtto]

This does not address this issue for spread, but you look in the description file for the package.

Depends: R (>= 3.1.0)

means that it requires a version later than 3.03. I would update R.

Mark

[hadley]

MRE:

require(tidyr)
require(dplyr)

df <- expand.grid(
    row = paste("r", 1:3, sep = ""),
    col = paste("c", 1:4, sep = "")   
  ) %>% mutate(
    value = paste0(rep(1:3, each = 4), 1:4)
  ) %>%
  filter(col != "c3")

df %>% spread(col, value, fill = 0)
df %>% reshape2::dcast(row ~ col, drop = FALSE, fill = 0)

[hadley]

I've called the argument drop, since that what it is in reshape2 (and it's consistent with other uses)