column names of summaryBy

[ggrothendieck]

Consider this code from here except we have used warpbreaks as the data. The problem is that the first column name is set to the factor name and here we wanted to call it level so that all the data frames generated below have the same column names because we rbindlist all the data frames together. We had to overwrite the names -- see the SetNames line. Perhaps summaryBy could provide some constrol over that through an argument or perhaps if we wrote summaryBy(list(nums, level = x), data, FUN = fun) it could use the name level or whatever name was used there.

library(doBy)
library(data.table)

data <- warpbreaks

facs <- names(Filter(is.factor, data))
nums <- names(Filter(is.numeric, data))

Map(function(x) {
  s <- summaryBy(list(nums, x), data, FUN = list("min", "mean", "median", "max"))
  setNames(s, c("level", names(s)[-1]))
}, facs)
rbindlist(res, idcol = "factor")

##     factor  level breaks.min breaks.mean breaks.median breaks.max
##     <char> <fctr>      <num>       <num>         <num>      <num>
## 1:    wool      A         10    31.03704          26.0         70
## 2:    wool      B         13    25.25926          24.0         44
## 3: tension      L         14    36.38889          29.5         70
## 4: tension      M         12    26.38889          27.0         42
## 5: tension      H         10    21.66667          20.5         43

[ggrothendieck]

The dplyr version of this is slightly shorter because of the simpler way to set the level column name as mentioned above but also because where(is.numeric) can select numeric columns. Maybe summaryBy could have some way of doing that too. For example maybe if the first arg to summaryBy has any components that are functions it would use those columns for which the function is TRUE. Then we could write:

summaryBy(list(is.numeric, level = x), data, FUN = list("min", "mean", "median", "max"))