diffs produces a data.frame with list columns

[jonocarroll]

I am using comparedf to compare two versions of a data.frame containing dates, and want to summarise the change ("value was x, now is y"). diffs.comparedf creates list-columns for values.x and values.y which don't behave nicely.

As a worked example, consider a data.frame with vector columns

dates <- as.Date(c("2024-01-01", "2024-01-02"))

df_veccols <- data.frame(num = 1:2, date = I(dates))
df_veccols
#>   num       date
#> 1   1 2024-01-01
#> 2   2 2024-01-02
str(df_veccols)
#> 'data.frame':    2 obs. of  2 variables:
#>  $ num : int  1 2
#>  $ date: AsIs, format: "2024-01-01" "2024-01-02"

paste(df_veccols$num, df_veccols$date, sep = ": ")
#> [1] "1: 2024-01-01" "2: 2024-01-02"

However, this internal function

arsenal/R/comparedf.internal.R

Lines 287 to 292 in 497db28

	tolist <- function(df)
	{
	df$values.x <- I(as.list(df$values.x)) # need the I() for factors and dates to show up right
	df$values.y <- I(as.list(df$values.y))
	df
	}

creates a version with list columns, like this

df_listcols <- data.frame(num = 1:2, date = I(as.list(dates)))
df_listcols
#>   num       date
#> 1   1 2024-01-01
#> 2   2 2024-01-02
str(df_listcols)
#> 'data.frame':    2 obs. of  2 variables:
#>  $ num : int  1 2
#>  $ date:List of 2
#>   ..$ : Date, format: "2024-01-01"
#>   ..$ : Date, format: "2024-01-02"
#>   ..- attr(*, "class")= chr "AsIs"

paste(df_listcols$num, df_listcols$date, sep = ": ")
#> [1] "1: 19723" "2: 19724"

I believe the core incompatibility is with

as.character(I(as.list(dates[1])))
#> [1] "19723"

# vs

as.character(I(dates[1]))
#> [1] "2024-01-01"

df_listcols$date
#> [[1]]
#> [1] "2024-01-01"
#> 
#> [[2]]
#> [1] "2024-01-02"

unlist(df_listcols$date)
#> [1] 19723 19724

In arsenal, this is

df1 <- data.frame(num = 1:2, dates = dates)
df2 <- data.frame(num = 1:2, dates = dates + c(0, 1))

diffs <- arsenal::diffs(arsenal::comparedf(df1, df2))
diffs
#>   var.x var.y ..row.names..   values.x   values.y row.x row.y
#> 1 dates dates             2 2024-01-02 2024-01-03     2     2
str(diffs)
#> 'data.frame':    1 obs. of  7 variables:
#>  $ var.x        : chr "dates"
#>  $ var.y        : chr "dates"
#>  $ ..row.names..: int 2
#>  $ values.x     :List of 1
#>   ..$ : AsIs, format: "2024-01-02"
#>   ..- attr(*, "class")= chr "AsIs"
#>  $ values.y     :List of 1
#>   ..$ : AsIs, format: "2024-01-03"
#>   ..- attr(*, "class")= chr "AsIs"
#>  $ row.x        : int 2
#>  $ row.y        : int 2

Should I expect this work?

paste(diffs$var.x, "was", diffs$values.x, "now is", diffs$values.y)
#> [1] "dates was 19724 now is 19725"

As a workaround, I currently need to do this (unless there is a better way)

paste(diffs$var.x, "was", diffs$values.x[[1]], "now is", diffs$values.y[[1]])
#> [1] "dates was 2024-01-02 now is 2024-01-03"