What is the equivalent of `unstack()` in data.table?-CodePudding

I want to convert the long to wide format of my data using data.table. Normally I use unstack(), but I don't know how to do it in data.table. Below is an example. As output I expect three columns and three rows.

library(data.table)
set.seed(1)
df <- data.frame(class = factor(rep(c("A", "B", "C"), times = 3)),
                 value = runif(9))
unstack(df, form = value ~ class)
#>           A         B         C
#> 1 0.2655087 0.3721239 0.5728534
#> 2 0.9082078 0.2016819 0.8983897
#> 3 0.9446753 0.6607978 0.6291140

dt <- data.table(df)
dcast(dt, formula = value ~ class, value.var = "value")
#>        value         A         B         C
#> 1: 0.2016819        NA 0.2016819        NA
#> 2: 0.2655087 0.2655087        NA        NA
#> 3: 0.3721239        NA 0.3721239        NA
#> 4: 0.5728534        NA        NA 0.5728534
#> 5: 0.6291140        NA        NA 0.6291140
#> 6: 0.6607978        NA 0.6607978        NA
#> 7: 0.8983897        NA        NA 0.8983897
#> 8: 0.9082078 0.9082078        NA        NA
#> 9: 0.9446753 0.9446753        NA        NA

Additionally, I don't want to use an aggregate function, but in data.table on real data I see: Aggregate function missing, defaulting to 'length', so there are fewer rows in the result.

CodePudding user response：

I think this is what you want:

dcast(dt, rowid(class) ~ class)
#   class         A          B         C
#1:     1 0.2655087 0.3721239 0.5728534
#2:     2 0.9082078 0.2016819 0.8983897
#3:     3 0.9446753 0.6607978 0.6291140

https://stackoverflow.com/a/52501268/10276092

CodePudding user response：

We can try split within dt like below (but this option is available only if we have the same number of values per class, thus, the dcast option by @M.Viking is recommended for general cases)

> dt[, split(value, class)]
           A         B         C
1: 0.2655087 0.3721239 0.5728534
2: 0.9082078 0.2016819 0.8983897
3: 0.9446753 0.6607978 0.6291140