show dimensions of list columns with DT

[randomgambit]

Hello everyone,

I am all IN for a better integration of DT in the tidyverse ecosystem. DT is really great and I think some features of the tidyverse could be very useful in a DT setting.

In particular, I am a huge fan of purrr and list columns. Consider this.

> tibble(group = c(1,1,1,2,2,2),
+            val = list(list(1,2,3),list(1,2,3),list(1,2,3),
+                       list(1,2,3),list(1,2,3),list(1,2,3))) 
# A tibble: 6 x 2
  group val       
  <dbl> <list>    
1     1 <list [3]>
2     1 <list [3]>
3     1 <list [3]>
4     2 <list [3]>
5     2 <list [3]>
6     2 <list [3]>

As you can see, I can clearly see the dimension (or length) of each element of the list-column.
As far as I know, this is not possible with DT

>  data.table(group = c(1,1,1,2,2,2),
+            val = list(list(1,2,3),list(1,2,3),list(1,2,3),
+                       list(1,2,3),list(1,2,3),list(1,2,3))) 
   group    val
1:     1 <list>
2:     1 <list>
3:     1 <list>
4:     2 <list>
5:     2 <list>
6:     2 <list>

This is a bit unfortunate. Here the idea is to use furrr (which leverages purrr) to easily run computations in parallel across the list columns.

> mydf[,  newval := future_map(val, ~length(.x))]
> mydf
   group    val newval
1:     1 <list>      3
2:     1 <list>      3
3:     1 <list>      3
4:     2 <list>      3
5:     2 <list>      3
6:     2 <list>      3

Being able to see the dimension of the embedded list (or DT) is very important when one creates list columns with whole data.frames and run regressions on them. What do you think? Could the printing of DT show an output similar to the printing of the tibble?

Thanks!

[randomgambit]

Similarly, this shows the dimensions of the embedded DT. Why not the opposite as well?

seq(1:3) %>% map(., ~ data.table(group = .x,
                                 mycol = seq(1, .x))) %>%
  enframe()
# A tibble: 3 x 2
   name value           
  <int> <list>          
1     1 <df[,2] [1 x 2]>
2     2 <df[,2] [2 x 2]>
3     3 <df[,2] [3 x 2]>

[franknarf1]

Related maybe, re user-chosen printing rules: #3414 , #1523 (comment) (and other items in that thread)

Side note: future_map(val, ~length(.x)) looks like a lot of overhead for lengths(val) from base R..?

[randomgambit]

@franknarf1 yes, thats. the future_map call was an example. In real life, I would run much more complex operations in parallel :)

[MichaelChirico]

#3414 will indeed be the start to this