data.table's := prints when data.table is not the first class

[rubenarslan]

I left this as a comment on a closed data.table issue previously, that probably was a bad idea. I think this might have to be fixed on the dplyr side, since the solution in data.table lives in the print.data.table function, which isn't called when "grouped_dt" is in the pole position.

The bug Rdatatable/data.table#505 happens when e.g. dplyr::group_by was used on a data.table. Presumably the error can also occur when other classes remove data.table from being the pole position class.

Here's a failing test:

cat(knit(text='`:=` prints in conjunction with dplyr, because dplyr adds all kind of classes to data.tables

```{r}
library(dplyr); library(data.table)
mtcars = data.table(mtcars)

grouped = mtcars %>% group_by(cyl)
grouped[, mpg := 2]
class(grouped)

class(grouped) = c("data.table", "grouped_dt", "tbl_dt", "tbl", "data.frame") # reorder classes
grouped[, mpg := 2]

    # this is how I currently "fix" it.
grouped = mtcars %>% group_by(cyl) %>% data.table()
grouped[, mpg := 2]
class(grouped)
```'))

This is with data.table_1.9.5, dplyr_0.4.1 and knitr_1.9.

[hadley]

I'm extending the data table class so I'm pretty sure my class names should come first.

[rubenarslan]

Sorry, I didn't mean to suggest your class names shouldn't come first – but if dplyr is extending the dt class, it would be better if printing didn't break when dt functionality (:=) is later used.