add column combining argument to unnest()

[mikabr]

There are use cases where it seems like it would be useful to have some control over the column names that unnest() produces, for example if there are multiple columns of data frames with the same columns:

df <- dplyr::data_frame(
  x = 1:3,
  y = list(data_frame(a = runif(3), b = runif(3))),
  z = list(data_frame(a = runif(3), b = runif(3)))
)

tidyr::unnest(df)

# Source: local data frame [9 x 5]
#
#       x         a         b         a         b
#   (int)     (dbl)     (dbl)     (dbl)     (dbl)
#1     1 0.1828750 0.6893545 0.4887932 0.6530871
#2     1 0.5282101 0.1400115 0.5172123 0.4216703
#3     1 0.2047363 0.9180311 0.5185549 0.8593084
#4     2 0.1828750 0.6893545 0.4887932 0.6530871
#5     2 0.5282101 0.1400115 0.5172123 0.4216703
#6     2 0.2047363 0.9180311 0.5185549 0.8593084
#7     3 0.1828750 0.6893545 0.4887932 0.6530871
#8     3 0.5282101 0.1400115 0.5172123 0.4216703
#9     3 0.2047363 0.9180311 0.5185549 0.8593084

Maybe something like .combine, that combines the nested column names:

tidyr::unnest(df, .combine = TRUE)

# Source: local data frame [9 x 5]
#
#       x       y.a       y.b       z.a       z.b
#   (int)     (dbl)     (dbl)     (dbl)     (dbl)
#1     1 0.1828750 0.6893545 0.4887932 0.6530871
#2     1 0.5282101 0.1400115 0.5172123 0.4216703
#3     1 0.2047363 0.9180311 0.5185549 0.8593084
#4     2 0.1828750 0.6893545 0.4887932 0.6530871
#5     2 0.5282101 0.1400115 0.5172123 0.4216703
#6     2 0.2047363 0.9180311 0.5185549 0.8593084
#7     3 0.1828750 0.6893545 0.4887932 0.6530871
#8     3 0.5282101 0.1400115 0.5172123 0.4216703
#9     3 0.2047363 0.9180311 0.5185549 0.8593084

Or supplying a function for how to combine them?

[hadley]

Do you have any examples of a similar API in other functions/packages? That would help me think about what would be most natural for tidyr.

[hadley]

If you dislike this API, can you please let me know ASAP?

ldf <- list(data_frame(x = 1))
data_frame(x = ldf, y = ldf) %>%
  unnest(.sep = "_") %>%
  names()
#> [1] "x_x" "y_x"

[mikabr]

That looks great.