reframe() is dropping columns when you use rowwise()

[mutahiwachira]

Minimal reprex

library(dplyr)
quantile_df <- function(x, probs = c(0.25, 0.5, 0.75)) {
  tibble(
    val = quantile(x, probs, na.rm = TRUE),
    quant = probs
  )
}

# Actual Behavioiur: Removes the grouping columns and just returns cols from rowwise
starwars %>%
  rowwise() %>% 
  reframe(quantile_df(height)) %>%
  ungroup()
  # 261 rows, 2 cols from quantile_df

# Expected Beavhiour: Preserve them like in this code
starwars %>% 
  rowwise() %>% 
  mutate(quantiles = list(quantile_df(height))) %>% 
  unnest(quantiles) %>%
  ungroup()
  # 261 rows, all columns preserved

I want to be able to go from 1 row to multiple rows while keeping the previous information.

The use case is described below.

Use case

• I was doing a simulation of the German Tank problem following a very functional/list-column heavy workflow. I have a dataframe called sensitivities which has columns that I need to generate and describe my samples. It looks like this:

## A tibble: 18 × 4
#  pop_size prop_of_pop all_tanks     sample_size
#       <dbl>       <dbl> <list>              <dbl>
#   1     1000         0.1 <int [1,000]>         100
#   2     1000         0.2 <int [1,000]>         200
#   3     1000         0.3 <int [1,000]>         300
#   4     1000         0.4 <int [1,000]>         400
#   5     1000         0.5 <int [1,000]>         500
#   6     1000         0.6 <int [1,000]>         600
#   7     1000         0.7 <int [1,000]>         700
#   8     1000         0.8 <int [1,000]>         800
#   9     1000         0.9 <int [1,000]>         900
# 10     2000         0.1 <int [2,000]>         200
# 11     2000         0.2 <int [2,000]>         400
# 12     2000         0.3 <int [2,000]>         600
# 13     2000         0.4 <int [2,000]>         800
# 14     2000         0.5 <int [2,000]>        1000
# 15     2000         0.6 <int [2,000]>        1200
# 16     2000         0.7 <int [2,000]>        1400
# 17     2000         0.8 <int [2,000]>        1600
# 18     2000         0.9 <int [2,000]>        1800

I have a function called simulate_samples. Each row of the above df defines one sensitivity. Given one sensitivity, simulate_sample generates a dataframe of 100 samples with a sample_id<int> and a list column for sample<list[int]>. One row to many. So I used reframe and got the behaviour like in the minimal reprex.

[DavisVaughan]

Do you just want this then?

starwars %>%
  rowwise(everything()) %>% 
  reframe(quantile_df(height)) %>%
  ungroup()

rowwise() works more like summarise() than like mutate()

[mutahiwachira]

That's perfect. I also read the docs and I see that the simulation case is mentioned.
Thanks so much this is very useful. Look forward to see what you guys do with by-row operations in future as they are very useful for simulations and nesting calculations without loops.

This solves my issue so I will close it.