dplyr-sample.md
In duckplyr: A 'DuckDB'-Backed Version of 'dplyr'

sample_*() gives meaningful error messages

Code
  df2 <- tibble(x = rep(1:2, 100), y = rep(c(0, 1), 100), g = rep(1:2, each = 100))
  grp <- df2 %>% duckplyr_group_by(g)
  (expect_error(sample_n(grp, nrow(df2) / 2, weight = y)))
Output
  <error/rlang_error>
  Error in `sample_n()`:
  ! Can't compute indices.
  i In group 1: `g = 1`.
  Caused by error in `sample.int()`:
  ! too few positive probabilities
Code
  (expect_error(sample_frac(grp, 1, weight = y)))
Output
  <error/rlang_error>
  Error in `sample_frac()`:
  ! Can't compute indices.
  i In group 1: `g = 1`.
  Caused by error in `sample.int()`:
  ! too few positive probabilities
Code
  (expect_error(mtcars %>% duckplyr_group_by(cyl) %>% sample_n(10)))
Output
  <error/rlang_error>
  Error in `sample_n()`:
  ! Can't compute indices.
  i In group 2: `cyl = 6`.
  Caused by error:
  ! `size` must be less than or equal to 7 (size of data).
  i set `replace = TRUE` to use sampling with replacement.
Code
  (expect_error(sample_n(list())))
Output
  <error/rlang_error>
  Error in `sample_n()`:
  ! `tbl` must be a data frame, not an empty list.
Code
  (expect_error(sample_frac(list())))
Output
  <error/rlang_error>
  Error in `sample_frac()`:
  ! `tbl` must be a data frame, not an empty list.
Code
  # # respects weight
  df <- data.frame(x = 1:2, y = c(0, 1))
  (expect_error(sample_n(df, 2, weight = y)))
Output
  <error/rlang_error>
  Error in `sample_n()`:
  ! Can't compute indices.
  Caused by error in `sample.int()`:
  ! too few positive probabilities
Code
  (expect_error(sample_frac(df, 2)))
Output
  <error/rlang_error>
  Error in `sample_frac()`:
  ! Can't compute indices.
  Caused by error:
  ! `size` of sampled fraction must be less or equal to one.
  i set `replace = TRUE` to use sampling with replacement.
Code
  (expect_error(sample_frac(df %>% duckplyr_group_by(y), 2)))
Output
  <error/rlang_error>
  Error in `sample_frac()`:
  ! Can't compute indices.
  i In group 1: `y = 0`.
  Caused by error:
  ! `size` of sampled fraction must be less or equal to one.
  i set `replace = TRUE` to use sampling with replacement.
Code
  (expect_error(sample_frac(df, 1, weight = y)))
Output
  <error/rlang_error>
  Error in `sample_frac()`:
  ! Can't compute indices.
  Caused by error in `sample.int()`:
  ! too few positive probabilities