On (1), duplicate(vec_slice(...)) appears stable across different machines using a complicated tibble. I also used the same tibble for benchmarks, and based on profiling, I don't think duplicate() will be a bottleneck except for the most extreme edge cases.

Details

library(targets)
library(tibble)

list(
  tar_target(x, {
    n <- 1e4
    tibble(
      id = seq_len(n),
      int1 = sample.int(1e6, n),
      int2 = sample.int(1e6, n),
      int3 = sample.int(1e6, n),
      dbl1 = rnorm(n),
      dbl2 = runif(n, -1e4, 1e4),
      dbl3 = rexp(n, 0.01),
      dbl4 = rbeta(n, 2, 5),
      dbl5 = rgamma(n, shape = 3, rate = 0.5),
      chr1 = paste0("item_", seq_len(n)),
      chr2 = sample(state.name, n, replace = TRUE),
      chr3 = format(Sys.time() + sample(-1e7:1e7, n), "%Y-%m-%d %H:%M:%S"),
      chr4 = replicate(
        n,
        paste0(sample(letters, 20, replace = TRUE), collapse = "")
      ),
      lgl1 = sample(c(TRUE, FALSE), n, replace = TRUE),
      lgl2 = runif(n) > 0.7,
      lgl3 = sample(
        c(TRUE, FALSE, NA),
        n,
        replace = TRUE,
        prob = c(0.4, 0.4, 0.2)
      ),
      fct1 = factor(sample(LETTERS[1:10], n, replace = TRUE)),
      fct2 = factor(
        sample(c("low", "med", "high"), n, replace = TRUE),
        levels = c("low", "med", "high"),
        ordered = TRUE
      ),
      date1 = as.Date("2020-01-01") + sample(0:1825, n, replace = TRUE),
      date2 = as.Date("2000-01-01") + sample(0:9125, n, replace = TRUE),
      posix1 = as.POSIXct("2020-01-01", tz = "UTC") + runif(n, 0, 1e8),
      cplx1 = complex(real = rnorm(n), imaginary = rnorm(n)),
      raw1 = as.raw(sample(0:255, n, replace = TRUE)),
      int_na = ifelse(
        runif(n) < 0.3,
        NA_integer_,
        sample.int(100, n, replace = TRUE)
      ),
      dbl_na = ifelse(runif(n) < 0.3, NA_real_, rnorm(n)),
      chr_na = ifelse(
        runif(n) < 0.3,
        NA_character_,
        sample(month.name, n, replace = TRUE)
      )
    )
  }),
  tar_target(y, x, pattern = map(x))
)

Failing tests on development R #1566

Description

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions