roll_any() and roll_all()

[hadley]

Would you consider extending roll to also think about logical vectors? roll_any() and roll_all() are admittedly esoteric, but can sometimes be useful, and would complete the matrix of summary functions at https://github.com/r-lib/vctrs/issues/9

[jjf234]

Thanks, that's an interesting suggestion. Just so I understand what you’re thinking, here’s an example of how I’d approach it in base R:

library(roll)
library(xts)

n_vars <- 3
n_obs <- 15
x <- matrix(sample(c(TRUE, FALSE), (n_obs * n_vars), replace = TRUE),
            nrow = n_obs, ncol = n_vars)

rollapplyr(x, 5, any)
rollapplyr(x, 5, all)

If so, I can see it fitting within my framework and existing arguments: min_obs, complete_obs, na_restore. An exception is the weights argument but that wouldn't make sense with a logical matrix anyway. Will look into it further.

[hadley]

Yeah, exactly.

[jjf234]

I have added both roll_any and roll_all in the development version. Here's an example:

# devtools::install_github("jjf234/roll")

library(roll)
library(xts)
library(testthat)

n_vars <- 30
n_obs <- 1500
x <- matrix(rnorm(n_obs * n_vars), nrow = n_obs, ncol = n_vars)

# rolling any ('min_obs = 1' equivalent to 'partial = TRUE')
expect_equivalent(roll_any(x < 0, 5, min_obs = 1),
                  rollapplyr(x < 0, 5, any, partial = TRUE))

# rolling all
expect_equivalent(roll_all(x < 0, 5, min_obs = 1),
                  rollapplyr(x < 0, 5, all, partial = TRUE))

Also, the performance is significantly faster. Let me know if you have any other feedback.

[hadley]

Looks great, thanks!