Dplyr verbs can strip the attributes (sdes and stype) from dtsurvey objects

[dcaseykc]

library('data.table')
library('dtsurvey')
library('dplyr')

a2 = data.frame(one = 1:26)
a = data.table(one = 1:26, two = letters)
loc = c(one = 1)
a[loc]
a2[loc]


b = dtsurvey::dtadmin(a)
dput(names(attributes(b)))

na = function(x) names(attributes(x))

# na(b %>% select(one))
options(datatable.verbose = TRUE)
distinct(b, one)
na(distinct(b, one))

[dcaseykc]

I think the approach will be to use select, capture the attributes, use NextMethod, and re-add the attributes

[dcaseykc]

Turns out it might be more simple: https://dplyr.tidyverse.org/reference/dplyr_extending.html

[dcaseykc]

@alastairmatheson : If you have some time, check out the https://github.com/PHSKC-APDE/dtsurvey/tree/dplyr_functionality and https://github.com/PHSKC-APDE/dtsurvey/blob/dplyr_functionality/inst/tinytest/test_dplyr_integration.R

Do these tests/use cases make sense?

[alastairmatheson]

I think so, but this you have greater expertise in S3 methods and package development than I do. One suggestion for the group_by error message (https://github.com/PHSKC-APDE/dtsurvey/blob/dplyr_functionality/R/dplyr_compatability.R#L83). If someone is trying to use group_by in the calc function, they probably actually want to use the by = '...' part of calc. Maybe change the error message to suggest that.