atorus-research / Tplyr

https://atorus-research.github.io/Tplyr/
Other
95 stars 17 forks source link

Update vignettes using options to reset options back to defaults to avoid CRAN inconsistencies #186

Open chelseadickens opened 2 months ago

chelseadickens commented 2 months ago

Prerequisites

For more information, see the CONTRIBUTING guide.

Description

The output I get by running the code below (copied directly from the Post-Processing vignette) does not produce the same output as the table under the "Collapsing Row Labels" section of https://cran.r-project.org/web/packages/Tplyr/vignettes/post_processing.html. Specifically, there are differences in the means and quartiles produced.

dat <- tplyr_table(tplyr_adsl, TRT01P) %>% 
    add_layer(
      group_count(RACE, by = vars("Race n (%)", SEX))
    ) %>% 
    add_layer(
      group_desc(AGE, by = vars("Age (years)", SEX))
    ) %>% 
    build()

 dat_show <- collapse_row_labels(dat, row_label1, row_label2, row_label3, indent = "&nbsp;&nbsp;") %>% 
    select(row_label, var1_Placebo)
 dat_show %>%  kable(escape=FALSE)

Steps to Reproduce (Bug Report Only)

  1. I checked to make sure the Tplyr version is the most recent (1.2.1) and that the data from the package matches what is under https://github.com/atorus-research/Tplyr/blob/main/data/tplyr_adsl.rda

  2. I tried calculating the means using dplyr tplyr_adsl %>% filter(TRT01P == "Placebo") %>% group_by(SEX) %>% summarize(mean = mean(AGE, na.rm = TRUE)) and get values that match the output I get from Tplyr (not those in the vignette). image

  3. When checking the Post-Processing article at https://atorus-research.github.io/Tplyr/articles/post_processing.html, the outputs match mine. So it seems perhaps the CRAN version of the vignette is out of date somehow?

Expected behavior: Screenshot from the vignette on CRAN: image

Actual behavior: image

Versions

image

mstackhouse commented 2 months ago

@chelseadickens well... This is incredibly weird. This is hard to decide where to even start looking.

mstackhouse commented 2 months ago

Well I can sleep tonight - this is actually a trickle effect from right here: https://github.com/atorus-research/Tplyr/blob/1fa6c4cd42da3434ee5118a2f61aed61b947dd48/vignettes/desc.Rmd#L178

The issue is that this option carries forward into subsequent vignettes. Apparently when you build a package, vignettes' R sessions are not truly independent.