Should we save tables in reports as separate files (similar to what is done for figures)?

[kelliemac]

Should we save tables in reports as separate files (similar to what is done for figures)? Maybe in csv/xlsx format? This can be helpful for sharing results with collaborators.

I am not sure if this is a good idea or not in general, so let's discuss. Does anyone else get asked to share individual tables like this, so that collaborators can annotate them as they want to?

[mayerbry]

I didn't see this before. I had to do this for a lot of tables in one project, so I developed this ad hoc procedure:

# top-level functions
save_table = function(df, fname){
  write_csv(df, paste0("local-dir-tables/", fname, ".csv"))
}
save_caption = function(cap, fname){
  cat(cap, file = paste0("local-dir-tables/", fname, "-cap.txt"))
}

# later in report, in actual table chunk

caption <- "caption for my table."
save_caption(caption, "name-of-table")

tbl_data %>%
  save_table("name-of-table") %>%
  table(
    format = output_type,
    digits = 2,
    booktabs = TRUE,
    caption = caption,
   ...
)

[kelliemac]

Thanks Bryan, this is great! I wasn't sure how useful this feature would be, but you've provided one good use case. Do we think other people need to do this regularly enough that it's worth including a version of this function in VISCtemplates?

I also wonder if there's a way to save the caption in the same file as the table. Maybe a hacky way to do it would be to write the caption as the first or last line in the csv.

[mayerbry]

The complete use case was compiling word doc ready tables for collaborators for a manuscript. This was the full workflow, but I would love a more efficient way to do this:

1. save tables + captions within each report subdirectory (as per last comment)
2. within a manuscript subdirectory had an Rmd that made one html document with every saved table compiled together
   • The Rmd re-used code from the reports to generate publication ready looking tables
3. copy and pasted the full html into a word doc
   • There is a quick manual step within the word doc to make tables look rendered correctly

This was a bit cumbersome in some ways, (and there is an obvious violation of DRY with code re-use to actually re-generate the tables), but at least I was sure the final tables shared the same source data as the reports.

Note: Rmd->doc couldn't re-generate the tables as they needed to be for the publication. I.e., these word tables looked terrible and required a lot of manual formatting in word to fix them. If there is a good kable equivalent for making word tables, that might be simpler and resolve your question re. saving tables and captions together.

source: https://github.com/FredHutch/ID50ZavalaAnalysis/tree/main/manuscript

[kelliemac]

It looks like the flextable package might help with saving nicely formatted tables directly to a Word doc (source: https://community.rstudio.com/t/nice-tables-when-knitting-to-word/3840). We could consider writing a VISCtemplates function to render tables that uses kable + kableExtra for pdf/html and flextable for word outputs. Maybe this is more useful than saving tables to csv.

The issue of code re-use and table/figure re-generation for manuscripts is I think a slightly separate, and important, one. I don't have a good solution to this, and I don't have a good sense of how universal the manuscript preparation process is across different projects, since I haven't been involved in very many.

[kelliemac]

Current thinking on this issue: we should just improve the table formatting in the docx report output, and then it's easy for people to pull desired tables from there. See https://github.com/FredHutch/VISCtemplates/issues/120

[kelliemac]

closing this issue as it is covered by others regarding improving table formatting in word docs