block_caption() in markdown corrups Word files

[jniedballa]

Hello,
thank you for this very useful package.
After updating R and all packages recently, I found that reports I used to create successfully with markdown and officer create corrupted docx files (both with the CRAN and GitHub version of officer). Word reports the content is causing problems but does not offer to recover the content of the file:

It seems the issue is related to block_caption() used in the markdown:

fig_num <- run_autonum(seq_id = "Figure", pre_label = "Figure ", bkm = "fig1")
block_caption("Test caption",
              style = "Normal",
              autonum = fig_num)

This worked flawlessly in officer 0.5.2 (from R 4.2.1), but corrupts the docx in the CRAN version and GitHub versions of officer (from R 4.3.1 - see sessionInfo below).

Attached is a zip file containing the R markdown file, the correct docx output file from officer 0.5.2 and the corrupted docx from officer_0.6.3.002.

Is the usage of block_caption() in a markdown correct, or am I doing something wrong here? Just wondering since it worked in previous versions. I guess I can achieve the same look with fpar(), but the convenience of block_caption is great.

sessionInfo() for the correct docx file:

> sessionInfo()
R version 4.2.1 (2022-06-23 ucrt)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows 10 x64 (build 17763)

Matrix products: default

locale:
[1] LC_COLLATE=German_Germany.1252  LC_CTYPE=German_Germany.1252    LC_MONETARY=German_Germany.1252 LC_NUMERIC=C                   
[5] LC_TIME=German_Germany.1252    

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] officer_0.5.2    officedown_0.3.0

loaded via a namespace (and not attached):
  [1] sf_1.0-9                lubridate_1.9.2         fontquiver_0.2.1        RColorBrewer_1.1-3      tools_4.2.1             backports_1.4.1        
  [7] utf8_1.2.3              R6_2.5.1                rpart_4.1.16            KernSmooth_2.23-20      Hmisc_4.8-0             DBI_1.1.3              
 [13] colorspace_2.1-0        nnet_7.3-17             tidyselect_1.2.0        gridExtra_2.3           curl_5.0.0              compiler_4.2.1         
 [19] cli_3.4.1               htmlTable_2.4.1         xml2_1.3.3              fontBitstreamVera_0.1.1 scales_1.2.1            checkmate_2.1.0        
 [25] classInt_0.4-8          proxy_0.4-27            askpass_1.1             systemfonts_1.0.4       stringr_1.5.0           digest_0.6.31          
 [31] foreign_0.8-82          rmarkdown_2.21          gfonts_0.2.0            base64enc_0.1-3         jpeg_0.1-10             pkgconfig_2.0.3        
 [37] htmltools_0.5.4         fastmap_1.1.1           rvg_0.3.2               htmlwidgets_1.6.1       rlang_1.0.6             readxl_1.4.2           
 [43] rstudioapi_0.14         httpcode_0.3.0          shiny_1.7.4             generics_0.1.3          jsonlite_1.8.4          dplyr_1.1.0            
 [49] zip_2.2.2               magrittr_2.0.3          Formula_1.2-5           interp_1.1-3            Matrix_1.5-3            Rcpp_1.0.10            
 [55] munsell_0.5.0           fansi_1.0.4             gdtools_0.3.1           lifecycle_1.0.3         terra_1.7-3             stringi_1.7.12         
 [61] yaml_2.3.7              plyr_1.8.8              grid_4.2.1              promises_1.2.0.1        crayon_1.5.2            deldir_1.0-6           
 [67] lattice_0.20-45         splines_4.2.1           knitr_1.42              pillar_1.8.1            uuid_1.1-0              reshape2_1.4.4         
 [73] codetools_0.2-18        crul_1.3                glue_1.6.2              evaluate_0.20           fontLiberation_0.1.0    latticeExtra_0.6-30    
 [79] data.table_1.14.8       png_0.1-8               vctrs_0.5.2             httpuv_1.6.9            cellranger_1.1.0        gtable_0.3.1           
 [85] openssl_2.0.5           cachem_1.0.7            ggplot2_3.4.1           xfun_0.39               mime_0.12               xtable_1.8-4           
 [91] e1071_1.7-13            later_1.3.0             class_7.3-20            survival_3.3-1          tibble_3.1.8            memoise_2.0.1          
 [97] units_0.8-1             cluster_2.1.3           timechange_0.2.0        ellipsis_0.3.2         

sessionInfo for the corrupted docx:

> sessionInfo()
R version 4.3.1 (2023-06-16 ucrt)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows 10 x64 (build 17763)

Matrix products: default


locale:
[1] LC_COLLATE=German_Germany.1252  LC_CTYPE=German_Germany.1252    LC_MONETARY=German_Germany.1252 LC_NUMERIC=C                   
[5] LC_TIME=German_Germany.1252    

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] officer_0.6.3.002 officedown_0.3.0 

loaded via a namespace (and not attached):
 [1] jsonlite_1.8.7          compiler_4.3.1          crayon_1.5.2            promises_1.2.0.1        Rcpp_1.0.11             zip_2.3.0              
 [7] xml2_1.3.5              later_1.3.1             fontquiver_0.2.1        systemfonts_1.0.4       textshaping_0.3.6       uuid_1.1-0             
[13] yaml_2.3.7              fastmap_1.1.1           mime_0.12               R6_2.5.1                gfonts_0.2.0            gdtools_0.3.3          
[19] curl_5.0.1              knitr_1.43              openssl_2.0.6           crul_1.4.0              shiny_1.7.4.1           rlang_1.1.1            
[25] rvg_0.3.3               cachem_1.0.8            httpcode_0.3.0          httpuv_1.6.11           xfun_0.39               memoise_2.0.1          
[31] cli_3.6.1               magrittr_2.0.3          digest_0.6.33           xtable_1.8-4            rstudioapi_0.15.0       askpass_1.1            
[37] lifecycle_1.0.3         evaluate_0.21           glue_1.6.2              fontLiberation_0.1.0    ragg_1.2.5              fontBitstreamVera_0.1.1
[43] rmarkdown_2.23          ellipsis_0.3.2          tools_4.3.1             htmltools_0.5.5          

P.S.1.: I wanted to try this with officer 0.5.2 in R 4.3.1, but after installing officer 0.5.2 got the error:

>library(officedown)     # using 0.3.0
Error: package or namespace load failed for ‘officedown’ in loadNamespace(j <- i[[1L]], c(lib.loc, .libPaths()), versionCheck = vI[[j]]):
 namespace ‘officer’ 0.5.2 is already loaded, but >= 0.6.2 is required

This is a little confusing since according to CRAN officedown imports officer (≥ 0.5.1). R 4.2.1 did not throw that error.

P.S.2: there are typos in autonum here:
https://github.com/davidgohel/officer/blob/c0e15dc488c33268cdb8a8ef633ecd20f2370db2/R/ooxml_block_objects.R#L41

officer_problem.zip

[grcatlin]

+1 on this. I've found that an .rmd that otherwise knits to word just fine will throw the corrupt error message if I even have library(officedown) included at all (not using any functions from officedown that I'm aware of). In my perhaps related presentation, it seems that using data.table's walrus operator := anywhere in the document will corrupt the file unless I have include=FALSE set for the chunk with the := in it.

Here is a sample rmd that corrupts on my end:

---
output:
  word_document
---

{r include=FALSE}
# libraries
library(officer)
library(officedown)

{r echo=FALSE}
# get data
faithful <- datasets::faithful |> data.table::as.data.table()

# make some operation
faithful[, nonsense := eruptions + sqrt(waiting)]

# print
faithful |>
    head() |>
    print()

This yields a corrupted file. However, as soon as I comment out library(officedown), it renders fine. Alternatively, if I set include=FALSE and split the chunk in two, it will render with officedown loaded.

{r include=FALSE}
# libraries
library(officer)
library(officedown)

{r include=FALSE}
# get data
faithful <- datasets::faithful |> data.table::as.data.table()

# make some operation
faithful[, nonsense := eruptions + sqrt(waiting)]

{r echo=FALSE}
# print
faithful |>
    head() |>
    print()

I'm not sure why that would matter, but apparently it does. As a follow-up, I was able knit the .rmd @jniedballa attached just fine. Anyway, here's my sessionInfo() if that helps.

R version 4.3.1 (2023-06-16 ucrt)
Platform: x86_64-w64-mingw32/x64 (64-bit)  
Running under: Windows 10 x64 (build 19045)

Matrix products: default


locale:
[1] LC_COLLATE=en_US.UTF-8  LC_CTYPE=en_US.UTF-8
[3] LC_MONETARY=en_US.UTF-8 LC_NUMERIC=C
[5] LC_TIME=en_US.UTF-8
system code page: 1252

attached base packages:
[1] stats     graphics  grDevices utils     datasets
[6] methods   base

other attached packages:
[1] officer_0.6.2    officedown_0.3.1

loaded via a namespace (and not attached):
 [1] jsonlite_1.8.7          compiler_4.3.1
 [3] crayon_1.5.2            promises_1.2.1
 [5] Rcpp_1.0.11             zip_2.3.0
 [7] xml2_1.3.5              later_1.3.1
 [9] fontquiver_0.2.1        systemfonts_1.0.4
[11] textshaping_0.3.6       yaml_2.3.7
[13] uuid_1.1-1              fastmap_1.1.1
[15] mime_0.12               R6_2.5.1
[17] gfonts_0.2.0            gdtools_0.3.3
[19] curl_5.0.2              knitr_1.43
[21] openssl_2.1.0           crul_1.4.0
[23] shiny_1.7.5             rlang_1.1.1
[25] rvg_0.3.3               cachem_1.0.8
[27] httpcode_0.3.0          httpuv_1.6.11
[29] xfun_0.40               memoise_2.0.1
[31] cli_3.6.1               magrittr_2.0.3
[33] digest_0.6.33           xtable_1.8-4
[35] askpass_1.2.0           lifecycle_1.0.3
[37] evaluate_0.21           glue_1.6.2
[39] fontLiberation_0.1.0    ragg_1.2.5
[41] fontBitstreamVera_0.1.1 rmarkdown_2.24
[43] tools_4.3.1             ellipsis_0.3.2
[45] htmltools_0.5.6

[davidgohel]

This happens because you both are not using the correct format. To work with officedown, you have to use its format (note it will process additional things outputted by officer to original word_document).

You should set format as below:

output: officedown::rdocx_document

See https://github.com/davidgohel/officedown#getting-started and https://ardata-fr.github.io/officeverse/officedown-for-word.html

[grcatlin]

@davidgohel Ah, I see thank you! I've been messing with output to word just this week and your packages have been invaluable. Sorry for the error and thanks for all of your work! Using output: officedown::rdocx_document fixes the issue. I believe I was swapping between that and bookdown:word_document2 and forgot to change it back.

If I run this chunk:

{r echo=FALSE}
# get data
faithful <- datasets::faithful |> data.table::as.data.table()

# make some operation
faithful[, nonsense := eruptions + sqrt(waiting)]

# print
faithful |>
    head() |>
    print()

it actually formats faithful[, nonsense := eruptions + sqrt(waiting)] as a table which is not intended. So, I get the whole dataset formatted and the actual print() after a few pages. This is probably why using {include=FALSE} stopped the corruption even without output: officedown::rdocx_document. Anyway, not a huge issue now that I know about it. Thanks for these great packages!

[davidgohel]

You're welcome

If you are refering to this:

It is expected, that's one of the feature of officedown, it automatically transform printed data.frame to table.

[grcatlin]

@davidgohel Thanks for the information! I still don't think the line faithful[, nonsense := eruptions + sqrt(waiting)] should print as it is equivalent to

# make same operation
faithful <- faithful |>
    dplyr::mutate(nonsense = eruptions + sqrt(waiting))

As opposed to

# make same operation (not saved)
faithful |>
    dplyr::mutate(nonsense = eruptions + sqrt(waiting))

Hopefully this demonstrates what I mean:

# get data
faithful <- datasets::faithful |> data.table::as.data.table()

# make some operation
faithful[, nonsense := eruptions + sqrt(waiting)]

# same but in dplyr
faithful_2 <- datasets::faithful |>
    dplyr::mutate(nonsense = eruptions + sqrt(waiting))

# check equal
faithful <- as.data.frame(faithful)
all.equal(faithful, faithful_2)

I use data.table's pretty regularly but I know they are not as widely-used as more tidyverse-oriented options. As such, I don't think it's a huge deal that data.table's behave this way in officedown/officer. However, I believe this behavior is what triggered my word file corruption as I wasn't expecting any lines similar to faithful[, nonsense := eruptions + sqrt(waiting)] to be turned into tables.