sjPlot::tab_model() file parameter ignored if function runs inside a targets pipeline

[gorkang]

Prework

• [ x ] Read and agree to the code of conduct and contributing guidelines.
• [ x ] Confirm that your issue is most likely a genuine bug in targets and not a known limitation or usage error.
• [ x ] If there is already a relevant issue, whether open or closed, comment on the existing thread instead of posting a new issue.
• [ x ] Post a minimal reproducible example like this one so the maintainer can troubleshoot the problems you identify. A reproducible example is:
  • [ x ] Runnable: post enough R code and data so any onlooker can create the error on their own computer.
  • [ x ] Minimal: reduce runtime wherever possible and remove complicated details that are irrelevant to the issue at hand.
  • [ x ] Readable: format your code according to the tidyverse style guide.

Description

The sjPlot::tab_model() function has a file parameter that allows you to save a nice .html table of your model results. But when running it inside a {targets} pipeline, the file parameter is ignored and no file is saved.

The saving behavior is triggered when we do tar_read(table_save) or tar_load(table_save); table_save.

I would expect the file to be created when running the pipeline with tar_make() with no further steps.

Please, let me know if you think I should report this in the {sjPlot} package instead of here.

Below a reproducible example. Given it involves creating files, I could not manage to do a proper reprex.

Reproducible example

This is the standard (expected) behavior for the table_save target. You input a model, and if the file parameter is there, it creates a file:

sjPlot::tab_model(lme4::lmer(year ~ displ + (1|manufacturer), data = ggplot2::mpg), file = "TABLE_standard.html")
file.exists("TABLE_standard.html")
# [1] TRUE
file.remove("TABLE_standard.html")
# [1] TRUE

But if the sjPlot::tab_model(model, file = "TABLE_targets.html") code runs inside {targets}, the file parameter is ignored.

library(targets)

tar_script({
  
  library(targets)
  options(crayon.enabled = FALSE)
  tar_option_set(packages = c("targets", "ggplot2", "lme4", "sjPlot"))

  list(
    tar_target(model, lme4::lmer(year ~ displ + (1|manufacturer), data = ggplot2::mpg)),
    tar_target(table_save, sjPlot::tab_model(model, file = "TABLE_targets.html"))
  )
})

tar_make()

file.exists("TABLE_targets.html") # File not saved
# [1] FALSE
tar_load(table_save); length(table_save) # Although table_save is there
# [1] 6
file.exists("TABLE_targets.html") # Just to be sure, file still not there
# [1] FALSE

But, if we take a look, the file is created

table_save
file.exists("TABLE_targets.html")
# [1] TRUE
file.remove("TABLE_targets.html")
# [1] TRUE

Also, if we read it, it is saved

tar_read(table_save)
file.exists("TABLE_targets.html")
# [1] TRUE
file.remove("TABLE_targets.html")
# [1] TRUE

Diagnostic information

> sessionInfo()
R version 4.0.3 (2020-10-10)
Platform: x86_64-pc-linux-gnu (64-bit)
Running under: Ubuntu 20.04.1 LTS

Matrix products: default
BLAS:   /usr/lib/x86_64-linux-gnu/openblas-pthread/libblas.so.3
LAPACK: /usr/lib/x86_64-linux-gnu/openblas-pthread/liblapack.so.3

locale:
 [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C               LC_TIME=es_CL.UTF-8        LC_COLLATE=en_US.UTF-8     LC_MONETARY=es_CL.UTF-8    LC_MESSAGES=en_US.UTF-8   
 [7] LC_PAPER=es_CL.UTF-8       LC_NAME=C                  LC_ADDRESS=C               LC_TELEPHONE=C             LC_MEASUREMENT=es_CL.UTF-8 LC_IDENTIFICATION=C       

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

other attached packages:
[1] targets_0.0.2.9000

loaded via a namespace (and not attached):
 [1] Rcpp_1.0.6          mvtnorm_1.1-1       lattice_0.20-41     tidyr_1.1.2         ps_1.5.0            zoo_1.8-8           assertthat_0.2.1    digest_0.6.27      
 [9] R6_2.5.0            backports_1.2.1     reprex_1.0.0        evaluate_0.14       coda_0.19-4         highr_0.8           ggplot2_3.3.3       pillar_1.4.7       
[17] rlang_0.4.10        multcomp_1.4-15     rstudioapi_0.13     minqa_1.2.4         data.table_1.13.6   performance_0.6.1.1 callr_3.5.1         nloptr_1.2.2.2     
[25] Matrix_1.3-2        effectsize_0.4.3    rmarkdown_2.6       styler_1.3.2        ggeffects_1.0.1     splines_4.0.3       lme4_1.1-26         statmod_1.4.35     
[33] stringr_1.4.0       igraph_1.2.6        munsell_0.5.0       sjPlot_2.8.7        broom_0.7.3         compiler_4.0.3      modelr_0.1.8        xfun_0.20          
[41] pkgconfig_2.0.3     parameters_0.11.0   clipr_0.7.1         htmltools_0.5.1.1   insight_0.12.0      tidyselect_1.1.0    tibble_3.0.5        codetools_0.2-18   
[49] fansi_0.4.2         withr_2.4.1         crayon_1.3.4        dplyr_1.0.3         MASS_7.3-53         sjmisc_2.8.6        grid_4.0.3          nlme_3.1-151       
[57] xtable_1.8-4        gtable_0.3.0        lifecycle_0.2.0     DBI_1.1.1           magrittr_2.0.1      bayestestR_0.8.2    scales_1.1.1        stringi_1.5.3      
[65] estimability_1.3    cli_2.2.0           fs_1.5.0            snakecase_0.11.0    ellipsis_0.3.1      generics_0.1.0      vctrs_0.3.6         boot_1.3-26        
[73] sandwich_3.0-0      sjlabelled_1.1.7    TH.data_1.0-10      tools_4.0.3         glue_1.4.2          purrr_0.3.4         sjstats_0.18.1      emmeans_1.5.3      
[81] processx_3.4.5      survival_3.2-7      yaml_2.2.1          colorspace_2.0-0    knitr_1.31         

[wlandau]

Oddly enough, sjPlot::tab_model() does not actually write a file. Instead, it returns an sjPlot object whose print() method actually does the writing. When the table runs in a target, it skips the interactive print-out and thus no file is written.

Personally, I find this behavior misleading and recommend filing an issue in the sjPlot package.

> sjPlot:::print.sjTable
function (x, ...) 
{
    if (!is.null(x$file)) {
        write(x$knitr, file = x$file)
    }
    else {
        x$page.complete <- replace_umlauts(x$page.complete)
        htmlFile <- tempfile(fileext = ".html")
        write(x$page.complete, file = htmlFile)
        viewer <- getOption("viewer")
        if (x$viewer && !is.null(viewer)) {
            viewer(htmlFile)
        }
        else {
            utils::browseURL(htmlFile)
        }
    }
}
<bytecode: 0x7fc28ce72df8>
<environment: namespace:sjPlot>

As a workaround, you can call the print method yourself in the target. If you want the file, I also recommend setting format = "file" so the output file is tracked for changes and automatically rebuilt if corrupted.

library(targets)
tar_script({
  options(crayon.enabled = FALSE)
  tar_option_set(
    packages = c("targets", "ggplot2", "lme4", "sjPlot")
  )
  
  save_table <- function(model, path) {
    print(sjPlot::tab_model(model, file = path))
    path
  }
  
  list(
    tar_target(model, lme4::lmer(year ~ displ + (1|manufacturer), data = ggplot2::mpg)),
    tar_target(table_save, save_table(model, "your_table.html"), format = "file")
  )
})

tar_make()
#> ● run target model
#> boundary (singular) fit: see ?isSingular
#> ● run target table_save
#> ● end pipeline

webshot::webshot("your_table.html")

^{Created on 2021-01-28 by the reprex package (v0.3.0)}