How to render two different tables with Elsevier extension in quarto?

[KipBalkcom-USDA]

Description

The code below uses the Elsevier journal extension (which must be installed) to create two tables. The first table created in a chunk labelled @tbl-data1 works as expected. The code for the second table labelled @tbl-data2 fails. The code for the second table is currently commented out using {{r}}. The document below will render unless the code for the second table is changed to{r}.

When I try to render both tables in the document, I get the following error:
ERROR:
compilation failed- error
Undefined control sequence.
l.207 \multirow
{2}{=}{\begin{minipage}[b]{\linewidth}\raggedright

The tables are made differently. Table 1 has one column under each header column. Table 2 has two columns under each header column, which seems to be causing the problem during rendering. Does anyone know a solution that will allow both tables to render?

---
title: Imaginary title
echo: false
warning: false
author:
  - name: FirstName LastName
    email: email@address
format:
  elsevier-pdf:
    keep-tex: true
---

```{r}
#| label: r-libraries

library(tidyverse)
library(gt)

```

```{r}
#| label: tbl-data1
#| tbl-cap: Table caption for data set one.

date_ops <- tibble::tibble(
    "Years" = c("Year1", "Year2"),
    "Measure 1" = c("11/12/2013", "12/13/2014"),
    "Measure 2" = c("12/13/2014", "01/02/2003")
) |>
    mutate(across(c(2:3), function(x)
        format(mdy(x), "%d %b %Y")))

date_ops_table <- date_ops |>
    gt() |>
    as_raw_html()

date_ops_table

```

```{{r}}
#| label: tbl-data2
#| tbl-cap: This table show information for two variables.

tbl_values <- tibble::tibble(
    "Effect" = c("Var1", "Var2"),
    "df" = c("3", "3"),
    "F.ratio" = c("3.456", "2.789"),
    "p.value" = c("0.16", "0.04")
) 

var1_data <- tbl_values |>
    filter(Effect == 'Var1') |>
    dplyr::select(-Effect)

var2_data <- tbl_values |>
    filter(Effect == 'Var2') |>
    dplyr::select(F.ratio, p.value)

df <- as_tibble(c(
    "Effect 1"
)) |>
    rename(Effect = value)

data_values <- cbind(df, var1_data, var2_data) |>
    janitor::clean_names() |>
    gt() |>
    tab_spanner(
        label = "Var1 Data",
        columns = c("f_ratio", "p_value")
    ) |>
    tab_spanner(
        label = "Var2 Data",
        columns = c("f_ratio_2", "p_value_2")
    ) |>
    cols_label(
        effect = "Effect",
        f_ratio = html("F-value"),
        p_value = html("<em>p</em>-value"),
        f_ratio_2 = html("F-value"),
        p_value_2 = html("<em>p</em>-value"),
    ) |>
    as_raw_html()

data_values

```

[mcanouil]

Could you remove pieces of your code one by one to end up with something small? Preferably something that does both require external dependencies such as janitor.
Either the error comes from those and they are unnecessary as they make your example harder and longer to investigate.

Also, please if we don't need the 100+ packages from the tidyverse, please don't include library(tidyverse) in your example. Focus on what is really needed. This will save you and us time to focus on the real issue.

I'm assuming the extension is the one that can be installed with quarto add quarto-journals/elsevier but as you did not give any details, hard to be sure about this.

Thank you for your help!

[mcanouil]

FWIW, I have no issues with the following code.
Either your code is producing bad code or your are using an outdated version of Quarto CLI.

---
title: Imaginary title
author:
  - name: FirstName LastName
    email: email@address
abstract: |
  {{< lipsum 1 >}}
format:
  elsevier-pdf: default
execute:
  echo: false
  warning: false
---

# Introduction

```{r}
#| label: r-libraries
library(gt)
data(penguins)
```

```{r}
#| label: tbl-data1
#| tbl-cap: Table caption for data set one.
#| tbl-pos: "!h"
gt(head(penguins, c(5L, 5L)))
```

```{r}
#| label: tbl-data2
#| tbl-cap: This table show information for two variables.
#| tbl-pos: "!h"
gt(head(penguins, c(5L, 5L)))
```

[KipBalkcom-USDA]

Yes, you were correct in assuming I was referring to the extension that can be installed with quarto add quarto-journals/elsevier

I took the example you provided and modified the code for the second table to create the output I was having problems with. The issue was with the html code contained in the second table of my original post. I added a comment to indicate the location in the code below.

I originally wrote the code to render as a docx before I realized there was an extension available. I used the html code to italicize some text in the table headers. If the html code is removed from the second table below, the document will render correctly, but it seems to create two other problems.

One, I still need to italicize some text in the table headers, but I don't know how to achieve that using the extension. Second, the second table is rendered wider than the page and I don't know if there are options to adjust the table and still be able to use the extension.

---
title: Imaginary title
author:
  - name: FirstName LastName
    email: email@address
abstract: |
  {{< lipsum 1 >}}
format:
  elsevier-pdf: default
execute:
  echo: false
  warning: false
---

# Introduction

```{r}
#| label: r-libraries
library(gt)
data(penguins)
```

```{r}
#| label: tbl-data1
#| tbl-cap: Table caption for data set one.
#| tbl-pos: "!h"
gt(head(penguins, c(5L, 5L)))
```

```{r}
#| label: tbl-data2
#| tbl-cap: This table shows information for two variables.
#| tbl-pos: "!h"

males <- penguins |> 
    dplyr::filter(island == 'Torgersen' & sex == 'male' & year == 2007) |> 
    head(c(4L, 4L))

females <- penguins |> 
    dplyr::filter(island == 'Torgersen' & sex == 'female' & year == 2007) |> 
    head(c(4L, 4L)) |> 
    dplyr::select(bill_len, bill_dep)  

first_four <- penguins |> 
    dplyr::filter(island == 'Torgersen' & year == 2007) |> 
    head(c(4L, 4L)) |> 
    dplyr::select(bill_len, bill_dep)  

peng_tbl <- cbind(males, females, first_four) |>
    janitor::clean_names() |> 
    gt() |>
    tab_spanner(label = 'Male', columns = c('bill_len', 'bill_dep')) |>
    tab_spanner(label = 'Female', columns = c('bill_len_2', 'bill_dep_2')) |>
    tab_spanner(label = 'M + F', columns = c('bill_len_3', 'bill_dep_3')) |>
    cols_label(
# note the html in the line below that is creating the problem
        bill_len_2 = html('<em>bill_len</em>'),
        bill_dep_2 = 'bill_dep',
        bill_len_3 = 'bill_len',
        bill_dep_3 = 'bill_dep') 

peng_tbl
   
```

[mcanouil]

Why overcomplicating the example?

Do as shown in gt, use Markdown and let gt/Quarto do the job.
What do you expect HTML to look like in LaTeX? LaTeX does not know how to handle HTML.

---
title: Imaginary title
author:
  - name: FirstName LastName
    email: email@address
abstract: |
  {{< lipsum 1 >}}
format:
  elsevier-pdf: default
execute:
  echo: false
  warning: false
---

# Introduction

```{r}
#| label: r-libraries
library(gt)
data(penguins)
```

```{r}
#| label: tbl-data1
#| tbl-cap: Table caption for data set one.
#| tbl-pos: "!h"
gt(head(penguins, c(5L, 5L)))
```

```{r}
#| label: tbl-data2
#| tbl-cap: This table show information for two variables.
#| tbl-pos: "!h"
gt(head(penguins, c(5L, 5L))) |>
  cols_label(species = md("*Species*"))
```

[cderv]

What do you expect HTML to look like in LaTeX? LaTeX does not know how to handle HTML.

To clarify and avoid any confusion, Quarto does parse HTML table to render a table in other format not supporting HTML. So it could be valid to output raw HTML from a cell and expect a table in PDF output. However, it would not be the same styled table as it would be processed through HTML -> Pandoc Table AST -> LaTeX conversion in Pandoc.

So in the examples shared above, if the cells are outputing HTML and there is a failure due to some LaTeX produced, this could be seen as an issue we need to handle in our HTML table processing.
And seeing the as_raw_html() in the example, I believe this is is.

However, gt can produce LaTeX directly, so it is probably indeed better to use LaTeX output, and use the markdown support from gt to let Quarto handle some markdown syntax processing.

I'll investigate to confirm the problem.

[mcanouil]

To be clear, from my example:

gt(head(penguins, c(5L, 5L))) |>
  cols_label(species = html("<em>Species</em>"))

produces a literal <em>Species</em> because gt does not emit HTML.

gt(head(penguins, c(5L, 5L))) |>
  cols_label(species = html("<em>Species</em>")) |>
  as_raw_html()

produces the italicise label (because of Quarto).

[cderv]

It was clear from your examples. Nothing to be added.

I added precision regarding original examples, as they have as_raw_html(). So their output should be parsed by Quarto and produce a table, or we should detect and error nicely if something is wrong with the parsing.

[KipBalkcom-USDA]

Thank you. As I stated, I was originally rendering to a docx, and I didn't realize the nuances between HTML and LaTeX when trying to create the pdf that required a modification. You have provided a very simple solution to the first problem I mentioned.

As for the complication of the example, I don't see how to make it simpler with my code experience because I need a table that is multiple columns wide or at least wider than the table you created in your example to illustrate how it does not fit the page. I'm sure there is a more elegant way to create the wide table, but nonetheless, I needed a wider table.

Landscape orientation could solve the problem, but this was not necessary when I originally rendered to a docx. I also tried tbl-colwidths but I'm sure I did not apply the option correctly.

---
title: Imaginary title
author:
  - name: FirstName LastName
    email: email@address
abstract: |
  {{< lipsum 1 >}}
format:
  elsevier-pdf: default
execute:
  echo: false
  warning: false
---

# Introduction

```{r}
#| label: r-libraries
library(gt)
data(penguins)
```

```{r}
#| label: tbl-data2
#| tbl-cap: This table shows information for two variables.
#| tbl-pos: "!h"

males <- penguins |> 
    dplyr::filter(island == 'Torgersen' & sex == 'male' & year == 2007) |> 
    head(c(4L, 4L))

females <- penguins |> 
    dplyr::filter(island == 'Torgersen' & sex == 'female' & year == 2007) |> 
    head(c(4L, 4L)) |> 
    dplyr::select(bill_len, bill_dep)  

first_four <- penguins |> 
    dplyr::filter(island == 'Torgersen' & year == 2007) |> 
    head(c(4L, 4L)) |> 
    dplyr::select(bill_len, bill_dep)  

peng_tbl <- cbind(males, females, first_four) |>
    janitor::clean_names() |> 
    gt() |>
    tab_spanner(label = 'Male', columns = c('bill_len', 'bill_dep')) |>
    tab_spanner(label = 'Female', columns = c('bill_len_2', 'bill_dep_2')) |>
    tab_spanner(label = 'M + F', columns = c('bill_len_3', 'bill_dep_3')) |>
    cols_label(
        bill_len_2 = md('*bill*_len'),
        bill_dep_2 = 'bill_dep',
        bill_len_3 = 'bill_len',
        bill_dep_3 = 'bill_dep') 

peng_tbl
   
```

[mcanouil]

When trying to figure out what is not working, it's no longer about your end goal.
You need to remove the pieces, one by one until the issue no longer appears.
Then put back the last piece you removed, if it triggers the issue again, then you are onto something.
After, you try to remove other pieces, etc.

In the end you get something small that exhibits the issue.

[mcanouil]

The following works: markdown + html + spanner.

So what version of Quarto CLI and gt / R are you using?

---
title: Imaginary title
author:
  - name: FirstName LastName
    email: email@address
abstract: |
  {{< lipsum 1 >}}
format:
  elsevier-pdf: default
execute:
  echo: false
  warning: false
---

# Introduction

```{r}
#| label: r-libraries
library(gt)
data(penguins)
```

```{r}
#| label: tbl-data2
#| tbl-cap: This table show information for two variables.
#| tbl-pos: "!h"
gt(head(penguins, c(5L, 5L))) |>
  tab_spanner(label = 'species_spanner', columns = c("species")) |>
  cols_label(
    island = html("<em>Island</em>"), # requires gt::as_raw_html() for Quarto to be able to process this into LaTeX/OpenXML, etc.
    bill_len = md("**Bill Length (mm)**")
  )
```

```{r}
#| label: tbl-data2
#| tbl-cap: This table show information for two variables.
#| tbl-pos: "!h"
gt(head(penguins, c(5L, 5L))) |>
  tab_spanner(label = 'species_spanner', columns = c("species")) |>
  cols_label(
    island = html("<em>Island</em>"), # requires gt::as_raw_html() for Quarto to be able to process this into LaTeX/OpenXML, etc.
  ) |>
  as_raw_html() # force Quarto to process the table as HTML because the table includes HTML
```

[KipBalkcom-USDA]

The previous example you provided will not work for me if I render the document. If I run each chunk individually, I can see the table output if I omit the as_raw_html()

Quarto 1.6.42

sessionInfo()
R version 4.5.0 (2025-04-11)
Platform: aarch64-apple-darwin20
Running under: macOS Sequoia 15.5

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

other attached packages:
[1] gt_1.0.0

loaded via a namespace (and not attached):
[1] vctrs_0.6.5 cli_3.6.5 knitr_1.50 rlang_1.1.6 xfun_0.52 stringi_1.8.7 generics_0.1.4
[8] litedown_0.7 glue_1.8.0 janitor_2.2.1 markdown_2.0 htmltools_0.5.8.1 sass_0.4.10 snakecase_0.11.1
[15] evaluate_1.0.3 tibble_3.2.1 fastmap_1.2.0 lifecycle_1.0.4 stringr_1.5.1 compiler_4.5.0 dplyr_1.1.4
[22] timechange_0.3.0 pkgconfig_2.0.3 rstudioapi_0.17.1 digest_0.6.37 R6_2.6.1 tidyselect_1.2.1 pillar_1.10.2
[29] commonmark_1.9.5 magrittr_2.0.3 tools_4.5.0 withr_3.0.2 lubridate_1.9.4 xml2_1.3.8

[mcanouil]

All my examples were executed and rendered fully using Quarto CLI. (I'm double checking them on a clean neutral environment)

Unfortunately, if you use an outdated version of Quarto CLI, there is not much we can do.

The current stable version is 1.7.32.

(I did not add screenshots because they bring almost nothing while they decrease accessibility)

[mcanouil]

Ok, It has actually nothing to do with the Elsevier extension, the use of Markdown or HTML inside the table.
The result is the same for both 1.7 and 1.8 (pre-release).
Three things are needed/required: tab_spanner, as_raw_html, and tbl- label then you get bad LaTeX.

---
format: pdf
---

```{r}
#| label: tbl-label-matters
library(gt)
data(penguins)
gt(head(penguins, c(5L, 5L))) |>
  tab_spanner(label = "species_spanner", columns = c("species")) |>
  as_raw_html()
```

(can't dive more into it now)

[KipBalkcom-USDA]

I thought you wanted me to run this and generate the log on my complex example from before. Code below generated the logs above.

---
title: Imaginary title
author:
  - name: FirstName LastName
    email: email@address
abstract: |
  {{< lipsum 1 >}}
format:
  elsevier-pdf: default
execute:
  echo: false
  warning: false
---

# Introduction

```{r}
#| label: r-libraries
library(gt)
data(penguins)
```

```{r}
#| label: tbl-data2
#| tbl-cap: This table shows information for two variables.
#| tbl-pos: "!h"

males <- penguins |> 
    dplyr::filter(island == 'Torgersen' & sex == 'male' & year == 2007) |> 
    head(c(4L, 4L))

females <- penguins |> 
    dplyr::filter(island == 'Torgersen' & sex == 'female' & year == 2007) |> 
    head(c(4L, 4L)) |> 
    dplyr::select(bill_len, bill_dep)  

first_four <- penguins |> 
    dplyr::filter(island == 'Torgersen' & year == 2007) |> 
    head(c(4L, 4L)) |> 
    dplyr::select(bill_len, bill_dep)  

peng_tbl <- cbind(males, females, first_four) |>
    janitor::clean_names() |> 
    gt() |>
    tab_spanner(label = 'Male', columns = c('bill_len', 'bill_dep')) |>
    tab_spanner(label = 'Female', columns = c('bill_len_2', 'bill_dep_2')) |>
    tab_spanner(label = 'M + F', columns = c('bill_len_3', 'bill_dep_3')) |>
    cols_label(
        bill_len_2 = html('<em>bill</em>_len'),
        bill_dep_2 = 'bill_dep',
        bill_len_3 = 'bill_len',
        bill_dep_3 = 'bill_dep') |> 
    as_raw_html()

peng_tbl
   
```

[mcanouil]

Bare with me, context matters. If you drop a log without giving the context it was produced just after saying "it works", expecting a reader to not link the two is unlikely.

Try to remove pieces of your example, for instance start with format: pdf as the YAML like in my example without the Elsevier extension, etc.
Does it work? or not.
Remove, one, two, three spanner. Does it work?
etc.

edit: I built 1.6 and my example works there. It seems to be a parsing issue of the HTML table that shift.

[KipBalkcom-USDA]

For the version I have (1.6.42), it does not appear that tab_spanner has any effect, but html() and/or as_html_raw() does. As you have already shown, md() can create the necessary emphasis.

cols_label(bill_len_2 = html('<em>bill_len</em>')) |> as_raw_html()

or

cols_label(bill_len_2 = html('<em>bill_len</em>'))

does not work.

---
format: pdf
---

# Introduction

```{r}
#| label: r-libraries
library(gt)
data(penguins)
```

```{r}
#| label: tbl-data2
#| tbl-cap: This table shows information for two variables.
#| tbl-pos: "!h"

males <- penguins |> 
    dplyr::filter(island == 'Torgersen' & sex == 'male' & year == 2007) |> 
    head(c(4L, 4L))

females <- penguins |> 
    dplyr::filter(island == 'Torgersen' & sex == 'female' & year == 2007) |> 
    head(c(4L, 4L)) |> 
    dplyr::select(bill_len, bill_dep)  

first_four <- penguins |> 
    dplyr::filter(island == 'Torgersen' & year == 2007) |> 
    head(c(4L, 4L)) |> 
    dplyr::select(bill_len, bill_dep)  

peng_tbl <- cbind(males, females, first_four) |>
    janitor::clean_names() |> 
    gt() |>
    tab_spanner(label = 'Male', columns = c('bill_len', 'bill_dep')) |>
    tab_spanner(label = 'Female', columns = c('bill_len_2', 'bill_dep_2')) |>
    tab_spanner(label = 'M + F', columns = c('bill_len_3', 'bill_dep_3')) |>
    cols_label(
        bill_len_2 = md('*bill_len*')) 

peng_tbl
   
```

[mcanouil]

As your example is still far too big despite your claim about cols_label, I took your original example, installed everything and remove each line one by one.
The only thing missing to make 1.6 crash was the caption:

---
format: pdf
---

```{r}
#| label: tbl-label-matters
#| tbl-cap: this will crash 1.6 in addition to all previous requirements already listed
library(gt)
gt(head(penguins, c(5L, 5L))) |>
  tab_spanner(label = "species_spanner", columns = "species") |>
  as_raw_html()
```

For 1.7/1.8:

• label using the reserved prefix for cross-reference tbl-
• gt::tab_spanner()
• HTML output

For 1.6:

• add a caption

The following will be using the latest version of Quarto CLI.
Using the cross-reference div syntax does not work around the issue.

---
format: pdf
---

:::: {#tbl-label-matters}

```{r}
library(gt)
gt(head(penguins, c(5L, 5L))) |>
  tab_spanner(label = "species_spanner", columns = "species") |>
  as_raw_html()
```

This will crash 1.6 in addition to all previous requirements already listed

:::

Looking at the intermediate markdown file which contain the HTML code for the table, trimming the table, I end up with the following that triggers the issue:

---
format: pdf
---

::: {#tbl-label-matters tbl-cap='this will crash 1.6 in addition to all previous requirements already listed'}

```{=html}
<table>
  <thead>
    <tr>
      <th>species_spanner</th>
      <th rowspan="2">island</th>
    </tr>
    <tr>
      <th rowspan="1">species</th>
    </tr>
  </thead>
</table>
```

:::

You can even go deeper with:

---
format: pdf
---

::: {#tbl-label tbl-cap='Caption'}

```{=html}
<table>
  <thead>
    <tr><th rowspan="2">island</th></tr>
    <tr></tr>
  </thead>
</table>
```

:::

From this, to avoid LaTeX crash, do any:

• remove tbl-cap in 1.6 (it does not crash anyway in 1.7 and 1.8)
• remove the cross-reference tbl-
• change to rowspan="1"
• remove one of the two rows in the header, *e.g., <tr></tr>

[cderv]

Thank you for trimming down to nice reprex !

[mcanouil]

TL;DR from #13084 (reply in thread)

---
format: pdf
---

::: {#tbl-label tbl-cap='Caption'}

```{=html}
<table>
  <thead>
    <tr><th rowspan="2">island</th></tr>
    <tr></tr>
  </thead>
</table>
```

:::

From this, to avoid LaTeX crash, do any:

• remove tbl-cap in 1.6 (it does not crash anyway in 1.7 and 1.8)
• remove the cross-reference tbl-
• change to rowspan="1"
• remove one of the two rows in the header, *e.g., <tr></tr>

[mcanouil]

cc @cderv here is the "summary" if you want to take over.

[cderv]

Yes, Thank you for trimming down to nice reprex !

This is a bug to me: HTML table will be parsed by Quarto, and this will create some LaTeX in output, that seems to be invalid when used in crossref. If this is expected to not work, we should probably detect, and stop early.

We do have support for rowspan, seeing the feature format matrix: https://github.com/quarto-dev/quarto-cli/blob/2f247767e6aa2eb62a4871c4e7ccff3ac1ac3641/dev-docs/feature-format-matrix/qmd-files/raw-html/cell/rowspan/document.qmd

So the issue is really when trying to reference the table.

[cderv]

So the issue is that \multirow is used, but somehow, \usepackage{multirow} is missing when

::: {#tbl-test}

```{=html}
<table>
  <thead>
    <tr>
      <th>species_spanner</th>
      <th rowspan="2">island</th>
    </tr>
    <tr>
      <th rowspan="1">species</th>
    </tr>
  </thead>
</table>
```

:::

but not if the table is not crossreferenced

```{=html}
<table>
  <thead>
    <tr>
      <th>species_spanner</th>
      <th rowspan="2">island</th>
    </tr>
    <tr>
      <th rowspan="1">species</th>
    </tr>
  </thead>
</table>
```

So probably related to multirow variable not set in one of the scenario

quarto-cli/src/resources/formats/pdf/pandoc/tables.tex

Lines 6 to 8 in 2f24776

	$if(multirow)$
	\usepackage{multirow}
	$endif$

[cderv]

• HTML Table in PDF - \multirow is used but somehow, \usepackage{multirow} is missing when table is crossreferenced #13093