table placement quarto-pdf

[bernardo-cruz]

I am trying to write my thesis using quarto. I have created a table as follows:

#| label: tbl-ohe_example
#| tbl-cap: One-hot vectorization example
#| tbl-pos: H

from IPython.display import display, Markdown, Latex, HTML, Image

Markdown(
    pd.DataFrame(
        data = {
            'Sentence': ['I like to eat apples', 'I like to eat bananas'],
            'Tokens' : [['I', 'like', 'to', 'eat', 'apples'], ['I', 'like', 'to', 'eat', 'bananas']],
            'Vector' : [[1, 1, 1, 1, 1, 0], [1, 1, 1, 1, 0, 1]],
        }
    )
   .to_html(index=False)
)

The table is perfectly rendered in the pdf - but placed in the top of the specific page of the document. The option tbl-pos: H seems not to function as fig-pos: H would.

I even tried to import \usepackage{float} - see yaml header:

---
...
    pdf:  
      include-in-header: 
        text: |
          \usepackage{float}
      papersize: a4paper
      df-print: paged
      fig-pos: 'H'
      output-file: "thesis"
      output-ext: "pdf"
...
---

Any hints how to avoid floating tables?

[rgaiacs]

@bernardo-cruz Can you add keep-tex: true to your _quarto.yaml file? You will have something like

---
...
    pdf:  
      papersize: a4paper
      keep-tex: true
      df-print: paged
      fig-pos: 'H'
      output-file: "thesis"
      output-ext: "pdf"
...
---

Inspect the .tex file will help to diagnose the problem as the LaTeX engine has a lot of rules for placement of float elements.

[bernardo-cruz]

@cscheid @rgaiacs thx very much for your answers - I am really grateful.

In summary, I cannot use the default quarto setup for python (pandas) tables, right? I understand the point about longtable etc, but I don't understand how to change that since I use quarto to create pandoc and pdfs. Where can I define if a table is longtable or not?

the code below yields to not output in the pdf

pd.DataFrame(
    data = {
        'Sentence': ['I like to eat apples', 'I like to eat bananas'],
        'Tokens' : [['I', 'like', 'to', 'eat', 'apples'], ['I', 'like', 'to', 'eat', 'bananas']],
        'Vector' : [[1, 1, 1, 1, 1, 0], [1, 1, 1, 1, 0, 1]],
    }
)
.to_latex(index=False)

[cscheid]

You need this:

---
format: latex
---

```{python}
import pandas as pd
```

```{python}
#| output: asis
#| warning: false
print(pd.DataFrame(
    data = {
        'Sentence': ['I like to eat apples', 'I like to eat bananas'],
        'Tokens' : [['I', 'like', 'to', 'eat', 'apples'], ['I', 'like', 'to', 'eat', 'bananas']],
        'Vector' : [[1, 1, 1, 1, 1, 0], [1, 1, 1, 1, 0, 1]],
    }
).to_latex(index=False))
```

[bernardo-cruz]

Thank you it works, but now the table is overflowing. Is there a reason why you have defined the format: latex?
My intention was to write a pdf document.

[cscheid]

Thank you it works, but now the table is overflowing.

You'll need to solve that by learning the pandas to_latex() API. Because of output: asis, quarto is not doing anything with that table. It's being emitted directly into the .tex source.

Is there a reason why you have defined the format: latex

Oh, sorry. The only difference between format: latex and format: pdf is whether we stop at the .tex file or we generate a .pdf from it.

[bernardo-cruz]

@cscheid Thx very much for all your explanations! 👍

[Jonasrg]

I tried the above answer suggested by @cscheid. The problem was that while it placed the table correctly, Quarto no longer added the table headings and labels. I found a workaround though. Given the following code, Quarto places the table correctly within the text (not on top of the page) and gives it the appropiate caption and label.

#| label: tbl-mytable
#| tbl-cap: This is awesome
#| tbl-pos: H

from IPython.display import display, Latex

display(Latex(df.to_latex(index=False)))

Hope that helps.

[grantmcdermott]

Relatedly: has anyone managed to make this work in a raw markdown table div? E.g.

See @tbl-ex for a demo.

| Default | Left | Right | Center |
|---------|:-----|------:|:------:|
| 12      | 12   |    12 |   12   |
| 123     | 123  |   123 |  123   |
| 1       | 1    |     1 |   1    |

: Demonstration of pipe table syntax {#tbl-ex tbl-pos="t"}

Doesn't work for me (the table still appears underneath the text rather than the top of the page.)

[mcanouil]

Did you look at the LaTeX code? keep-tex: true

[mcanouil]

I can't check right now but here is a test/example from the codebase:

https://github.com/quarto-dev/quarto-cli/blob/d86da0b118220ab8b9aeeb98e831abfc279b4073/tests/docs/smoke-all/2024/10/23/issue-10001.qmd

It might be a longtable (Pandoc) / table issue.

[mcanouil]

It seems tbl-pos is only respected when using layout.

• tbl-pos="t" is not respected

---
format: pdf
keep-tex: true
---

{{< lipsum 1 >}}


::: {#tbl-tm2 tbl-pos="t"}

| A | B | C |
|---|---|---|
| 1 | 2 | 3 |

Not so nice: Where Latex Wants it
:::

• tbl-pos="t" is respected

---
format: pdf
keep-tex: true
---

{{< lipsum 1 >}}


::: {#tbl-tm2 tbl-pos="t" layout-ncol="1"}

| A | B | C |
|---|---|---|
| 1 | 2 | 3 |

Not so nice: Where Latex Wants it
:::

That is because the table is actually encapsulated in a table/minipage via Quarto cross-reference div syntax.
This does not work via Pandoc syntax.

[grantmcdermott]

Thanks @mcanouil (you are a scholar and gent). I was slammed with deadlines at work yesterday, so couldn't test myself. I appreciate the layout workaround.