Reproduce any tricks used in native pandoc table in the example of inserting midrules in LaTeX

[pbsds]

Hi, i'm touching up my thesis and could use some midrules in my tables.
I found that simply inserting them in standard markdown tables works fine, but not with pantable

example:

|1 | 1|1 |
|--|--|--|
|1 | 1|1 |
|1 | 1|1 |
|\midrule 1 | 1|1 |
|1 | 1|1 |


```table
---
markdown: false
---
1,1,1,1,1
1,1,1,1,1
1,1,1,1,1
\midrule 1,1,1,1,1
1,1,1,1,1
1,1,1,1,1
```

```table
---
markdown: true
---
1,1,1,1,1
1,1,1,1,1
1,1,1,1,1
\midrule 1,1,1,1,1
1,1,1,1,1
1,1,1,1,1
```

command:

pandoc test.md -F pantable -t latex

output:

\begin{longtable}[]{@{}lll@{}}
\toprule
1 & 1 & 1 \\
\midrule
\endhead
1 & 1 & 1 \\
1 & 1 & 1 \\
\midrule 1 & 1 & 1 \\
1 & 1 & 1 \\
\bottomrule
\end{longtable}

\begin{longtable}[]{@{}lllll@{}}
\toprule
1 & 1 & 1 & 1 & 1 \\
\midrule
\endhead
1 & 1 & 1 & 1 & 1 \\
1 & 1 & 1 & 1 & 1 \\
\textbackslash midrule 1 & 1 & 1 & 1 & 1 \\
1 & 1 & 1 & 1 & 1 \\
1 & 1 & 1 & 1 & 1 \\
\bottomrule
\end{longtable}

\begin{longtable}[]{@{}
  >{\raggedright\arraybackslash}p{(\columnwidth - 8\tabcolsep) * \real{0.20}}
  >{\raggedright\arraybackslash}p{(\columnwidth - 8\tabcolsep) * \real{0.20}}
  >{\raggedright\arraybackslash}p{(\columnwidth - 8\tabcolsep) * \real{0.20}}
  >{\raggedright\arraybackslash}p{(\columnwidth - 8\tabcolsep) * \real{0.20}}
  >{\raggedright\arraybackslash}p{(\columnwidth - 8\tabcolsep) * \real{0.20}}@{}}
\toprule
1 & 1 & 1 & 1 & 1 \\
\midrule
\endhead
1 & 1 & 1 & 1 & 1 \\
1 & 1 & 1 & 1 & 1 \\
\begin{minipage}[t]{\linewidth}\raggedright
\midrule 1
\end{minipage} & 1 & 1 & 1 & 1 \\
1 & 1 & 1 & 1 & 1 \\
1 & 1 & 1 & 1 & 1 \\
\bottomrule
\end{longtable}

My use-case is markdown: true. The minipage block is likely added here.
Is it possible to add a fix or workaround for this?

[ickc]

It can be done: I turned this into a discussion and show you how it find out how to do this.

First, say in ex1.md

| 1	| 2	| 3	|  
| --	| --	| --	|  
| 4	| 5	| 6	|  
| 7	| 8	| 9	|  
| \midrule 10	| 11	| 12	|  

assuming this native pandoc table can achieve what you want to do (in this case inject a certain TeX command to the LaTeX output.)

To see what should be written as a pantable, use

pandoc ex1.md -F pantable2csv -o ex2.md

where ex2.md now has

``` {.table}
---
markdown: true
...
1,2,3
4,5,6
7,8,9
`\midrule 10`{=tex},11,12
```

So you see the key idea is to mark explicitly the \midrule as raw tex.

To see that ex1.md and ex2.md produces identical native AST:

comm -3 <(pandoc ex1.md -F pantable2csv -t native -F pantable) <(pandoc ex1.md -t native)

as it shows no difference, it means it has identical AST. qed.

Edit:

A few interesting points:

Look at the native of

``` {.table}
---
markdown: true
...
1,2,3
4,5,6
7,8,9
\midrule 10,11,12
```

The problem is that the cell \midrule 10, when converted to AST, is interpreted as RawBlock, instead of RawInline. As it is a block level element, pandoc inserted a minipage to handle that. Normally putting them inside a minipage is no big deal, but since this hack requires the \midrule to be interpreted as part of the table, it fails.

Another worth noting thing is, it would be tempted to think the more native way to produce a table like that in pantable is

``` {.table}
---
markdown: true
ms:
    - 1
    - 0
    - 2
    - 1
...
1,2,3
4,5,6
7,8,9
10,11,12
```

But pandoc currently does not put draw a midrule before the foot row. This is an upstream issue and is subjective to say which is right.