qmd writer turns a loose bullet list into a tight one when items contain a nested sublist

[rundel]

A BulletList parsed as loose (each item's first block is Paragraph) round-trips as tight (Plain) when at least one item contains a nested sublist. The writer emits blank-line separators between top-level items but the re-parser does not treat the resulting source as loose, so the outer list collapses. Symptom: items[[i]][[1]]$tag = "Plain" after the round trip where the original was "Paragraph".

$ printf -- '-   a\n\n    -   c\n    -   d\n\n-   b\n' 
-   a

    -   c
    -   d

-   b

$ printf -- '-   a\n\n    -   c\n    -   d\n\n-   b\n' | cargo run --bin pampa --
[ BulletList [[Para [Str "a"], BulletList [[Plain [Str "c"]], [Plain [Str "d"]]]], [Para [Str "b"]]] ]

$ printf -- '-   a\n\n    -   c\n    -   d\n\n-   b\n' | cargo run --bin pampa -- -t qmd
* a
  
  * c
  * d

* b

$ printf -- '-   a\n\n    -   c\n    -   d\n\n-   b\n' | cargo run --bin pampa -- -t qmd | cargo run --bin pampa --
[ BulletList [[Plain [Str "a"], BulletList [[Plain [Str "c"]], [Plain [Str "d"]]]], [Plain [Str "b"]]] ]

Examples in quarto-web:

• https://github.com/quarto-dev/quarto-web/blob/8b1be24df1c9301f2700c77e41e7a9f789535717/docs/blog/posts/2024-04-01-manuscripts-rmedicine/index.qmd#L19
• https://github.com/quarto-dev/quarto-web/blob/8b1be24df1c9301f2700c77e41e7a9f789535717/docs/authoring/_shortcodes.qmd#L30
• https://github.com/quarto-dev/quarto-web/blob/8b1be24df1c9301f2700c77e41e7a9f789535717/docs/journals/authors.qmd#L142

[cscheid]

Thanks for the report! This is going to be a tricky one to solve in general, I think. We end up having to define loose vs tight lists pretty poorly in order to support both of them in our parser, and I don't have a good description of the implied spec for "what's a tight list" the way we do it.

I remember thinking at the time that I didn't believe Pandoc handled it particularly consistently, but now is the time for us to do it right.

[rundel]

Another seemingly related issue that cropped up - When a list item contains a code block (or any non-leading block) separated from the leading paragraph by a blank line, the original parses as a loose list (Para per item). The writer emits the inter-block separator as an indented line of trailing whitespace instead of a true blank line; on re-parse the list is read as tight (Plain per item).

$ printf '1.  First item.\n\n    ```r\n    x <- 1\n    ```\n\n2.  Second item.\n'
1.  First item.

    ```r
    x <- 1
    ```

2.  Second item.

$ printf '1.  First item.\n\n    ```r\n    x <- 1\n    ```\n\n2.  Second item.\n' | cargo run --bin pampa --
[ OrderedList (1, Decimal, Period) [[Para [Str "First", Space, Str "item."], CodeBlock ( "" , ["r"] , [] ) "x <- 1"], [Para [Str "Second", Space, Str "item."]]] ]

$ printf '1.  First item.\n\n    ```r\n    x <- 1\n    ```\n\n2.  Second item.\n' | cargo run --bin pampa -- -t qmd
1.  First item.
    
    ```r
    x <- 1
    ```

2.  Second item.

$ printf '1.  First item.\n\n    ```r\n    x <- 1\n    ```\n\n2.  Second item.\n' | cargo run --bin pampa -- -t qmd | cargo run --bin pampa --
[ OrderedList (1, Decimal, Period) [[Plain [Str "First", Space, Str "item."], CodeBlock ( "" , ["r"] , [] ) "x <- 1"], [Plain [Str "Second", Space, Str "item."]]] ]

From quarto-web:

• https://github.com/quarto-dev/quarto-web/blob/8b1be24df1c9301f2700c77e41e7a9f789535717/docs/extensions/_shortcode-escaping.qmd#L5
• https://github.com/quarto-dev/quarto-web/blob/8b1be24df1c9301f2700c77e41e7a9f789535717/docs/manuscripts/authoring/_setup.qmd#L46
• https://github.com/quarto-dev/quarto-web/blob/8b1be24df1c9301f2700c77e41e7a9f789535717/docs/journals/authors.qmd#L142