Inline rmarkdown results inserted in wrong place

[hadley]

Something is going wrong with the Rd snipped generated by the link() function:

And message needs to mention "v" and "!", I'd suggest something like: Elements named "*", "i", "v", and "x", "!", are formatted as regular, info, success, failure, and error bullets respectively.

[lionel-]

This looks like a roxygen2 issue. This:

#' Topic
#' @name topic2
#' @param message foo
#'   . bar `r "\\ifelse{foo}{bar}{baz}"`.
NULL

Generates

% Generated by roxygen2: do not edit by hand
% Please edit documentation in R/cnd-abort.R
\name{topic2}
\alias{topic2}
\title{Topic}
\arguments{
\item{message}{foo
. ba\ifelse{foo}{bar}{baz}"`.}
}
\description{
Topic
}

It seems the . and newlines are necessary to cause the bug.

[hadley]

Probably in bug in str_set_all_pos(). I have a vague recollection of hitting this bug elsewhere. Maybe in downlit?

...

Maybe r-lib/downlit@3254d19 ?

[hadley]

Hmmm, I can't reprex:

library(roxygen2)

roc_proc_text(rd_roclet(), r"(
#' Topic
#' @name topic2
#' @param message foo
#'   . bar `r "\\ifelse{foo}{bar}{baz}"`.
NULL
)")[[1]]
#> % Generated by roxygen2: do not edit by hand
#> % Please edit documentation in ./<text>
#> \name{topic2}
#> \alias{topic2}
#> \title{Topic}
#> \arguments{
#> \item{message}{foo
#> . bar `r "\\ifelse{foo}{bar}{baz}"`.}
#> }
#> \description{
#> Topic
#> }

^{Created on 2022-07-06 by the reprex package (v2.0.1)}

[hadley]

Ah, needed to force md on:

library(roxygen2)

roc_proc_text(rd_roclet(), r"(
#' Topic
#' @md
#' @name topic2
#' @param message foo
#'   . bar `r "\\ifelse{foo}{bar}{baz}"`.
NULL
)")[[1]]
#> % Generated by roxygen2: do not edit by hand
#> % Please edit documentation in ./<text>
#> \name{topic2}
#> \alias{topic2}
#> \title{Topic}
#> \arguments{
#> \item{message}{foo
#> . ba\ifelse{foo}{bar}{baz}"`.}
#> }
#> \description{
#> Topic
#> }

^{Created on 2022-07-06 by the reprex package (v2.0.1)}

[hadley]

Somewhat simpler reprex:

library(roxygen2)

roc_proc_text(rd_roclet(), r"(
#' Title
#'
#' line1
#'    111 `r "222"` 333
#' @md
foo <- function() {}
)")[[1]]
#> % Generated by roxygen2: do not edit by hand
#> % Please edit documentation in ./<text>
#> \name{foo}
#> \alias{foo}
#> \title{Title}
#> \usage{
#> foo()
#> }
#> \description{
#> line1
#> 12222"` 333
#> }

^{Created on 2022-07-06 by the reprex package (v2.0.1)}

It's something to do with the leading indent — if you change the number of spaces, the corruption changes.

[hadley]

Probably related to this comment in str_set_all_pos():

  # Cmark has a bug when reporting source positions for multi-line
  # code tags, and it does not count the indenting space in the
  # continuation lines. However, the bug might get fixed later, so
  # for now we just simply error for multi-line inline code.

[gaborcsardi]

Yeah, unfortunately this seems to be a cmark bug:

src <- "Title\n\nline1\n   111 `r \"222\"` 333\n"
cat(src)

Title

line1
   111 `r "222"` 333

cat(commonmark::markdown_xml(src, sourcepos = TRUE))

<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE document SYSTEM "CommonMark.dtd">
<document sourcepos="1:1-4:20" xmlns="http://commonmark.org/xml/1.0">
  <paragraph sourcepos="1:1-1:5">
    <text sourcepos="1:1-1:5" xml:space="preserve">Title</text>
  </paragraph>
  <paragraph sourcepos="3:1-4:20">
    <text sourcepos="3:1-3:5" xml:space="preserve">line1</text>
    <softbreak />
    <text sourcepos="4:1-4:4" xml:space="preserve">111 </text>
    <code sourcepos="4:6-4:12" xml:space="preserve">r &quot;222&quot;</code>
    <text sourcepos="4:14-4:17" xml:space="preserve"> 333</text>
  </paragraph>
</document>

The sourcepos of the last line is clearly off. Let me see if a newer cmark version fixes this.

[gaborcsardi]

Seems that the R package uses https://github.com/github/cmark-gfm, for the GFM extensions, I guess and there is no fix for this: https://github.com/github/cmark-gfm/releases

I'll report it upstream.

As for roxygen2, let me see if we can work around this.

[gaborcsardi]

Found it in the cmark issue tracker, it does not seem like a thing that will be fixed soon: commonmark/cmark#296