Lengthen fence characters when needed (fix #1621)

[atusy]

Fixes #1621.

• Update source
• Add tests
• Update NEWS
• pass CI

Reprex

rmd = tempfile(fileext = ".Rmd")
c(
  "````{r}",
  "invisible('",
  "```",
  "')",
  "````"
) |> writeLines(rmd)

knitr::knit(rmd, tempfile(), quiet = TRUE) |>
  readLines() |>
  cat(sep = "\n")

#> ````r
#> invisible('
#> ```
#> ')
#> ````

^{Created on 2021-09-21 by the reprex package (v2.0.1)}

[atusy]

For some reason following output occurs in https://github.com/yihui/knitr-examples/blob/master/115-engine-sql.Rmd

````r
readrPackage
``` 
````

```
## Error: attempt to use zero-length variable name
```

I do not understand why, but the error disappears if I trim following lines
https://github.com/yihui/knitr-examples/blob/08cf670e3296b5febffb454bad7b5acf5f1f6295/115-engine-sql.Rmd#L71-L77

I have to investigate more.

[atusy]

It seems the error happens when

• duplicated chunks exist.
• if there are any extra lines after the last duplicated chunk.

rmd = tempfile(fileext = ".Rmd")
c(
  "```{r}", "x=1", "```",
  "",
  "```{r}", "x=1", "``` ",
  ""
) |> writeLines(rmd)
cat(readLines(knitr::knit(rmd, tempfile())), sep = "\n")
#> processing file: /tmp/RtmppEyr1p/file37db7e54a1b1.Rmd
#> output file: /tmp/RtmppEyr1p/file37db661c6a48
#> 
#> ```r
#> x=1
#> ```
#> 
#> 
#> ````r
#> x=1
#> ``` 
#> ````
#> 
#> ```
#> #> Error: attempt to use zero-length variable name
#> ```

^{Created on 2021-09-22 by the reprex package (v2.0.1)}

[atusy]

The previous post is wrong.
The problem happens if chunk is ending with backticks followed by space(s).

rmd = tempfile(fileext = ".Rmd")
c(
  "```{r}", "x = 1", "``` ", ""
) |> writeLines(rmd)
cat(readLines(knitr::knit(rmd, tempfile())), sep = "\n")
#> processing file: /tmp/RtmpdRpzv0/file6f376822dca7.Rmd
#> output file: /tmp/RtmpdRpzv0/file6f374be0aa19
#> 
#> ````r
#> x = 1
#> ``` 
#> ````
#> 
#> ```
#> #> Error: attempt to use zero-length variable name
#> ```

^{Created on 2021-09-23 by the reprex package (v2.0.1)}

[atusy]

Yes, CI is green! Now, I'm ready for the review!!

[yihui]

This implementation is okay. I just wish to avoid the repetition of code in filter_chunk_end_general() and filter_chunk_end_md() if possible. Thank you!

[yihui]

In this case, I hope to update chunk.end and turn its elements to FALSE if the matched lines are not really the end of code chunks (i.e., if they do not match the number of backticks in chunk.begin).

After chunk.end truly represents chunk endings, we can use the original filter_chunk_end() instead of calling filter_chunk_end_md(), because the latter contains duplicated code from filter_chunk_end(), which I hope to avoid.

Could you try that?

[atusy]

I hope my simplification is fine which employs an approach different from yours.

If I understand right, updating chunk.end is the exactly what is done by filter_chunk_end*().
It means calling filter_chunk_end after updating the chunk.end will attempts unnecessary filtering.
This is why I chose the different approach.

https://github.com/yihui/knitr/pull/2047/files#diff-92c9f6892b4000a9291d262dbbdb4519edfc0437c89e7eeb686d84c1ee13119eL402-R417

[yihui]

Perfect! c302685 achieved exactly what I wanted, although it was a different method. Thank you!

[atusy]

@yihui Thanks for the review! I updated the codes based on it.

[yihui]

Thank you! I'll merge this PR later today.