[[<-.custom produces unexpected exported NAMESPACE values

[schloerke]

I've made a reprex below showing that the tag's $val which appears to be a "" is not actually a "". This causes the roxy_tag_ns.roxy_tag_export function logic to fail.

tmpfile <- tempfile()
cat(file = tmpfile, "
#' @export​
`[[<-.custom` <- function(a,b,c) {
  a
}
​
#' @export​
`[<-.custom` <- function(a,b,c) {
  a
}
"
)

blocks <- roxygen2::parse_file(tmpfile)
blocks
#> [[1]]
#> <roxy_block> [file11399528b5575:3]
#>   $tag
#>     [line:  2] @export '​' {parsed}
#>     [????:???] @usage '<generated>' {parsed}
#>     [????:???] @.formals '<generated>' {parsed}
#>     [????:???] @backref '/var/folders/0k/bxg5lhr92sq74mb1d446ql540000gp/...' {parsed}
#>   $call   `[[<-.custom` <- function(a, b, c) { ...
#>   $object <s3method> 
#>     $topic [[<-.custom
#>     $alias [[<-.custom
#> 
#> [[2]]
#> <roxy_block> [file11399528b5575:8]
#>   $tag
#>     [line:  7] @export '​' {parsed}
#>     [????:???] @usage '<generated>' {parsed}
#>     [????:???] @.formals '<generated>' {parsed}
#>     [????:???] @backref '/var/folders/0k/bxg5lhr92sq74mb1d446ql540000gp/...' {parsed}
#>   $call   `[<-.custom` <- function(a, b, c) { ...
#>   $object <s3method> 
#>     $topic [<-.custom
#>     $alias [<-.custom

# Unexpected behavior
roxygen2:::blocks_to_ns(blocks)
#> [1] "export(\"​\")"

# The values appear to be a double quote
blocks[[1]]$tags[[1]]$raw
#> [1] "​"
blocks[[1]]$tags[[1]]$val
#> [1] "​"
# Does not equal a double quote
blocks[[1]]$tags[[1]]$raw == ""
#> [1] FALSE
blocks[[1]]$tags[[1]]$val == ""
#> [1] FALSE

# Since it does not equal a `""`, the then the `identical(x$val, "")` check fails in
roxygen2:::roxy_tag_ns.roxy_tag_export
#> function (x, block, env, import_only = FALSE) 
#> {
#>     if (import_only) {
#>         return()
#>     }
#>     if (identical(x$val, "")) {
#>         default_export(block$object, block)
#>     }
#>     else {
#>         export(x$val)
#>     }
#> }
#> <bytecode: 0x7fa5114865c8>
#> <environment: namespace:roxygen2>

# `$raw` is copied over here...
roxygen2:::tag_words_line
#> function (x) 
#> {
#>     x$val <- str_trim(x$raw)
#>     if (str_detect(x$val, "\n")) {
#>         roxy_tag_warning(x, "may only span a single line")
#>     }
#>     else if (!rdComplete(x$val)) {
#>         roxy_tag_warning(x, "mismatched braces or quotes")
#>     }
#>     else {
#>         x$val <- str_split(x$val, "\\s+")[[1]]
#>         x
#>     }
#> }
#> <bytecode: 0x7fa5211d3600>
#> <environment: namespace:roxygen2>

# Expected behavior
fixed_blocks <- blocks
fixed_blocks[[1]]$tags[[1]]$val <- ""
fixed_blocks[[2]]$tags[[1]]$val <- ""
roxygen2:::blocks_to_ns(fixed_blocks)
#> [1] "S3method(\"[<-\",custom)"  "S3method(\"[[<-\",custom)"

^{Created on 2021-03-31 by the reprex package (v0.3.0)}

---

I don't know if the original error is here:

roxygen2/src/parser2.cpp

Line 153 in 6c1e42f

_["raw"] = Rcpp::String(stripTrailingNewline(vals[i])),

Changing the string value to "" fixes my problem, but is not a solution.

---

Link to where $val is set (at least for my example):

roxygen2/R/tag-parser.R

Lines 135 to 136 in 6c1e42f

	tag_words_line <- function(x) {
	x$val <- str_trim(x$raw)

[schloerke]

Did more debugging with @cpsievert ...

Turns out, x$raw is not equal to "". If we call charToRaw(x$raw), I was expecting raw(0). It is actually charToRaw(x$raw) # e2 80 8b. This value is a "Zero width space". Weird!

Looking at the c++ code more, if the line below was disabled, the Zero Width Space problem would go away.

roxygen2/src/parser2.cpp

Line 133 in 6c1e42f

line.consumeText(&curVal);

[schloerke]

Work-around for now. #' @rawNamespace S3method("[[<-", custom)

[schloerke]

I'm going to blame slack.app. I can see there is a non-ascii character in my code (even the reprex).

When copying a text from a snippet in slack, the hidden UTF-8 characters were added in. 🤦