Merge pull request #4 from r-lib/main

Update

NEWS.md

@@ -1,5 +1,6 @@
 # pkgdown (development version)
 
+* `build_reference()` does a better job of parsing `\value{}` blocks (#2371).
 * When built on GitHub, source urls now use the name of the current upstream branch (rather than `HEAD`), which is more likely to generate correct links (#2597).
 * New `vignette("non-english")` that discusses non-English sites including how to submit new translations (#2605).
 * `build_reference()` now generates the usage that users actually type for infix and replacement methods (#2303).

R/rd-data.R

@@ -78,25 +78,46 @@ as_data.tag_value <- function(x, ...) {
 }
 
 describe_contents <- function(x, ..., id_prefix = NULL) {
-  # Drop pure whitespace nodes between items
-  is_ws <- purrr::map_lgl(x, is_whitespace)
-
-  # Group contiguous \items{}/whitespace into a <dl>
-  is_item <- purrr::map_lgl(x, inherits, "tag_item") | is_ws
-  changed <- is_item[-1] != is_item[-length(is_item)]
-  group <- cumsum(c(TRUE, changed))
-
-  parse_piece <- function(x) {
-    if (length(x) == 0) {
-      NULL
-    } else if (any(purrr::map_lgl(x, inherits, "tag_item"))) {
-      paste0("<dl>\n", parse_descriptions(x, ..., id_prefix = id_prefix), "</dl>")
+  if (length(x) == 0) {
+    return("")
+  }
+
+  # Group contiguous \items{}/whitespace into a <dl>; everything else
+  # is handled as is
+  block_id <- integer(length(x))
+  block_id[[1]] <- 1
+  cur_block_is_dl <- inherits(x[[1]], "tag_item")
+
+  for (i in seq2(2, length(x))) {
+    is_item <- inherits(x[[i]], "tag_item")
+    if (cur_block_is_dl) {
+      same_type <- is_item || is_whitespace(x[[i]])
+    } else {
+      same_type <- !is_item
+    }
+
+    if (same_type) {
+      block_id[[i]] <- block_id[[i - 1]]
+    } else {
+      block_id[[i]] <- block_id[[i - 1]] + 1
+      cur_block_is_dl <- !cur_block_is_dl
+    }
+  }
+
+  parse_block <- function(x) {
+    is_dl <- any(purrr::map_lgl(x, inherits, "tag_item"))
+    if (is_dl) {
+      paste0(
+        "<dl>\n",
+        parse_descriptions(x, ..., id_prefix = id_prefix),
+        "</dl>"
+      )
     } else {
       flatten_para(x, ...)
     }
   }
-  pieces <- split(x, group)
-  out <- purrr::map(pieces, parse_piece)
+  blocks <- split(x, block_id)
+  out <- purrr::map(blocks, parse_block)
 
   paste(unlist(out), collapse = "\n")
 }

R/utils.R

@@ -89,7 +89,7 @@ skip_if_no_pandoc <- function(version = "1.12.3") {
 }
 
 has_internet <- function() {
-  return(getOption("pkgdown.internet", default = TRUE))
+  getOption("pkgdown.internet", default = TRUE)
 }
 
 # remove '' quoting

tests/testthat/test-rd-data.R

@@ -48,7 +48,6 @@ test_that("leading whitespace doesn't break items", {
     value2html("\n\\item{a}{b}\n\n\\item{c}{d}\n\n\\item{e}{f}"),
     c(
       "<dl>",
-        "",
         "<dt>a</dt>", "<dd><p>b</p></dd>", "", "",
         "<dt>c</dt>", "<dd><p>d</p></dd>", "", "",
         "<dt>e</dt>", "<dd><p>f</p></dd>",
@@ -57,14 +56,10 @@ test_that("leading whitespace doesn't break items", {
   )
 })
 
-test_that("whitespace between text is preserved", {
+test_that("whitespace between text is not preserved", {
   expect_equal(
     value2html("a\n\nb\n\nc"),
-    c(
-      "<p>a</p>", "", "",
-      "<p>b</p>", "", "",
-      "<p>c</p>"
-    )
+    c("<p>a</p>", "<p>b</p>", "<p>c</p>")
   )
 })
 
@@ -82,3 +77,15 @@ test_that("can have multiple interleaved blocks", {
     )
   )
 })
+
+test_that("other tags don't affect breaking (#2371)", {
+  expect_equal(
+    value2html("1\\code{xxx}\n2\n3"),
+    c("<p>1<code>xxx</code>", "2", "3</p>")
+  )
+  # additionally teading whitespace
+  expect_equal(
+    value2html("1\\code{xxx}\n  2\n  3"),
+    c("<p>1<code>xxx</code>", "2", "3</p>")
+  )
+})

vignettes/translations.Rmd

@@ -38,7 +38,7 @@ Current available translations are:
 -   `tr`: Turkish
 -   `zh_CN`: Chinese (simplified)
 
-As you can see, most language codes are two letters, but if a language has multiple variants, it gets a longer form which can be used to disamiguate the options. For example, Chinese can use one of two forms: simplified (used in China and Singapore) or traditional (used in Taiwan and Hong Kong). Another example would be providing specific French Canadian translations by using code `fr_CN`.
+As you can see, most language codes are two letters, but if a language has multiple variants, it gets a longer form which can be used to disambiguate the options. For example, Chinese can use one of two forms: simplified (used in China and Singapore) or traditional (used in Taiwan and Hong Kong). Another example would be providing specific French Canadian translations by using code `fr_CN`.
 
 ## Translations