Improve md link parsing

By prohibiting [ inside of links; fixes #555.

NEWS.md

@@ -108,6 +108,9 @@
       space (#628, @egnha).
 
     * `%` in inline code blocks is now automatically escaped (#640).
+
+    * Parsing of markdown links has been tweaked to reduce false positives 
+      (#555)
 
 # roxygen2 6.0.1
 

R/markdown.R

@@ -278,10 +278,10 @@ add_linkrefs_to_md <- function(text) {
     regex(
       comments = TRUE,
       "
-        (?<=[^\\]]|^)        # must not be preceded by ]
-        \\[([^\\]]+)\\]      # match anything inside of []
-        (?:\\[([^\\]]+)\\])? # match optional second pair of []
-        (?=[^\\[]|$)         # must not be followed by [
+        (?<=[^\\]]|^)           # must not be preceded by ]
+        \\[([^\\]\\[]+)\\]      # match anything inside of []
+        (?:\\[([^\\]\\[]+)\\])? # match optional second pair of []
+        (?=[^\\[]|$)            # must not be followed by [
       "
     )
   )[[1]]
@@ -303,6 +303,12 @@ add_linkrefs_to_md <- function(text) {
   )
 }
 
+# Helper designed primarily for testing
+
+md_link_html <- function(x) {
+  commonmark::markdown_html(add_linkrefs_to_md(x))
+}
+
 #' Parse a MarkDown link, to see if we should create an Rd link
 #'
 #' See the table above.

tests/testthat/test-rd-markdown-links.R

@@ -24,6 +24,13 @@ test_that("proper link references are added", {
   }
 })
 
+test_that("can not have [ inside of link", {
+  expect_equal(
+    md_link_html("`[[`. [subset()]"),
+    "<p><code>[[</code>. <a href=\"R:subset()\">subset()</a></p>\n"
+  )
+})
+
 test_that("commonmark picks up the various link references", {
   cases <- list(
     c("foo [func()] bar",