From 164ec5f93c333a87a9530dbf02a821681c087ea5 Mon Sep 17 00:00:00 2001
From: Logan Barnett <logustus@gmail.com>
Date: Mon, 18 May 2026 16:22:58 -0700
Subject: [PATCH] fix: allow newlines inside bracket link path

Emacs org-mode treats bracket links as inline objects bounded by their
enclosing element (paragraph, list item, etc.), not by a single line.
A `[[path]]` or `[[path][desc]]` may span a newline inside either part
and is still recognised as one link.

orgize 0.10 previously excluded `\n` from the path `take_while`, so
multi-line links written by Emacs were dropped on parse.  Description
parsing already allowed newlines.

The element-level parsers (e.g. `paragraph_node`) already bound the
input passed to `link_node` at blank lines, so removing the per-link
newline restriction cannot let a link escape its element.
---
 src/syntax/link.rs | 65 +++++++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 64 insertions(+), 1 deletion(-)

diff --git a/src/syntax/link.rs b/src/syntax/link.rs
index a2aa4d9..447b5a1 100644
--- a/src/syntax/link.rs
+++ b/src/syntax/link.rs
@@ -22,7 +22,7 @@ pub fn link_node(input: Input) -> IResult<Input, GreenElement, ()> {
     let mut parser = map(
         tuple((
             l_bracket2_token,
-            take_while(|c: char| c != '<' && c != '>' && c != '\n' && c != ']'),
+            take_while(|c: char| c != '<' && c != '>' && c != ']'),
             opt(tuple((
                 r_bracket_token,
                 l_bracket_token,
@@ -110,3 +110,66 @@ fn parse() {
 
     assert!(link_node(("[[#id][desc]", config).into()).is_err());
 }
+
+/// Emacs org-mode accepts a single newline inside a bracket link — both inside
+/// the path and inside the description.  These tests document the discrepancy
+/// against orgize and serve as the reproducer for the upstream issue/PR.
+#[test]
+fn parse_multiline_link() {
+    use crate::{ast::Link, tests::to_ast};
+
+    let to_link = to_ast::<Link>(link_node);
+
+    // Path-only link with a newline inside the path.
+    let link = to_link("[[really really long link\nthat just keeps going]]");
+    insta::assert_debug_snapshot!(
+        link.syntax,
+        @r###"
+    LINK@0..49
+      L_BRACKET2@0..2 "[["
+      LINK_PATH@2..47 "really really long li ..."
+      R_BRACKET2@47..49 "]]"
+    "###
+    );
+
+    // Link with separate description that contains a newline.
+    let link = to_link("[[https://example.com][some\ndescription]]");
+    insta::assert_debug_snapshot!(
+        link.syntax,
+        @r###"
+    LINK@0..41
+      L_BRACKET2@0..2 "[["
+      LINK_PATH@2..21 "https://example.com"
+      R_BRACKET@21..22 "]"
+      L_BRACKET@22..23 "["
+      TEXT@23..39 "some\ndescription"
+      R_BRACKET2@39..41 "]]"
+    "###
+    );
+}
+
+/// `Org::parse` should also pick up multi-line links inside paragraphs.  This
+/// is the end-to-end case relevant to downstream formatters (e.g. org-fmt).
+#[test]
+fn parse_multiline_link_in_paragraph() {
+    use crate::{ast::Link, Org};
+    use rowan::ast::AstNode;
+
+    let org = Org::parse(
+        "Here is a [[really really long link\nthat just keeps going]] in prose.\n",
+    );
+
+    let links: Vec<_> = org
+        .document()
+        .syntax()
+        .descendants()
+        .filter_map(Link::cast)
+        .collect();
+
+    assert_eq!(
+        links.len(),
+        1,
+        "expected one link, found {}",
+        links.len(),
+    );
+}