Commit `6e2ef80a16`

6e2ef80a16a9eb61bff7f9928c230f4c5dff8240

parent: 1873fef702

Unsigned

cmc <hello@cleberg.net> · 2026-08-31 21:44 UTC

Match orgo's emphasis boundary characters

The characters allowed before an opening marker and after a closing one
were a narrower set than org's, so `=verbatim=-adjacent` did not close and
`-=code=` did not open. Both sets now match orgo, which is what the
conformance corpus is generated from: `-({'"` before, and
`-.,;:!?)}["'` after.

`]` leaves the post set, where it was never part of org's. That is safe
because a link's description is parsed as its own run, so a closing `]`
never follows a marker — asserted rather than assumed.

Closes #1

Layout: unified · split

Sources/OrgSwift/AST/OrgInlineParser.swift +13 −3

		@@ -131,11 +131,21 @@ extension OrgParser {
131	131	}
132	132	}
133	133
134		/// org requires the opening marker to follow whitespace or start the run.
	134	/// The character before an opening marker. org allows the start of the run,
	135	/// whitespace, or one of `-({'"` — the same set orgo uses, which is what the
	136	/// conformance corpus is generated from.
135	137	private static func boundaryBefore(_ chars: [Character], _ i: Int) -> Bool {
136		i == 0 \|\| chars[i - 1].isWhitespace \|\| "([{'\"".contains(chars[i - 1])
	138	i == 0 \|\| chars[i - 1].isWhitespace \|\| preChars.contains(chars[i - 1])
137	139	}
138	140
	141	private static let preChars: Set<Character> = ["-", "(", "{", "'", "\""]
	142
	143	/// The character after a closing marker, likewise from orgo. Note `[` is in the
	144	/// set and `]` is not, which looks backwards until you remember a link's
	145	/// description is parsed as its own run, so a `]` never follows a closer.
	146	private static let postChars: Set<Character> =
	147	["-", ".", ",", ";", ":", "!", "?", ")", "}", "[", "\"", "'"]
	148
139	149	private static func isWordCharacter(_ c: Character) -> Bool {
140	150	c.isLetter \|\| c.isNumber
141	151	}
		@@ -154,7 +164,7 @@ extension OrgParser {
154	164	if chars[j] == marker, let last = body.last, !last.isWhitespace {
155	165	// The closer must end the run or be followed by space/punctuation.
156	166	let after = j + 1 < chars.count ? chars[j + 1] : " "
157		if after.isWhitespace \|\| ".,;:!?)]}'\"".contains(after) \|\| j + 1 == chars.count {
	167	if after.isWhitespace \|\| postChars.contains(after) {
158	168	return (body, j + 1)
159	169	}
160	170	}

Tests/OrgSwiftTests/OrgTreeTests.swift +28

		@@ -344,6 +344,34 @@ struct OrgRenderOptionsTests {
344	344	#expect(bold.contains(.bold([.text("two words")])))
345	345	}
346	346
	347	/// The characters org allows around emphasis markers, matched to orgo — which is
	348	/// what the conformance corpus is generated from. The set was narrower here, so a
	349	/// closer against a dash or a bracket did not close.
	350	@Test
	351	func emphasisBoundaryCharactersMatchOrgo() {
	352	func verbatims(_ text: String) -> [String] {
	353	OrgParser.parseInline(text).compactMap {
	354	if case .verbatim(let t) = $0 { return t } else { return nil }
	355	}
	356	}
	357	// Newly closing: a dash or an opening bracket after the closer.
	358	#expect(verbatims("=verbatim=-adjacent") == ["verbatim"])
	359	#expect(verbatims("see =code=[1]") == ["code"])
	360	// Newly opening: a dash before the opener.
	361	#expect(verbatims("-=code= wrapped") == ["code"])
	362
	363	// `]` is not a post character in org, so this does not close.
	364	#expect(verbatims("=code=] trailing").isEmpty)
	365
	366	// Which is safe because a link's description is parsed as its own run, so a
	367	// closing `]` never actually follows the marker.
	368	let link = OrgParser.parseInline("[[https://example.com][=code=]]")
	369	guard case .link(let l) = link.first else {
	370	Issue.record("expected a link"); return
	371	}
	372	#expect(l.description ?? [] == [.verbatim("code")])
	373	}
	374
347	375	@Test
348	376	func rejectsUnsafeSchemes() {
349	377	let html = OrgRenderer.renderToHTML("[[javascript:alert(1)][click]]")

Commit 6e2ef80a16

Commit `6e2ef80a16`