Commit `f4591a4d3b`

f4591a4d3bcd7a90e2d01b7f8b6c9a6ca9a69b3f

parent: 3f7601917c

Unsigned

cmc <hello@cleberg.net> · 2026-10-04 23:22 UTC

OrgCore parser foundation (!1)

Adds the design doc, the first phase 1 plan, and the `OrgCore` package:

- `SourceText`: UTF-8 with or without BOM; invalid input is read-only and re-encodes to its original bytes.
- Lossless green/red syntax tree with UTF-16 lengths.
- Line classifier, element-aware TODO/priority settings scan, and a block-level parser (sections, headings, planning, drawers, blocks, keywords, tables, lists, footnotes, comments, fixed width).
- Tests: 44 unit, fuzz (2,000 seeded documents) and encoding cases; opt-in corpus test via `ORGSTAR_CORPUS`.

Layout: unified · split

Package.swift added +14

@@ -0,0 +1,14 @@
	1	// swift-tools-version: 6.2
	2	import PackageDescription
	3
	4	let package = Package(
	5	name: "Orgstar",
	6	platforms: [.macOS(.v26), .iOS(.v26)],
	7	products: [
	8	.library(name: "OrgCore", targets: ["OrgCore"])
	9	],
	10	targets: [
	11	.target(name: "OrgCore"),
	12	.testTarget(name: "OrgCoreTests", dependencies: ["OrgCore"])
	13	]
	14	)

Sources/OrgCore/Parser/Lines.swift added +153

@@ -0,0 +1,153 @@
	1	/// One line of source: its content and its terminator ("", "\n" or "\r\n"), both as slices of
	2	/// the original text.
	3	struct RawLine {
	4	let content: Substring
	5	let ending: Substring
	6	}
	7
	8	/// Splits on "\n" without normalizing anything. Works on unicode scalars, because String
	9	/// treats "\r\n" as a single Character.
	10	func splitRawLines(_ text: String) -> [RawLine] {
	11	var lines: [RawLine] = []
	12	let scalars = text.unicodeScalars
	13	var lineStart = scalars.startIndex
	14	var i = lineStart
	15	while i != scalars.endIndex {
	16	if scalars[i] == "\n" {
	17	var contentEnd = i
	18	if contentEnd > lineStart, scalars[scalars.index(before: i)] == "\r" {
	19	contentEnd = scalars.index(before: i)
	20	}
	21	let next = scalars.index(after: i)
	22	lines.append(RawLine(content: text[lineStart..<contentEnd], ending: text[contentEnd..<next]))
	23	lineStart = next
	24	i = next
	25	} else {
	26	i = scalars.index(after: i)
	27	}
	28	}
	29	if lineStart != scalars.endIndex {
	30	lines.append(RawLine(content: text[lineStart...], ending: ""))
	31	}
	32	return lines
	33	}
	34
	35	enum LineClass: Equatable {
	36	case blank
	37	case heading(level: Int)
	38	case blockBegin(name: String)
	39	case blockEnd(name: String)
	40	case dynamicBegin
	41	case dynamicEnd
	42	case drawerBegin(name: String)
	43	case drawerEnd
	44	case keyword(key: String)
	45	case comment
	46	case fixedWidth
	47	case horizontalRule
	48	case tableRow
	49	case footnoteDefinition
	50	case clock
	51	case planning
	52	case listItem
	53	case plain
	54	}
	55
	56	struct ClassifiedLine {
	57	let cls: LineClass
	58	/// Column of the first non-blank character, with tabs advancing to the next multiple of 8.
	59	let indent: Int
	60	}
	61
	62	func classifyLine(_ line: Substring) -> ClassifiedLine {
	63	var column = 0
	64	var rest = line
	65	while let c = rest.first, c == " " \|\| c == "\t" {
	66	column = c == "\t" ? (column / 8 + 1) * 8 : column + 1
	67	rest = rest.dropFirst()
	68	}
	69	if rest.isEmpty { return ClassifiedLine(cls: .blank, indent: column) }
	70	return ClassifiedLine(cls: lineClass(rest, columnZero: column == 0), indent: column)
	71	}
	72
	73	private func lineClass(_ rest: Substring, columnZero: Bool) -> LineClass {
	74	let trimmed = rest.trimmingTrailingWhitespace
	75
	76	if columnZero, rest.first == "*" {
	77	let stars = rest.prefix { $0 == "*" }
	78	let after = rest.dropFirst(stars.count)
	79	if after.isEmpty \|\| after.first == " " \|\| after.first == "\t" {
	80	return .heading(level: stars.count)
	81	}
	82	}
	83
	84	if rest.hasPrefix("#+") {
	85	let lower = trimmed.lowercased()
	86	if lower.hasPrefix("#+begin_") {
	87	let name = lower.dropFirst(8).prefix { !$0.isWhitespace }
	88	if !name.isEmpty { return .blockBegin(name: String(name)) }
	89	}
	90	if lower.hasPrefix("#+end_") {
	91	let name = lower.dropFirst(6)
	92	if !name.isEmpty, !name.contains(where: \.isWhitespace) { return .blockEnd(name: String(name)) }
	93	}
	94	if lower.hasPrefix("#+begin:") { return .dynamicBegin }
	95	if lower == "#+end:" { return .dynamicEnd }
	96	if let colon = rest.firstIndex(of: ":") {
	97	let key = rest[rest.index(rest.startIndex, offsetBy: 2)..<colon]
	98	if !key.isEmpty, !key.contains(where: \.isWhitespace) { return .keyword(key: key.uppercased()) }
	99	}
	100	}
	101
	102	if trimmed == "#" \|\| rest.hasPrefix("# ") \|\| rest.hasPrefix("#\t") { return .comment }
	103
	104	if rest.first == ":" {
	105	if trimmed == ":" \|\| rest.hasPrefix(": ") \|\| rest.hasPrefix(":\t") { return .fixedWidth }
	106	if trimmed.uppercased() == ":END:" { return .drawerEnd }
	107	if trimmed.count >= 3, trimmed.last == ":" {
	108	let name = trimmed.dropFirst().dropLast()
	109	if name.allSatisfy({ $0.isLetter \|\| $0.isNumber \|\| $0 == "_" \|\| $0 == "-" }) {
	110	return .drawerBegin(name: String(name))
	111	}
	112	}
	113	}
	114
	115	if rest.first == "\|" { return .tableRow }
	116	if trimmed.count >= 5, trimmed.allSatisfy({ $0 == "-" }) { return .horizontalRule }
	117
	118	if columnZero, rest.hasPrefix("[fn:"), let close = rest.firstIndex(of: "]"),
	119	close > rest.index(rest.startIndex, offsetBy: 4) {
	120	return .footnoteDefinition
	121	}
	122
	123	if rest.hasPrefix("CLOCK:") { return .clock }
	124	if rest.hasPrefix("SCHEDULED:") \|\| rest.hasPrefix("DEADLINE:") \|\| rest.hasPrefix("CLOSED:") { return .planning }
	125	if isListBullet(rest, indented: !columnZero) { return .listItem }
	126	return .plain
	127	}
	128
	129	/// `-`, `+`, `*` (indented only), `1.` or `1)`, followed by whitespace or end of line.
	130	/// Alphabetical bullets are off, as in org's default.
	131	private func isListBullet(_ rest: Substring, indented: Bool) -> Bool {
	132	guard let first = rest.first else { return false }
	133	let afterBullet: Substring
	134	if first == "-" \|\| first == "+" \|\| (first == "*" && indented) {
	135	afterBullet = rest.dropFirst()
	136	} else if first.isASCII, first.isNumber {
	137	let digits = rest.prefix { $0.isASCII && $0.isNumber }
	138	let tail = rest.dropFirst(digits.count)
	139	guard let separator = tail.first, separator == "." \|\| separator == ")" else { return false }
	140	afterBullet = tail.dropFirst()
	141	} else {
	142	return false
	143	}
	144	return afterBullet.isEmpty \|\| afterBullet.first == " " \|\| afterBullet.first == "\t"
	145	}
	146
	147	extension Substring {
	148	var trimmingTrailingWhitespace: Substring {
	149	var s = self
	150	while let last = s.last, last == " " \|\| last == "\t" { s = s.dropLast() }
	151	return s
	152	}
	153	}

Sources/OrgCore/Parser/Parser.swift added +348

@@ -0,0 +1,348 @@
	1	public enum OrgParser {
	2	public static func parse(_ text: String, defaults: OrgSettings = .default) -> OrgTree {
	3	var parser = Parser(text: text, defaults: defaults)
	4	return parser.run()
	5	}
	6	}
	7
	8	struct Parser {
	9	let lines: [RawLine]
	10	let info: [ClassifiedLine]
	11	/// Begin line → end line, for blocks, dynamic blocks and drawers that are closed before the
	12	/// next heading.
	13	let blockEnds: [Int: Int]
	14	let settings: OrgSettings
	15	var builder = GreenBuilder()
	16	var i = 0
	17
	18	init(text: String, defaults: OrgSettings) {
	19	lines = splitRawLines(text)
	20	info = lines.map { classifyLine($0.content) }
	21	blockEnds = Parser.matchEnds(info)
	22	settings = SettingsScanner.scan(lines: lines, info: info, blockEnds: blockEnds, defaults: defaults)
	23	}
	24
	25	static func matchEnds(_ info: [ClassifiedLine]) -> [Int: Int] {
	26	var ends: [Int: Int] = [:]
	27	var k = 0
	28	while k < info.count {
	29	let isEnd: ((LineClass) -> Bool)?
	30	switch info[k].cls {
	31	case .blockBegin(let name): isEnd = { $0 == .blockEnd(name: name) }
	32	case .dynamicBegin: isEnd = { $0 == .dynamicEnd }
	33	case .drawerBegin: isEnd = { $0 == .drawerEnd }
	34	default: isEnd = nil
	35	}
	36	if let isEnd {
	37	var j = k + 1
	38	while j < info.count {
	39	if case .heading = info[j].cls { break }
	40	if isEnd(info[j].cls) { ends[k] = j; break }
	41	j += 1
	42	}
	43	// Block contents are verbatim, so nothing inside starts another element.
	44	if let end = ends[k], !isDrawer(info[k].cls) { k = end }
	45	}
	46	k += 1
	47	}
	48	return ends
	49	}
	50
	51	static func isDrawer(_ cls: LineClass) -> Bool {
	52	if case .drawerBegin = cls { return true }
	53	return false
	54	}
	55
	56	mutating func run() -> OrgTree {
	57	builder.start(.document)
	58	if !lines.isEmpty, !isHeading(0) {
	59	builder.start(.zerothSection)
	60	parseContent(limit: lines.count)
	61	builder.finish()
	62	}
	63	while i < lines.count, case .heading(let level) = info[i].cls {
	64	parseSection(level: level)
	65	}
	66	builder.finish()
	67	return OrgTree(green: builder.build(), settings: settings)
	68	}
	69
	70	func isHeading(_ k: Int) -> Bool {
	71	if case .heading = info[k].cls { return true }
	72	return false
	73	}
	74
	75	mutating func parseSection(level: Int) {
	76	builder.start(.section)
	77	headingLine(lines[i])
	78	i += 1
	79	if i < lines.count, info[i].cls == .planning {
	80	builder.start(.planning)
	81	line(i)
	82	i += 1
	83	builder.finish()
	84	}
	85	if i < lines.count, case .drawerBegin(let name) = info[i].cls, name.uppercased() == "PROPERTIES",
	86	let end = blockEnds[i] {
	87	propertyDrawer(end: end)
	88	}
	89	parseContent(limit: lines.count)
	90	while i < lines.count, case .heading(let child) = info[i].cls, child > level {
	91	parseSection(level: child)
	92	}
	93	builder.finish()
	94	}
	95
	96	mutating func propertyDrawer(end: Int) {
	97	builder.start(.propertyDrawer)
	98	line(i)
	99	i += 1
	100	while i < end {
	101	if info[i].cls == .blank {
	102	line(i)
	103	} else {
	104	builder.start(.nodeProperty)
	105	line(i)
	106	builder.finish()
	107	}
	108	i += 1
	109	}
	110	line(i)
	111	i += 1
	112	builder.finish()
	113	}
	114
	115	/// Elements until `limit` or the next heading.
	116	mutating func parseContent(limit: Int) {
	117	while i < limit, !isHeading(i) {
	118	element(limit: limit, floor: nil)
	119	}
	120	}
	121
	122	static let affiliatedKeys: Set<String> = ["NAME", "CAPTION", "RESULTS", "HEADER", "PLOT"]
	123
	124	/// One element starting at `i`. `floor` is the indent of the enclosing list item, if any:
	125	/// non-blank lines at or left of it end the element.
	126	mutating func element(limit: Int, floor: Int?) {
	127	switch info[i].cls {
	128	case .blank:
	129	line(i)
	130	i += 1
	131	case .blockBegin, .dynamicBegin:
	132	if let end = blockEnds[i], end < limit {
	133	builder.start(info[i].cls == .dynamicBegin ? .dynamicBlock : .block)
	134	while i <= end {
	135	line(i)
	136	i += 1
	137	}
	138	builder.finish()
	139	} else {
	140	paragraph(limit: limit, floor: floor)
	141	}
	142	case .drawerBegin:
	143	if let end = blockEnds[i], end < limit {
	144	builder.start(.drawer)
	145	line(i)
	146	i += 1
	147	parseContent(limit: end)
	148	line(i)
	149	i += 1
	150	builder.finish()
	151	} else {
	152	paragraph(limit: limit, floor: floor)
	153	}
	154	case .keyword(let key):
	155	let affiliated = Self.affiliatedKeys.contains(key) \|\| key.hasPrefix("ATTR_")
	156	single(affiliated ? .affiliatedKeyword : .keyword)
	157	case .comment:
	158	consecutive(.comment, limit: limit, floor: floor) { $0 == .comment }
	159	case .fixedWidth:
	160	consecutive(.fixedWidth, limit: limit, floor: floor) { $0 == .fixedWidth }
	161	case .horizontalRule:
	162	single(.horizontalRule)
	163	case .clock:
	164	single(.clock)
	165	case .tableRow:
	166	table(limit: limit, floor: floor)
	167	case .footnoteDefinition:
	168	footnoteDefinition(limit: limit)
	169	case .listItem:
	170	list(limit: limit, floor: floor)
	171	default:
	172	paragraph(limit: limit, floor: floor)
	173	}
	174	}
	175
	176	mutating func single(_ kind: SyntaxKind) {
	177	builder.start(kind)
	178	line(i)
	179	i += 1
	180	builder.finish()
	181	}
	182
	183	mutating func consecutive(_ kind: SyntaxKind, limit: Int, floor: Int?, matching: (LineClass) -> Bool) {
	184	builder.start(kind)
	185	repeat {
	186	line(i)
	187	i += 1
	188	} while i < limit && matching(info[i].cls) && within(floor, i)
	189	builder.finish()
	190	}
	191
	192	mutating func table(limit: Int, floor: Int?) {
	193	builder.start(.table)
	194	while i < limit, info[i].cls == .tableRow, within(floor, i) {
	195	single(.tableRow)
	196	}
	197	while i < limit, info[i].cls == .keyword(key: "TBLFM"), within(floor, i) {
	198	single(.tableFormula)
	199	}
	200	builder.finish()
	201	}
	202
	203	mutating func footnoteDefinition(limit: Int) {
	204	builder.start(.footnoteDefinition)
	205	line(i)
	206	i += 1
	207	while i < limit, info[i].cls == .plain {
	208	line(i)
	209	i += 1
	210	}
	211	builder.finish()
	212	}
	213
	214	mutating func list(limit: Int, floor: Int?) {
	215	let base = info[i].indent
	216	builder.start(.plainList)
	217	while i < limit, info[i].cls == .listItem, info[i].indent == base, within(floor, i) {
	218	item(base: base, limit: limit)
	219	}
	220	builder.finish()
	221	}
	222
	223	/// An item's first line, then everything indented past its bullet. One blank line stays
	224	/// inside the item when the item or list continues after it; two end the list.
	225	mutating func item(base: Int, limit: Int) {
	226	builder.start(.item)
	227	line(i)
	228	i += 1
	229	while i < limit, !isHeading(i) {
	230	if info[i].cls == .blank {
	231	var j = i
	232	while j < limit, info[j].cls == .blank { j += 1 }
	233	guard j - i < 2, j < limit else { break }
	234	let continuesItem = info[j].indent > base
	235	let nextSibling = info[j].cls == .listItem && info[j].indent == base
	236	guard continuesItem \|\| nextSibling else { break }
	237	line(i)
	238	i += 1
	239	if nextSibling { break }
	240	continue
	241	}
	242	guard info[i].indent > base else { break }
	243	element(limit: limit, floor: base)
	244	}
	245	builder.finish()
	246	}
	247
	248	mutating func paragraph(limit: Int, floor: Int?) {
	249	builder.start(.paragraph)
	250	line(i)
	251	i += 1
	252	while i < limit, within(floor, i), continuesParagraph(i) {
	253	line(i)
	254	i += 1
	255	}
	256	builder.finish()
	257	}
	258
	259	/// Lines that don't start an element of their own.
	260	func continuesParagraph(_ k: Int) -> Bool {
	261	switch info[k].cls {
	262	case .plain, .planning, .blockEnd, .dynamicEnd, .drawerEnd:
	263	return true
	264	case .blockBegin, .dynamicBegin, .drawerBegin:
	265	return blockEnds[k] == nil
	266	default:
	267	return false
	268	}
	269	}
	270
	271	func within(_ floor: Int?, _ k: Int) -> Bool {
	272	guard let floor else { return true }
	273	return info[k].indent > floor
	274	}
	275
	276	// MARK: - Tokens
	277
	278	/// A whole line as leading whitespace, content and line ending.
	279	mutating func line(_ k: Int) {
	280	let rest = whitespace(lines[k].content)
	281	builder.token(.text, rest)
	282	builder.token(.newline, lines[k].ending)
	283	}
	284
	285	mutating func whitespace(_ s: Substring) -> Substring {
	286	let ws = s.prefix { $0 == " " \|\| $0 == "\t" }
	287	builder.token(.whitespace, ws)
	288	return s.dropFirst(ws.count)
	289	}
	290
	291	mutating func headingLine(_ raw: RawLine) {
	292	builder.start(.heading)
	293	var rest = raw.content
	294	let stars = rest.prefix { $0 == "*" }
	295	builder.token(.stars, stars)
	296	rest = whitespace(rest.dropFirst(stars.count))
	297
	298	let word = rest.prefix { $0 != " " && $0 != "\t" }
	299	if !word.isEmpty, settings.todoKeywordNames.contains(String(word)) {
	300	builder.token(.todoKeyword, word)
	301	rest = whitespace(rest.dropFirst(word.count))
	302	}
	303
	304	if let cookie = priorityCookie(rest) {
	305	builder.token(.priority, cookie)
	306	rest = whitespace(rest.dropFirst(cookie.count))
	307	}
	308
	309	let parts = splitTags(rest)
	310	builder.token(.title, parts.title)
	311	builder.token(.whitespace, parts.gap)
	312	builder.token(.tags, parts.tags)
	313	builder.token(.whitespace, parts.trailing)
	314	builder.token(.newline, raw.ending)
	315	builder.finish()
	316	}
	317
	318	/// `[#A]` or `[#10]`, followed by whitespace or end of line.
	319	func priorityCookie(_ s: Substring) -> Substring? {
	320	guard s.hasPrefix("[#"), let close = s.firstIndex(of: "]") else { return nil }
	321	let value = s[s.index(s.startIndex, offsetBy: 2)..<close]
	322	let valid = (value.count == 1 && value.first!.isLetter && value.first!.isUppercase)
	323	\|\| (!value.isEmpty && value.allSatisfy { $0.isASCII && $0.isNumber })
	324	guard valid else { return nil }
	325	let after = s[s.index(after: close)...]
	326	guard after.isEmpty \|\| after.first == " " \|\| after.first == "\t" else { return nil }
	327	return s[...close]
	328	}
	329
	330	func splitTags(_ s: Substring) -> (title: Substring, gap: Substring, tags: Substring, trailing: Substring) {
	331	let trimmed = s.trimmingTrailingWhitespace
	332	let trailing = s[trimmed.endIndex...]
	333	let none = (title: trimmed, gap: Substring(), tags: Substring(), trailing: trailing)
	334	guard trimmed.last == ":" else { return none }
	335	let tagStart = trimmed.lastIndex { $0 == " " \|\| $0 == "\t" }.map { trimmed.index(after: $0) } ?? trimmed.startIndex
	336	let tags = trimmed[tagStart...]
	337	guard tags.count >= 3, tags.first == ":", isTagString(tags) else { return none }
	338	let before = trimmed[..<tagStart]
	339	let title = before.trimmingTrailingWhitespace
	340	return (title, before[title.endIndex...], tags, trailing)
	341	}
	342
	343	func isTagString(_ tags: Substring) -> Bool {
	344	tags.dropFirst().dropLast().split(separator: ":", omittingEmptySubsequences: false).allSatisfy { tag in
	345	!tag.isEmpty && tag.allSatisfy { $0.isLetter \|\| $0.isNumber \|\| "_@#%".contains($0) }
	346	}
	347	}
	348	}

Sources/OrgCore/Parser/Settings.swift added +129

@@ -0,0 +1,129 @@
	1	public struct TodoKeyword: Sendable, Hashable {
	2	public var name: String
	3	public var fastKey: Character?
	4	/// Logging flag when entering the state (`!` or `@`), from `NAME(k!/@)`.
	5	public var logOnEnter: String?
	6	/// Logging flag when leaving the state.
	7	public var logOnLeave: String?
	8
	9	public init(name: String, fastKey: Character? = nil, logOnEnter: String? = nil, logOnLeave: String? = nil) {
	10	self.name = name
	11	self.fastKey = fastKey
	12	self.logOnEnter = logOnEnter
	13	self.logOnLeave = logOnLeave
	14	}
	15	}
	16
	17	public struct TodoSequence: Sendable, Equatable {
	18	public enum Kind: Sendable, Equatable { case sequence, type }
	19
	20	public var kind: Kind
	21	public var active: [TodoKeyword]
	22	public var done: [TodoKeyword]
	23
	24	public init(kind: Kind, active: [TodoKeyword], done: [TodoKeyword]) {
	25	self.kind = kind
	26	self.active = active
	27	self.done = done
	28	}
	29	}
	30
	31	public struct Priorities: Sendable, Equatable {
	32	public var highest: String
	33	public var lowest: String
	34	public var `default`: String
	35
	36	public init(highest: String, lowest: String, default: String) {
	37	self.highest = highest
	38	self.lowest = lowest
	39	self.default = `default`
	40	}
	41	}
	42
	43	public struct OrgSettings: Sendable, Equatable {
	44	public var todoSequences: [TodoSequence]
	45	public var priorities: Priorities
	46
	47	public init(todoSequences: [TodoSequence], priorities: Priorities) {
	48	self.todoSequences = todoSequences
	49	self.priorities = priorities
	50	}
	51
	52	public static let `default` = OrgSettings(
	53	todoSequences: [TodoSequence(kind: .sequence, active: [TodoKeyword(name: "TODO")], done: [TodoKeyword(name: "DONE")])],
	54	priorities: Priorities(highest: "A", lowest: "C", default: "B")
	55	)
	56
	57	public var todoKeywordNames: Set<String> {
	58	Set(todoSequences.flatMap { ($0.active + $0.done).map(\.name) })
	59	}
	60
	61	public func isDone(_ name: String) -> Bool {
	62	todoSequences.contains { $0.done.contains { $0.name == name } }
	63	}
	64	}
	65
	66	enum SettingsScanner {
	67	/// Reads `#+TODO`, `#+SEQ_TODO`, `#+TYP_TODO` and `#+PRIORITIES` outside blocks. Any TODO
	68	/// line replaces the default sequences, as in org.
	69	static func scan(lines: [RawLine], info: [ClassifiedLine], blockEnds: [Int: Int], defaults: OrgSettings) -> OrgSettings {
	70	var sequences: [TodoSequence] = []
	71	var priorities = defaults.priorities
	72	var k = 0
	73	while k < lines.count {
	74	switch info[k].cls {
	75	case .blockBegin, .dynamicBegin:
	76	if let end = blockEnds[k] { k = end }
	77	case .keyword(let key):
	78	let value = keywordValue(lines[k].content)
	79	switch key {
	80	case "TODO", "SEQ_TODO":
	81	if let s = todoSequence(value, kind: .sequence) { sequences.append(s) }
	82	case "TYP_TODO":
	83	if let s = todoSequence(value, kind: .type) { sequences.append(s) }
	84	case "PRIORITIES":
	85	let words = value.split(whereSeparator: \.isWhitespace)
	86	if words.count == 3 {
	87	priorities = Priorities(highest: String(words[0]), lowest: String(words[1]), default: String(words[2]))
	88	}
	89	default:
	90	break
	91	}
	92	default:
	93	break
	94	}
	95	k += 1
	96	}
	97	return OrgSettings(todoSequences: sequences.isEmpty ? defaults.todoSequences : sequences, priorities: priorities)
	98	}
	99
	100	static func keywordValue(_ line: Substring) -> Substring {
	101	guard let colon = line.firstIndex(of: ":") else { return "" }
	102	return line[line.index(after: colon)...]
	103	}
	104
	105	static func todoSequence(_ value: Substring, kind: TodoSequence.Kind) -> TodoSequence? {
	106	let words = value.split(whereSeparator: \.isWhitespace)
	107	guard !words.isEmpty else { return nil }
	108	if let bar = words.firstIndex(of: "\|") {
	109	return TodoSequence(kind: kind, active: words[..<bar].map(todoKeyword), done: words[(bar + 1)...].map(todoKeyword))
	110	}
	111	return TodoSequence(kind: kind, active: words.dropLast().map(todoKeyword), done: [todoKeyword(words.last!)])
	112	}
	113
	114	/// `NAME`, or `NAME(spec)` where spec is an optional fast key followed by `enter/leave`
	115	/// logging flags.
	116	static func todoKeyword(_ word: Substring) -> TodoKeyword {
	117	guard let open = word.firstIndex(of: "("), word.last == ")" else { return TodoKeyword(name: String(word)) }
	118	var spec = word[word.index(after: open)..<word.index(before: word.endIndex)]
	119	var fastKey: Character?
	120	if let first = spec.first, first != "!", first != "@", first != "/" {
	121	fastKey = first
	122	spec = spec.dropFirst()
	123	}
	124	let parts = spec.split(separator: "/", omittingEmptySubsequences: false)
	125	let enter = parts.first.flatMap { $0.isEmpty ? nil : String($0) }
	126	let leave = parts.count > 1 && !parts[1].isEmpty ? String(parts[1]) : nil
	127	return TodoKeyword(name: String(word[..<open]), fastKey: fastKey, logOnEnter: enter, logOnLeave: leave)
	128	}
	129	}

Sources/OrgCore/SourceText.swift added +40

@@ -0,0 +1,40 @@
	1	import Foundation
	2
	3	/// A file's bytes and their decoded text. Only UTF-8 (with or without a BOM) is editable;
	4	/// anything else decodes with replacement characters for display and is never written back.
	5	public struct SourceText: Sendable {
	6	public let originalBytes: [UInt8]
	7	public let hasBOM: Bool
	8	public let isValidUTF8: Bool
	9	/// Decoded text without the BOM.
	10	public let text: String
	11
	12	private static let bom: [UInt8] = [0xEF, 0xBB, 0xBF]
	13
	14	public init(bytes: [UInt8]) {
	15	originalBytes = bytes
	16	hasBOM = bytes.starts(with: Self.bom)
	17	let body = hasBOM ? Array(bytes.dropFirst(3)) : bytes
	18	if let decoded = String(validating: body, as: UTF8.self) {
	19	text = decoded
	20	isValidUTF8 = true
	21	} else {
	22	text = String(decoding: body, as: UTF8.self)
	23	isValidUTF8 = false
	24	}
	25	}
	26
	27	public init(_ text: String) {
	28	self.init(bytes: Array(text.utf8))
	29	}
	30
	31	public var isEditable: Bool { isValidUTF8 }
	32
	33	/// Bytes to write for `newText`. Unchanged text returns the original bytes. Valid UTF-8
	34	/// round-trips through `String` unchanged, so untouched spans keep their exact bytes.
	35	public func encode(_ newText: String) -> [UInt8] {
	36	if newText == text { return originalBytes }
	37	precondition(isEditable, "a read-only document cannot be re-encoded")
	38	return (hasBOM ? Self.bom : []) + Array(newText.utf8)
	39	}
	40	}

Sources/OrgCore/Syntax/GreenTree.swift added +87

@@ -0,0 +1,87 @@
	1	public struct GreenToken: Sendable, Equatable {
	2	public let kind: SyntaxKind
	3	public let text: String
	4	/// Length in UTF-16 code units.
	5	public let length: Int
	6
	7	public init(kind: SyntaxKind, text: String) {
	8	self.kind = kind
	9	self.text = text
	10	self.length = text.utf16.count
	11	}
	12	}
	13
	14	/// An immutable node. Stores only kind, children and length, so unchanged subtrees can be
	15	/// shared between versions of a document.
	16	public final class GreenNode: Sendable, Equatable {
	17	public let kind: SyntaxKind
	18	public let children: [GreenElement]
	19	/// Length in UTF-16 code units.
	20	public let length: Int
	21
	22	public init(kind: SyntaxKind, children: [GreenElement]) {
	23	self.kind = kind
	24	self.children = children
	25	self.length = children.reduce(0) { $0 + $1.length }
	26	}
	27
	28	public static func == (lhs: GreenNode, rhs: GreenNode) -> Bool {
	29	lhs === rhs \|\| (lhs.kind == rhs.kind && lhs.children == rhs.children)
	30	}
	31
	32	public var text: String {
	33	var out = ""
	34	write(to: &out)
	35	return out
	36	}
	37
	38	func write(to out: inout String) {
	39	for child in children {
	40	switch child {
	41	case .node(let node): node.write(to: &out)
	42	case .token(let token): out += token.text
	43	}
	44	}
	45	}
	46	}
	47
	48	public enum GreenElement: Sendable, Equatable {
	49	case node(GreenNode)
	50	case token(GreenToken)
	51
	52	public var length: Int {
	53	switch self {
	54	case .node(let node): node.length
	55	case .token(let token): token.length
	56	}
	57	}
	58	}
	59
	60	struct GreenBuilder {
	61	private var stack: [(kind: SyntaxKind, children: [GreenElement])] = []
	62	private var root: GreenNode?
	63
	64	mutating func start(_ kind: SyntaxKind) {
	65	stack.append((kind, []))
	66	}
	67
	68	mutating func token(_ kind: SyntaxKind, _ text: some StringProtocol) {
	69	guard !text.isEmpty else { return }
	70	stack[stack.count - 1].children.append(.token(GreenToken(kind: kind, text: String(text))))
	71	}
	72
	73	mutating func finish() {
	74	let (kind, children) = stack.removeLast()
	75	let node = GreenNode(kind: kind, children: children)
	76	if stack.isEmpty {
	77	root = node
	78	} else {
	79	stack[stack.count - 1].children.append(.node(node))
	80	}
	81	}
	82
	83	func build() -> GreenNode {
	84	precondition(stack.isEmpty, "unfinished nodes")
	85	return root!
	86	}
	87	}

Sources/OrgCore/Syntax/SyntaxKind.swift added +12

@@ -0,0 +1,12 @@
	1	public enum SyntaxKind: String, Sendable {
	2	// Tokens
	3	case text, newline, whitespace
	4	case stars, todoKeyword, priority, title, tags
	5
	6	// Nodes
	7	case document, zerothSection, section, heading
	8	case planning, propertyDrawer, nodeProperty, drawer, clock
	9	case paragraph, plainList, item, table, tableRow, tableFormula
	10	case block, dynamicBlock, keyword, affiliatedKeyword
	11	case comment, fixedWidth, horizontalRule, footnoteDefinition
	12	}

Sources/OrgCore/Syntax/SyntaxNode.swift added +59

@@ -0,0 +1,59 @@
	1	/// A view of a green node at an absolute offset, with a parent link. Created on demand.
	2	public final class SyntaxNode: Sendable {
	3	public let green: GreenNode
	4	public let offset: Int
	5	public let parent: SyntaxNode?
	6
	7	init(green: GreenNode, offset: Int, parent: SyntaxNode?) {
	8	self.green = green
	9	self.offset = offset
	10	self.parent = parent
	11	}
	12
	13	public var kind: SyntaxKind { green.kind }
	14	public var range: Range<Int> { offset..<(offset + green.length) }
	15	public var text: String { green.text }
	16
	17	public var children: [SyntaxNode] {
	18	var result: [SyntaxNode] = []
	19	var at = offset
	20	for child in green.children {
	21	if case .node(let node) = child {
	22	result.append(SyntaxNode(green: node, offset: at, parent: self))
	23	}
	24	at += child.length
	25	}
	26	return result
	27	}
	28
	29	public var tokens: [SyntaxToken] {
	30	var result: [SyntaxToken] = []
	31	var at = offset
	32	for child in green.children {
	33	if case .token(let token) = child {
	34	result.append(SyntaxToken(kind: token.kind, text: token.text, range: at..<(at + token.length)))
	35	}
	36	at += child.length
	37	}
	38	return result
	39	}
	40
	41	/// This node and every node below it, in document order.
	42	public func descendants() -> [SyntaxNode] {
	43	[self] + children.flatMap { $0.descendants() }
	44	}
	45	}
	46
	47	public struct SyntaxToken: Sendable, Equatable {
	48	public let kind: SyntaxKind
	49	public let text: String
	50	public let range: Range<Int>
	51	}
	52
	53	public struct OrgTree: Sendable {
	54	public let green: GreenNode
	55	public let settings: OrgSettings
	56
	57	public var root: SyntaxNode { SyntaxNode(green: green, offset: 0, parent: nil) }
	58	public var text: String { green.text }
	59	}

Tests/OrgCoreTests/LinesTests.swift added +64

@@ -0,0 +1,64 @@
	1	import Testing
	2	@testable import OrgCore
	3
	4	struct LinesTests {
	5	@Test func splitKeepsEveryEnding() {
	6	let lines = splitRawLines("a\r\nb\n\nc")
	7	#expect(lines.map { String($0.content) } == ["a", "b", "", "c"])
	8	#expect(lines.map { String($0.ending) } == ["\r\n", "\n", "\n", ""])
	9	#expect(splitRawLines("").isEmpty)
	10	#expect(splitRawLines("x\n").count == 1)
	11	}
	12
	13	@Test func splitIsLossless() {
	14	let text = "\r\n\n a\r b\r\n😀\n"
	15	#expect(splitRawLines(text).map { String($0.content) + String($0.ending) }.joined() == text)
	16	}
	17
	18	@Test(arguments: [
	19	("", LineClass.blank),
	20	(" \t", .blank),
	21	("* a", .heading(level: 1)),
	22	("*** ", .heading(level: 3)),
	23	("*", .heading(level: 1)),
	24	("bold text", .plain),
	25	(" * a", .listItem),
	26	("#+BEGIN_SRC sh :results output", .blockBegin(name: "src")),
	27	("#+end_src", .blockEnd(name: "src")),
	28	("#+BEGIN: clocktable :scope file", .dynamicBegin),
	29	("#+END:", .dynamicEnd),
	30	("#+TITLE: x", .keyword(key: "TITLE")),
	31	("#+tblfm: $2=$1", .keyword(key: "TBLFM")),
	32	("# comment", .comment),
	33	("#", .comment),
	34	("#hashtag", .plain),
	35	(": fixed", .fixedWidth),
	36	(":", .fixedWidth),
	37	(":PROPERTIES:", .drawerBegin(name: "PROPERTIES")),
	38	(" :LOGBOOK:", .drawerBegin(name: "LOGBOOK")),
	39	(":END:", .drawerEnd),
	40	("\| a \| b \|", .tableRow),
	41	("-----", .horizontalRule),
	42	("----", .plain),
	43	("[fn:1] note", .footnoteDefinition),
	44	("CLOCK: [2026-10-04 Sun 10:00]", .clock),
	45	("SCHEDULED: <2026-10-04 Sun>", .planning),
	46	("- item", .listItem),
	47	("+ item", .listItem),
	48	("1. item", .listItem),
	49	("2) item", .listItem),
	50	("-", .listItem),
	51	("-x", .plain),
	52	("1.5 apples", .plain),
	53	("plain text", .plain),
	54	])
	55	func classify(line: String, expected: LineClass) {
	56	#expect(classifyLine(line[...]).cls == expected)
	57	}
	58
	59	@Test func indentCountsTabsToEight() {
	60	#expect(classifyLine("\t- a").indent == 8)
	61	#expect(classifyLine(" \t- a").indent == 8)
	62	#expect(classifyLine(" - a").indent == 3)
	63	}
	64	}

Tests/OrgCoreTests/ParserElementTests.swift added +79

@@ -0,0 +1,79 @@
	1	import Testing
	2	@testable import OrgCore
	3
	4	func childKinds(_ text: String) -> [SyntaxKind] {
	5	let root = OrgParser.parse(text).root
	6	let container = root.children.first { $0.kind == .zerothSection \|\| $0.kind == .section }!
	7	return container.children.map(\.kind)
	8	}
	9
	10	struct ParserElementTests {
	11	@Test func planningAndPropertiesFollowHeading() {
	12	let text = "* a\nSCHEDULED: <2026-10-04 Sun>\n:PROPERTIES:\n:ID: x\n:END:\nbody\n"
	13	#expect(childKinds(text) == [.heading, .planning, .propertyDrawer, .paragraph])
	14	}
	15
	16	@Test func planningElsewhereIsText() {
	17	#expect(childKinds("SCHEDULED: <2026-10-04 Sun>\n") == [.paragraph])
	18	}
	19
	20	@Test func blocks() {
	21	#expect(childKinds("#+begin_src sh\n,* escaped\n:END:\n#+end_src\nafter\n") == [.block, .paragraph])
	22	#expect(childKinds("#+BEGIN: clocktable\n#+END:\n") == [.dynamicBlock])
	23	}
	24
	25	@Test func headingsEndBlocks() {
	26	#expect(childKinds("#+begin_src sh\n* heading\n#+end_src\n") == [.paragraph])
	27	}
	28
	29	@Test func unclosedBlockIsParagraph() {
	30	#expect(childKinds("#+begin_src sh\necho\n") == [.paragraph])
	31	}
	32
	33	@Test func drawersHoldElements() {
	34	let root = OrgParser.parse(":LOGBOOK:\nCLOCK: [2026-10-04 Sun 10:00]\n:END:\n").root
	35	let drawer = root.children[0].children[0]
	36	#expect(drawer.kind == .drawer)
	37	#expect(drawer.children.map(\.kind) == [.clock])
	38	}
	39
	40	@Test func keywords() {
	41	#expect(childKinds("#+TITLE: x\n#+NAME: t\n#+ATTR_HTML: :width 10\n") == [.keyword, .affiliatedKeyword, .affiliatedKeyword])
	42	}
	43
	44	@Test func commentsAndFixedWidthGroup() {
	45	#expect(childKinds("# a\n# b\n: c\n: d\n-----\n") == [.comment, .fixedWidth, .horizontalRule])
	46	}
	47
	48	@Test func tableWithFormulas() {
	49	let root = OrgParser.parse("\| a \|\n\|---\|\n\| 1 \|\n#+TBLFM: $1=2\n#+TBLFM: $1=3\n").root
	50	let table = root.children[0].children[0]
	51	#expect(table.kind == .table)
	52	#expect(table.children.map(\.kind) == [.tableRow, .tableRow, .tableRow, .tableFormula, .tableFormula])
	53	}
	54
	55	@Test func footnoteDefinition() {
	56	#expect(childKinds("[fn:1] note\ncontinued\n\nafter\n") == [.footnoteDefinition, .paragraph])
	57	}
	58
	59	@Test func listsNestByIndent() {
	60	let text = "- a\n more\n - b\n- c\n\nafter\n"
	61	let root = OrgParser.parse(text).root
	62	let list = root.children[0].children[0]
	63	#expect(list.kind == .plainList)
	64	#expect(list.children.map(\.kind) == [.item, .item])
	65	#expect(list.children[0].children.map(\.kind) == [.paragraph, .plainList])
	66	#expect(childKinds(text) == [.plainList, .paragraph])
	67	}
	68
	69	@Test func twoBlankLinesEndAList() {
	70	#expect(childKinds("- a\n\n\n- b\n") == [.plainList, .plainList])
	71	#expect(childKinds("- a\n\n- b\n") == [.plainList])
	72	}
	73
	74	@Test func paragraphStopsAtElementStart() {
	75	#expect(childKinds("text\n\| a \|\n") == [.paragraph, .table])
	76	#expect(childKinds("text\n#+begin_quote\nq\n#+end_quote\n") == [.paragraph, .block])
	77	#expect(childKinds("text\n#+begin_quote\nq\n") == [.paragraph])
	78	}
	79	}

Tests/OrgCoreTests/ParserSectionTests.swift added +66

@@ -0,0 +1,66 @@
	1	import Testing
	2	@testable import OrgCore
	3
	4	func nodeKinds(_ text: String) -> [SyntaxKind] {
	5	OrgParser.parse(text).root.descendants().map(\.kind)
	6	}
	7
	8	func tokens(of kind: SyntaxKind, in text: String) -> [SyntaxToken] {
	9	OrgParser.parse(text).root.descendants().filter { $0.kind == kind }.flatMap(\.tokens)
	10	}
	11
	12	struct ParserSectionTests {
	13	@Test func emptyDocument() {
	14	let tree = OrgParser.parse("")
	15	#expect(tree.text == "")
	16	#expect(nodeKinds("") == [.document])
	17	}
	18
	19	@Test func zerothSectionHoldsPreamble() {
	20	#expect(nodeKinds("text\n* a\n") == [.document, .zerothSection, .paragraph, .section, .heading])
	21	}
	22
	23	@Test func sectionsNestByLevel() {
	24	let text = "* a\n b\n* c\n** d\n* e\n"
	25	let root = OrgParser.parse(text).root
	26	let top = root.children
	27	#expect(top.map(\.kind) == [.section, .section])
	28	#expect(top[0].children.map(\.kind) == [.heading, .section, .section])
	29	#expect(top[0].children[1].children.map(\.kind) == [.heading, .section])
	30	}
	31
	32	@Test func headingTokens() {
	33	let parts = tokens(of: .heading, in: "** TODO [#A] Write the plan :work:urgent: \n")
	34	#expect(parts.map(\.kind) == [.stars, .whitespace, .todoKeyword, .whitespace, .priority, .whitespace, .title, .whitespace, .tags, .whitespace, .newline])
	35	#expect(parts.first { $0.kind == .tags }?.text == ":work:urgent:")
	36	#expect(parts.first { $0.kind == .title }?.text == "Write the plan")
	37	}
	38
	39	@Test func todoKeywordsComeFromSettings() {
	40	let text = "#+TODO: NEXT \| DONE\n* NEXT a\n* TODO b\n"
	41	let todo = tokens(of: .heading, in: text).filter { $0.kind == .todoKeyword }.map(\.text)
	42	#expect(todo == ["NEXT"])
	43	}
	44
	45	@Test func priorityNeedsValidValueAndSpace() {
	46	#expect(tokens(of: .heading, in: "* [#B] x\n").contains { $0.kind == .priority })
	47	#expect(tokens(of: .heading, in: "* [#10] x\n").contains { $0.kind == .priority })
	48	#expect(!tokens(of: .heading, in: "* [#AB] x\n").contains { $0.kind == .priority })
	49	#expect(!tokens(of: .heading, in: "* [#A]x\n").contains { $0.kind == .priority })
	50	}
	51
	52	@Test func tagsNeedValidCharacters() {
	53	#expect(tokens(of: .heading, in: "* a :b@c_1:\n").contains { $0.kind == .tags })
	54	#expect(!tokens(of: .heading, in: "* a :b c:\n").contains { $0.kind == .tags })
	55	#expect(!tokens(of: .heading, in: "* a :b:c\n").contains { $0.kind == .tags })
	56	}
	57
	58	@Test func headingWithoutNewlineAtEnd() {
	59	#expect(OrgParser.parse("* a").text == "* a")
	60	}
	61
	62	@Test(arguments: ["* a\n", "\n", " TODO\n", "text\r\n* a\r\n** b\r\n", "\n\n* a\n\n"])
	63	func roundTrip(text: String) {
	64	#expect(OrgParser.parse(text).text == text)
	65	}
	66	}

Tests/OrgCoreTests/RoundTripTests.swift added +75

@@ -0,0 +1,75 @@
	1	import Foundation
	2	import Testing
	3	@testable import OrgCore
	4
	5	/// SplitMix64, so failures reproduce from the seed.
	6	struct SeededGenerator: RandomNumberGenerator {
	7	var state: UInt64
	8	mutating func next() -> UInt64 {
	9	state &+= 0x9E37_79B9_7F4A_7C15
	10	var z = state
	11	z = (z ^ (z >> 30)) &* 0xBF58_476D_1CE4_E5B9
	12	z = (z ^ (z >> 27)) &* 0x94D0_49BB_1331_11EB
	13	return z ^ (z >> 31)
	14	}
	15	}
	16
	17	let fragments = [
	18	"* ", " TODO [#A] title :a:b:", "* DONE", "#+TODO: NEXT \| DONE", "#+begin_src sh", "#+end_src",
	19	"#+BEGIN_QUOTE", "#+end_quote", "#+BEGIN: clocktable", "#+END:", ":PROPERTIES:", ":ID: x", ":END:",
	20	":LOGBOOK:", "CLOCK: [2026-10-04 Sun 10:00]", "SCHEDULED: <2026-10-04 Sun>", "- item", " - nested",
	21	"\t+ tab", "1. one", "\| a \| b \|", "\|---+---\|", "#+TBLFM: $2=$1", "# comment", ": fixed", "-----",
	22	"[fn:1] note", "#+NAME: x", "plain text", "é", "😀", "e\u{301}", " ", "\t", "\n", "\n", "\r\n", "\r", "",
	23	]
	24
	25	func randomDocument(_ rng: inout SeededGenerator) -> String {
	26	(0..<Int.random(in: 0...40, using: &rng)).map { _ in
	27	fragments.randomElement(using: &rng)! + (Bool.random(using: &rng) ? "\n" : "")
	28	}.joined()
	29	}
	30
	31	func checkLengths(_ node: SyntaxNode) -> Bool {
	32	let sum = node.green.children.reduce(0) { $0 + $1.length }
	33	return sum == node.green.length && node.children.allSatisfy(checkLengths)
	34	}
	35
	36	struct RoundTripTests {
	37	@Test func fuzzedDocumentsRoundTrip() {
	38	var rng = SeededGenerator(state: 20261004)
	39	for n in 0..<2_000 {
	40	let text = randomDocument(&rng)
	41	let tree = OrgParser.parse(text)
	42	#expect(tree.text == text, "document \(n)")
	43	#expect(checkLengths(tree.root), "document \(n)")
	44	}
	45	}
	46
	47	@Test(arguments: [
	48	[0xEF, 0xBB, 0xBF] + Array("* a\r\n".utf8),
	49	Array("* a\r\n- b\n\tc".utf8),
	50	Array("* a".utf8),
	51	Array("😀 e\u{301}\n".utf8),
	52	[0x2A, 0x20, 0xFF, 0x0A] as [UInt8],
	53	])
	54	func encodingFixturesRoundTrip(bytes: [UInt8]) {
	55	let source = SourceText(bytes: bytes)
	56	let tree = OrgParser.parse(source.text)
	57	#expect(source.encode(tree.text) == bytes)
	58	}
	59
	60	/// Private corpus: `ORGSTAR_CORPUS=/path/to/org swift test --filter corpus`.
	61	@Test(.enabled(if: ProcessInfo.processInfo.environment["ORGSTAR_CORPUS"] != nil))
	62	func corpusRoundTrips() throws {
	63	let root = URL(fileURLWithPath: ProcessInfo.processInfo.environment["ORGSTAR_CORPUS"]!)
	64	let files = FileManager.default.enumerator(at: root, includingPropertiesForKeys: nil)!
	65	.compactMap { $0 as? URL }
	66	.filter { ["org", "org_archive"].contains($0.pathExtension) }
	67	#expect(!files.isEmpty)
	68	for file in files {
	69	let bytes = try [UInt8](Data(contentsOf: file))
	70	let source = SourceText(bytes: bytes)
	71	let tree = OrgParser.parse(source.text)
	72	#expect(source.encode(tree.text) == bytes, "\(file.path)")
	73	}
	74	}
	75	}

Tests/OrgCoreTests/SettingsTests.swift added +48

@@ -0,0 +1,48 @@
	1	import Testing
	2	@testable import OrgCore
	3
	4	struct SettingsTests {
	5	func scan(_ text: String, blockEnds: [Int: Int] = [:]) -> OrgSettings {
	6	let lines = splitRawLines(text)
	7	let info = lines.map { classifyLine($0.content) }
	8	return SettingsScanner.scan(lines: lines, info: info, blockEnds: blockEnds, defaults: .default)
	9	}
	10
	11	@Test func defaultsWithoutKeywords() {
	12	let settings = scan("* TODO a\n")
	13	#expect(settings.todoKeywordNames == ["TODO", "DONE"])
	14	#expect(settings.isDone("DONE"))
	15	}
	16
	17	@Test func fileKeywordsReplaceDefaults() {
	18	let settings = scan("#+TODO: NEXT(n) WAIT(w@/!) \| DONE(d!) CANCELED(c@)\n")
	19	#expect(settings.todoKeywordNames == ["NEXT", "WAIT", "DONE", "CANCELED"])
	20	let sequence = settings.todoSequences[0]
	21	#expect(sequence.active.map(\.name) == ["NEXT", "WAIT"])
	22	#expect(sequence.done.map(\.name) == ["DONE", "CANCELED"])
	23	#expect(sequence.active[1] == TodoKeyword(name: "WAIT", fastKey: "w", logOnEnter: "@", logOnLeave: "!"))
	24	#expect(sequence.done[0] == TodoKeyword(name: "DONE", fastKey: "d", logOnEnter: "!", logOnLeave: nil))
	25	}
	26
	27	@Test func lastWordIsDoneWithoutSeparator() {
	28	let settings = scan("#+SEQ_TODO: A B C\n")
	29	#expect(settings.todoSequences[0].active.map(\.name) == ["A", "B"])
	30	#expect(settings.todoSequences[0].done.map(\.name) == ["C"])
	31	}
	32
	33	@Test func severalLinesMakeSeveralSequences() {
	34	let settings = scan("#+TODO: A \| B\n#+TYP_TODO: X \| Y\n")
	35	#expect(settings.todoSequences.count == 2)
	36	#expect(settings.todoSequences[1].kind == .type)
	37	}
	38
	39	@Test func keywordsInsideBlocksAreIgnored() {
	40	let text = "#+begin_example\n#+TODO: X \| Y\n#+end_example\n"
	41	#expect(scan(text, blockEnds: [0: 2]).todoKeywordNames == ["TODO", "DONE"])
	42	}
	43
	44	@Test func priorities() {
	45	#expect(scan("#+PRIORITIES: 1 10 5\n").priorities == Priorities(highest: "1", lowest: "10", default: "5"))
	46	#expect(scan("").priorities == Priorities(highest: "A", lowest: "C", default: "B"))
	47	}
	48	}

Tests/OrgCoreTests/SourceTextTests.swift added +40

@@ -0,0 +1,40 @@
	1	import Testing
	2	@testable import OrgCore
	3
	4	struct SourceTextTests {
	5	@Test func plainUTF8() {
	6	let source = SourceText(bytes: Array("* a\n".utf8))
	7	#expect(source.text == "* a\n")
	8	#expect(!source.hasBOM)
	9	#expect(source.isEditable)
	10	}
	11
	12	@Test func bomIsStrippedAndRestored() {
	13	let bytes: [UInt8] = [0xEF, 0xBB, 0xBF] + Array("x\n".utf8)
	14	let source = SourceText(bytes: bytes)
	15	#expect(source.hasBOM)
	16	#expect(source.text == "x\n")
	17	#expect(source.encode(source.text) == bytes)
	18	#expect(source.encode("y\n") == [0xEF, 0xBB, 0xBF] + Array("y\n".utf8))
	19	}
	20
	21	@Test func crlfAndMixedEndingsSurvive() {
	22	let bytes = Array("a\r\nb\nc\r\n".utf8)
	23	let source = SourceText(bytes: bytes)
	24	#expect(source.encode(source.text) == bytes)
	25	}
	26
	27	@Test func invalidUTF8IsReadOnlyAndUnchanged() {
	28	let bytes: [UInt8] = [0x61, 0xFF, 0x0A]
	29	let source = SourceText(bytes: bytes)
	30	#expect(!source.isValidUTF8)
	31	#expect(!source.isEditable)
	32	#expect(source.encode(source.text) == bytes)
	33	}
	34
	35	@Test func nonBMPAndCombiningSurvive() {
	36	let bytes = Array("😀 e\u{301}\n".utf8)
	37	let source = SourceText(bytes: bytes)
	38	#expect(source.encode(source.text) == bytes)
	39	}
	40	}

Tests/OrgCoreTests/SyntaxTreeTests.swift added +45

@@ -0,0 +1,45 @@
	1	import Testing
	2	@testable import OrgCore
	3
	4	struct SyntaxTreeTests {
	5	func sample() -> GreenNode {
	6	var b = GreenBuilder()
	7	b.start(.document)
	8	b.start(.paragraph)
	9	b.token(.text, "hé😀")
	10	b.token(.newline, "\n")
	11	b.finish()
	12	b.token(.newline, "\r\n")
	13	b.finish()
	14	return b.build()
	15	}
	16
	17	@Test func lengthsAreUTF16() {
	18	let green = sample()
	19	#expect(green.length == 4 + 1 + 2)
	20	#expect(green.text == "hé😀\n\r\n")
	21	}
	22
	23	@Test func redNodesCarryOffsets() {
	24	let root = SyntaxNode(green: sample(), offset: 0, parent: nil)
	25	let paragraph = root.children[0]
	26	#expect(paragraph.kind == .paragraph)
	27	#expect(paragraph.range == 0..<5)
	28	#expect(paragraph.parent === root)
	29	#expect(root.tokens.map(\.range) == [5..<7])
	30	#expect(paragraph.tokens.map(\.kind) == [.text, .newline])
	31	}
	32
	33	@Test func builderSkipsEmptyTokens() {
	34	var b = GreenBuilder()
	35	b.start(.document)
	36	b.token(.whitespace, "")
	37	b.finish()
	38	#expect(b.build().children.isEmpty)
	39	}
	40
	41	@Test func descendantsArePreorder() {
	42	let root = SyntaxNode(green: sample(), offset: 0, parent: nil)
	43	#expect(root.descendants().map(\.kind) == [.document, .paragraph])
	44	}
	45	}

docs/design.md added +393

@@ -0,0 +1,393 @@
	1	# Orgstar Design Doc
	2
	3	Oct 4, 2026
	4
	5	## Summary
	6
	7	A native macOS app, with iOS to follow, that edits a folder of org files directly and can replace Emacs for someone who only uses org-mode. The buffer is the file: the app is a text editor with an org-aware syntax tree, rendering drawn over the text, and views (agenda, table recalculation, export) that write back minimal edits.
	8
	9	Goals
	10
	11	- Cover the org features used daily: cycling, TODO/priority/tags, planning and timestamps, tables with formulas, source blocks with execution, `M-q` fill, agenda, export.
	12	- Byte-for-byte fidelity: text the user did not edit never changes, so Emacs, beorg and Syncthing can work on the same files.
	13	- macOS first. Every core decision keeps an iOS port possible without a rewrite.
	14
	15	Non-goals for v1
	16
	17	- Emacs features unrelated to org (an elisp runtime, packages, other major modes).
	18	- Block-based storage, or a database as the source of truth.
	19	- Mac App Store distribution.
	20	- Babel execution on iOS.
	21	- Native Calc symbolic math in table formulas.
	22	- Sync providers other than Syncthing and iCloud Drive.
	23
	24	## Decisions
	25
	26	\| Area \| Decision \| Reason \|
	27	\| --- \| --- \| --- \|
	28	\| Editing model \| Text editor; the buffer is the file. Rendering and views sit on top. \| Fill, table alignment and fidelity need plain text. Block editors can't round-trip. \|
	29	\| Platforms \| macOS v1. iOS after the Mac app is mature. \| One platform at a time; core stays portable. \|
	30	\| Distribution \| Direct download (notarized, Sparkle) and Homebrew cask. No sandbox. \| Babel needs to spawn interpreters. \|
	31	\| Language \| Swift throughout. Core in a separate package with no AppKit/UIKit imports. \| TextKit uses UTF-16 offsets natively; no FFI layer; direct path to iOS. \|
	32	\| Parser \| New lossless parser (`OrgCore`) with source ranges. Port inline rules from OrgSwift. \| OrgSwift and orgo are export parsers: no positions, drop drawers/comments, flat headings. \|
	33	\| Key bindings \| Command registry; keymaps as data. Presets: Emacs (default), Mac, Doom. \| One registry serves keys, palette, menus and iOS. Doom needs a modal engine, so it ships last. \|
	34	\| Babel \| Execute `sh`/`bash`/`shell` and `python` with real header args; generic interpreter runner; `emacs-lisp` via `emacs --batch`. Tree-sitter highlighting for ~20 languages. iOS execution: maybe. \| Covers the most-used languages; generic runner adds ruby, node, R, sqlite3, awk cheaply. \|
	35	\| Table formulas \| Native evaluator for the common subset; `emacs --batch` fallback for the rest. \| Runs on iOS; Calc can't be reproduced exactly. \|
	36	\| Storage \| Plain folders: Syncthing-synced and iCloud Drive. SQLite index is a rebuildable cache. \| Files stay the source of truth. \|
	37
	38	## Architecture
	39
	40	Three layers: a thin app shell per platform, a platform layer for system services, and `OrgCore`, which holds all org logic and is shared unchanged by both apps.
	41
	42	Diagram (described in text):
	43
	44	- macOS app (v1, AppKit shell, menus) and iOS app (later, UIKit shell, touch input) both sit on the platform layer.
	45	- Platform layer (AppKit or UIKit): Editor view (TextKit 2, rendering over text), Notifications (scheduled from agenda queries), Processes (Babel runs and `emacs --batch`, talks to external interpreters: sh, python, emacs), Files (bookmarks, watching, saving; reads/writes the org folders on Syncthing and iCloud, which are the source of truth).
	46	- The platform layer calls into OrgCore (commands, tree queries).
	47	- OrgCore (Swift package, no UI imports): Parser (lossless tree, UTF-16 ranges), Commands (pure functions, minimal edits), Agenda (index queries, saved views), Compute (TBLFM evaluator, Babel headers). OrgCore writes to OrgIndex (SQLite + FTS5, rebuildable cache).
	48	- Keymaps (TOML; Emacs, Mac, Doom) resolve keys to OrgCore commands.
	49
	50	Packages
	51
	52	- `OrgCore`: parser, syntax tree, semantic layer, commands, agenda queries, table formula evaluator, Babel header parsing and results writing. Foundation only.
	53	- `OrgIndex`: GRDB schema, indexing and queries. Foundation only.
	54	- `OrgPlatform`: TextKit 2 editor view, file access and watching, process runner, notifications. Separate macOS and iOS implementations behind shared protocols.
	55	- `OrgSwift` (existing): HTML and SwiftUI rendering, used for export and preview.
	56	- App target: windows, sidebar, menus, settings, keymap loading.
	57
	58	## OrgCore data model
	59
	60	`OrgCore` parses a file into a lossless syntax tree: concatenating its tokens reproduces the file exactly, and every node knows its range in UTF-16 code units, the unit `NSTextStorage` uses.
	61
	62	Tree structure
	63
	64	- Two layers, as in rust-analyzer's rowan: immutable green nodes store kind and length and can be shared between versions; red nodes are created on demand with absolute offsets and parent links.
	65	- Whitespace, blank lines and newlines are tokens in the tree, not discarded.
	66	- Anything the parser doesn't recognize becomes a `Raw` node that keeps its text, so unknown syntax survives every edit.
	67
	68	Bytes and encoding
	69
	70	The tree holds text, so byte fidelity needs its own contract.
	71
	72	- Supported: UTF-8, with or without a BOM. Anything else (invalid UTF-8, UTF-16, legacy encodings) opens read-only with a notice; it is never converted on save.
	73	- Each document keeps its original bytes plus metadata: BOM present, line-ending style per line (LF, CRLF, mixed), trailing newline present.
	74	- Line endings stay as tokens in the tree; nothing normalizes CRLF.
	75	- Saving writes untouched byte spans from the original buffer and encodes only edited spans. A UTF-16 edit maps to a byte range through a per-line offset table.
	76	- Fixtures: BOM, CRLF, mixed endings, no trailing newline, non-BMP characters, combining sequences, tabs, invalid UTF-8.
	77
	78	Node kinds
	79
	80	- Containers: `Document`, `Section` (heading line, its content, child sections).
	81	- Heading parts: stars, TODO keyword, priority cookie, title objects, tags.
	82	- Heading metadata: `Planning` (SCHEDULED/DEADLINE/CLOSED), `PropertyDrawer`, `Drawer` (incl. LOGBOOK), `Clock`.
	83	- Elements: `Paragraph`, `PlainList`/`Item` (with checkbox and description term), `Table`/`TableRow`/`TableFormula`, blocks (src, example, quote, center, verse, export, special, dynamic), `Keyword`, `AffiliatedKeyword` (`#+NAME`, `#+CAPTION`, `#+RESULTS`), `Comment`, `FixedWidth`, `HorizontalRule`, `FootnoteDefinition`.
	84	- Objects: emphasis, link, timestamp (with repeater and warning cookies), footnote reference, entity, inline src block, macro, statistics cookie, LaTeX fragment, target.
	85
	86	Parse settings
	87
	88	Some in-buffer settings change how text parses.
	89
	90	- Syntax-affecting: `#+TODO`/`#+SEQ_TODO`/`#+TYP_TODO`, `#+PRIORITIES`. Semantic only: `#+FILETAGS`, `#+STARTUP`, `#+PROPERTY`.
	91	- The settings pass is element-aware: a `#+TODO` line inside a src, example or export block is not a setting.
	92	- TODO precedence follows org: any file-level TODO line replaces the app default sequences for that file; several lines define several sequences; `\|` separates active from done states; fast-select keys and logging suffixes (`TODO(t!)`) are parsed and kept.
	93	- An edit that adds, removes or changes a syntax-affecting line triggers a full reparse. A semantic-only change invalidates the semantic layer and the file's index rows, not the tree.
	94
	95	Incremental reparse
	96
	97	1. Compute the damaged range from both the old and the new text: the edited lines plus one line on each side.
	98	2. Classify the damaged lines in both versions (heading, block or drawer delimiter, list item, footnote definition, keyword, table row, plain). If any class differs between old and new, the edit changed structure.
	99	3. No structural change: reparse the enclosing element and reuse everything else.
	100	4. Structural change: reparse forward from the start of the enclosing section until the new tree and the old tree agree on a section boundary at the same shifted offset (a synchronization point), then reuse the old tree from there.
	101	5. Content before the first heading is a zeroth section, so every offset has an enclosing section.
	102	6. When the classifier is unsure, or no synchronization point is found, reparse the whole file.
	103
	104	A differential test asserts that the incremental result always equals a full parse, with edits that create and delete delimiters, join lines, remove indentation, add list tabs and change keyword lines. Latency targets are in Phases.
	105
	106	Semantic layer
	107
	108	Typed views over the tree, computed and cached per tree version: `HeadingInfo` (todo, priority, tags with inheritance, properties with inheritance, planning, clocks, ID), `TableModel` (cells, column widths, formulas), `SrcBlockInfo` (language, header args merged from `#+PROPERTY`, property drawers and the block line). Commands and the index read these, never raw text.
	109
	110	Inheritance is not one rule, so the semantic layer stores local values and resolved values separately, and each consumer picks a policy:
	111
	112	- Tags: inherited by default, minus an exclusion list (`org-tags-exclude-from-inheritance`); `#+FILETAGS` apply to the whole file.
	113	- Properties: not inherited for search by default; opt-in per property (`org-use-property-inheritance`). Special properties (`ID`, `CATEGORY`, `ARCHIVE`, `COLUMNS`) have their own rules; `ID` is never inherited.
	114	- Babel header args: resolved from `#+PROPERTY: header-args`, language-specific `header-args:lang`, additive `+` forms, heading properties, `#+HEADER:` lines and the block line, in org's order.
	115	- Every resolved value records where it came from, so the UI can show it and tests can check it.
	116
	117	From OrgSwift
	118
	119	Port the inline rules: emphasis border characters (matched to orgo), link forms, timestamp parsing and ranges. OrgSwift itself stays the HTML/SwiftUI renderer; later it can render from the `OrgCore` tree so there is one parser.
	120
	121	## Commands and keymaps
	122
	123	Every feature is a named command, implemented as a pure function in `OrgCore`. Keys, menus, the command palette and (later) iOS touch controls all invoke the same commands.
	124
	125	Command shape
	126
	127	```swift
	128	protocol OrgCommand {
	129	static var id: String { get } // "org.todo.cycle"
	130	static var title: String { get } // shown in the palette
	131	func applies(in ctx: EditContext) -> Bool
	132	func run(in ctx: EditContext) throws -> CommandStep
	133	}
	134
	135	struct EditContext {
	136	let document: DocumentID
	137	let revision: Int // the revision the command read
	138	let text: String
	139	let tree: OrgTree
	140	let selection: [Range<Int>]
	141	let settings: OrgSettings
	142	let now: Date // injected; commands never read the clock
	143	let calendar: Calendar // includes time zone
	144	let answers: [String: String] // replies to earlier prompts
	145	}
	146
	147	enum CommandStep {
	148	case commit(EditResult)
	149	case prompt(Prompt) // ask, then rerun with the answer in `answers`
	150	}
	151
	152	struct EditResult { var baseRevision: Int; var edits: [TextEdit]; var selection: [Range<Int>]?; var effects: [Effect] }
	153	struct TextEdit { let range: Range<Int>; let replacement: String } // UTF-16, against baseRevision, non-overlapping
	154	```
	155
	156	- `edits` are minimal replacements, applied as one undo group.
	157	- `effects` cover anything that isn't a text change: run a src block, open the agenda, show a message, fold a subtree. The platform layer performs them.
	158	- Context dispatch works like org's TAB: one key can bind several commands, and the first whose `applies` returns true runs (cycle on a heading, next cell in a table, indent in a list).
	159
	160	Revisions, time and prompts
	161
	162	- Every edit result names the revision it was computed against. The document session applies it only if that is still the current revision; otherwise the command reruns on the new state.
	163	- Prompts (a note when logging a state change, a date for SCHEDULED) use the prepare/prompt/commit loop above. Cancelling at any prompt produces no edit.
	164	- Repeaters, `LOGBOOK` notes and `CLOSED` stamps use the injected `now` and `calendar`, so tests are deterministic.
	165	- Asynchronous effects (a Babel run) carry the revision and the target element's identity. On completion they re-find the target in the current tree (by `#+NAME`, or by the block's position relative to its neighbours) and are rejected with a message if it is gone or ambiguous. Result insertion is its own undo step.
	166
	167	Document session
	168
	169	One session per open file owns the text, original bytes, tree, revision counter, merge base and undo history. Each window or split has its own selection, fold state and narrowing, mapped through every edit. On an external reload or merge, view state is mapped through the diff; folds on headings that no longer exist are dropped.
	170
	171	Keymap format
	172
	173	Keymaps are TOML files, user-editable, loaded in layers: preset, then user overrides.
	174
	175	```toml
	176	[[bind]]
	177	keys = "C-c C-t"
	178	command = "org.todo.cycle"
	179
	180	[[bind]]
	181	keys = "TAB"
	182	command = "org.cycle"
	183	when = "heading"
	184
	185	[[bind]]
	186	keys = "SPC m t"
	187	command = "org.todo.cycle"
	188	mode = "normal" # Doom preset only
	189	```
	190
	191	- Key sequences of any length (`C-c C-x C-i`), with an echo area for the pending prefix and a which-key popup after a short delay.
	192	- `when` names a context predicate; `mode` names a modal state. Both exist in the format from day one even though only Doom uses `mode`.
	193	- An "Option as Meta" setting, per side (left/right), like Terminal and iTerm2.
	194	- `⌘` shortcuts (save, undo, find) stay active in every preset.
	195
	196	Presets
	197
	198	\| Preset \| Default \| Needs \| Ships \|
	199	\| --- \| --- \| --- \| --- \|
	200	\| Emacs \| Yes \| Prefix-key engine \| Phase 2 \|
	201	\| Mac \| No \| Menu wiring \| Phase 2 \|
	202	\| Doom \| No \| Modal engine: normal/insert/visual, operators + motions + text objects, counts, `.` repeat, registers; `SPC` leader; evil-org bindings \| After phase 3 \|
	203
	204	## Workspace, storage and index
	205
	206	A workspace is one or more root folders. The files are the only source of truth; the SQLite index can be deleted and rebuilt at any time.
	207
	208	Folders and access
	209
	210	- The user adds folders through an open panel. Store security-scoped bookmarks even though the Mac app is unsandboxed, because iOS will require them.
	211	- Three scopes, configured separately:
	212	- Discovery: every `.org` and `.org_archive` file under the roots. Indexed for search and link resolution.
	213	- Agenda: folders or globs, the equivalent of `org-agenda-files`. Archive files and subtrees tagged `ARCHIVE` are excluded by default.
	214	- Link resolution: all discovered files, including archives, so `id:` links into archived subtrees still resolve. Duplicate IDs are reported. A link to a file outside the roots opens it but doesn't index it.
	215	- Syncthing conflict copies are discovered but excluded from agenda, search and ID resolution; they appear only in the conflict UI.
	216
	217	Watching for external changes
	218
	219	- FSEvents (macOS) and `NSFilePresenter` are hints that something changed, not a complete log.
	220	- On launch, and whenever FSEvents reports dropped events, a root change, or must-scan-subdirs, rescan the affected tree and reconcile the index: compare path, size, mtime and hash; detect renames by hash; delete rows for missing files. Reconciliation runs in one transaction per root.
	221	- Roots are identified by bookmark, not path, so a moved root is followed or reported.
	222	- iOS (later): `NSFilePresenter` plus `NSMetadataQuery` for iCloud.
	223	- Ignore Syncthing temp files (`.syncthing..tmp`). Show Syncthing conflict files (`.sync-conflict-*`) with a diff against the original.
	224	- iCloud placeholders: a file not yet downloaded is requested with `startDownloadingUbiquitousItem` and shown as loading.
	225
	226	Saving
	227
	228	- Save mode is a user setting: autosave after an idle delay, or explicit save only.
	229	- Save sequence, inside a coordinated write:
	230	1. Read the current disk bytes and hash.
	231	2. If the hash equals the merge base, write. Otherwise three-way merge (base = merge base, ours = buffer, theirs = disk). On conflict, stop and show the conflict; write nothing.
	232	3. Write to a temp file in the same folder and replace atomically.
	233	4. Read back the hash; it becomes the new merge base.
	234	- Emacs and Syncthing don't use `NSFileCoordinator`, so another writer can still replace the file between steps 1 and 3. We can't fully prevent that; we narrow the window and detect it: after the write, if FSEvents reports a change whose hash is neither ours nor the base, treat it as a new external edit and merge again.
	235	- Before any write that replaces a version we didn't produce, keep a copy of both versions in a recovery folder (last 20 per file) so nothing is lost.
	236	- An unedited open file reloads silently on external change, keeping cursor and folds. An edited one merges automatically when the merge is clean, and shows the conflict otherwise.
	237	- Fault-injection tests replace the file at each step of the sequence.
	238
	239	Index schema (GRDB, SQLite + FTS5)
	240
	241	\| Table \| Columns \|
	242	\| --- \| --- \|
	243	\| `files` \| id, root, path, size, mtime, hash, parsed_at \|
	244	\| `headings` \| id, file_id, parent_id, start, end, level, todo, priority, title, outline_path, org_id \|
	245	\| `tags` \| heading_id, tag, inherited \|
	246	\| `properties` \| heading_id, key, value, inherited \|
	247	\| `timestamps` \| heading_id, kind (scheduled, deadline, closed, active, inactive), start, end, repeater, warning \|
	248	\| `clocks` \| heading_id, start, end, minutes \|
	249	\| `links` \| heading_id, type, target \|
	250	\| `headings_fts` \| FTS5 over title and body text \|
	251
	252	- A file is reindexed when its disk hash changes, and when an app setting that changes semantics (TODO defaults, inheritance policies) changes. Index rows record the settings version they were built with.
	253	- Open documents with unsaved edits overlay the index: their rows are computed in memory from the session's tree and replace that file's disk rows in every query. Agenda and search therefore reflect unsaved TODO, tag and planning edits in explicit-save mode.
	254	- Index rows store positions with the revision they came from. Jumping to a heading checks the target session's revision and re-finds the heading by ID or outline path if the positions are stale.
	255	- Every query the UI runs (agenda, tag search, TODO list, saved views) is a query over the index plus the overlay.
	256
	257	## Babel and table formulas
	258
	259	Both live in `OrgCore` as parsing and planning logic; only the process launch is platform code. Results are written back as a minimal edit to `#+RESULTS:`.
	260
	261	Babel execution (macOS)
	262
	263	- Execution plan first: resolve all header args, then check them before launching anything. `:eval never`/`no` blocks execution; `:eval query` always asks. Any header that changes what runs and isn't supported yet (`:session`, `:noweb yes`, `:prologue`, `:epilogue`, `:file`) stops execution with a message naming it, rather than running different code.
	264	- Runner: write the expanded body to a temp file, launch the interpreter with `Process` in `:dir` (default: the file's folder), capture stdout and stderr, support timeout and cancel.
	265	- Each language has an adapter for `:var` serialization (scalars, lists, tables) and for `:results value` (python wraps the body in a function, as org does; shells take the last output). Interpreter invocation alone doesn't provide org semantics.
	266	- Never execute on open or on export without confirmation. Confirm per block, with a per-file trust setting, like `org-confirm-babel-evaluate`. Trust is keyed to the block's content hash, so an external edit that changes the code asks again.
	267
	268	\| Language \| Runner \| Header args in v1 \|
	269	\| --- \| --- \| --- \|
	270	\| `sh`, `bash`, `shell` \| Native \| `:results`, `:var`, `:dir`, `:cmd`, `:exports` \|
	271	\| `python` \| Native; `:results value` wraps the body in a function, as org does \| Same \|
	272	\| `ruby`, `js`, `R`, `sqlite`, `awk` \| Generic: configured command per language (`ruby`, `node`, `Rscript`, `sqlite3`, `awk`) \| `:results output`, `:dir`, `:cmd` \|
	273	\| `emacs-lisp`, `elisp` \| `emacs --batch` if Emacs is installed \| `:results` \|
	274
	275	- `:results` handling in v1: `output`/`value`, `verbatim`/`table`/`list`/`raw`/`drawer`, `replace`/`append`/`silent`.
	276	- Result ownership, matched to org: the results element is the `#+RESULTS:` keyword plus exactly one element after it (fixed-width lines, a table, a list, an example block, or a `:RESULTS:` drawer). `replace` swaps that element; `append` adds after it; `silent` writes nothing. `raw` output has no boundary, so `raw` without `drawer` refuses to replace existing results and asks first. Changing the format replaces the old element by its old format's boundary.
	277	- Results are re-found on completion as described in Commands; affiliated keywords (`#+NAME`, `#+CAPTION`) on the results stay.
	278	- Tests run the same block repeatedly and switch formats between runs.
	279	- Later: `:session`, `:file` with inline images (dot, plantuml, mermaid), `:noweb`, `:tangle`.
	280	- Highlighting: tree-sitter (SwiftTreeSitter) grammars for about 20 languages: shells, emacs-lisp, python, C, R, js, java, lisp, scheme, clojure, haskell, rust, go, sql, ruby, org, latex, dot, plantuml, mermaid.
	281
	282	Table formula evaluator
	283
	284	The native evaluator handles a defined numeric domain. A table that uses anything outside it is recalculated through Emacs.
	285
	286	- Native domain: integers within ±2^53 and decimals with at most 15 significant digits, using Decimal. Results within the domain must match Calc's output byte for byte; an input or intermediate outside it (overflow, cancellation flagged by precision loss, unsupported mode flags) routes the whole table to Emacs.
	287	- Emacs fallback: run `emacs -Q --batch` on a snapshot of the whole file (so `#+CONSTANTS`, properties, named tables for `remote()` and file-local settings are present) in the file's own folder. Put point in the table, select the chosen `#+TBLFM` line, and call `(org-table-recalculate 'all)`; without that argument it recalculates only the current row. Splice back only the table's text.
	288	- Execution authorization: Lisp formulas (`'(...)`) and anything that evaluates Lisp run arbitrary code, so they need the same confirmation and content-hash trust as Babel. The fallback disables file-local variable evaluation (`enable-local-variables` set to `:safe`).
	289
	290	\| Area \| Native \| Notes \|
	291	\| --- \| --- \| --- \|
	292	\| Arithmetic, `^`, parentheses \| Yes \| \|
	293	\| References: `$N`, `@N`, `@N$M`, `@<`, `@>`, `@I`/`@II`, `$#`, `@#`, relative `@-1` \| Yes \| \|
	294	\| Ranges: `$1..$3`, `@2$1..@>$1` \| Yes \| \|
	295	\| `vsum`, `vmean`, `vmax`, `vmin`, `vcount`, `vmedian` \| Yes \| \|
	296	\| `sqrt`, `exp`, `ln`, `log10`, trig \| Yes \| \|
	297	\| Column vs field formulas, field takes precedence \| Yes \| \|
	298	\| Several `#+TBLFM:` lines, chosen by cursor line \| Yes \| \|
	299	\| Format flags `;%.Nf`, `;N`, `;E` \| Yes \| \|
	300	\| Calc display format (e.g. `1.4142136`) \| Yes, matched \| Must be tested against Emacs output byte for byte. \|
	301	\| Precision \| Bounded \| Native within the defined domain only; outside it, Emacs. \|
	302	\| Dates and durations (`;T`, `;t`, `HH:MM`, timestamp subtraction) \| Later \| Separate piece of work. \|
	303	\| `remote(name, ref)` \| Later \| \|
	304	\| Calc symbolic math: `taylor`, `deriv`, `integ`, `solve` \| No \| Emacs fallback. \|
	305	\| Lisp formulas `'(...)` \| No \| Emacs fallback. \|
	306	\| Calc units, vectors/matrices beyond `v*` \| No \| Emacs fallback. \|
	307	\| Iterate to convergence (`C-u C-u C-c C-c`) \| No \| Emacs fallback. \|
	308
	309	On iOS, tables that need Emacs keep their last computed values and show a "recalculate on Mac" marker.
	310
	311	## Testing
	312
	313	Fidelity is checked against Emacs itself, using the same oracle pattern as orgo's `tests/oracle.rs`.
	314
	315	\| Layer \| Test \| Corpus \|
	316	\| --- \| --- \| --- \|
	317	\| Parser \| Round trip: tree text equals file bytes \| Your org files, org-conformance cases, org manual examples \|
	318	\| Parser \| Incremental equals full: random edits, then compare incremental tree with a fresh parse \| Same, plus fuzzed edits \|
	319	\| Parser \| Conformance: HTML skeleton from an `OrgCore`-based renderer matches the goldens \| org-conformance \|
	320	\| Commands \| Oracle: same file, cursor and command in `emacs --batch`; diff resulting bytes \| Generated cases per command \|
	321	\| Table formulas \| Oracle: `org-table-recalculate` output vs native, byte for byte \| Tables from your files plus a generated set \|
	322	\| Babel \| Results block text vs Emacs for the same block and header args \| sh and python cases \|
	323	\| Performance \| Parse, reparse and restyle times on the largest real files \| Your largest files, plus synthetic 10x copies \|
	324
	325	- Command oracle mapping examples: `org.todo.cycle` → `org-todo`, `org.heading.demote` → `org-metaright`, `org.fill` → `org-fill-paragraph`, `org.table.align` → `org-table-align`.
	326	- The command oracle is stateful. Each case fixes: file, point and mark (converted from UTF-16 offsets to Emacs character positions), prefix argument, active region, the org settings that matter (`org-todo-keywords`, `org-log-done`, `fill-column`, `tab-width`, `indent-tabs-mode`), frozen time via a stubbed `current-time`, locale and time zone, and scripted answers to prompts. Emacs runs with `-Q` plus that explicit configuration.
	327	- Oracle comparisons cover resulting bytes, resulting point and mark, and the text after one undo.
	328	- Pin Emacs 31.1 and Org 9.8.7, recorded in the repo. CI requires the pinned Emacs; a missing oracle fails the job instead of skipping.
	329	- Encoding fixtures (BOM, CRLF, mixed endings, invalid UTF-8, non-BMP, combining characters) are public, in the repo.
	330	- Save fault-injection tests replace the file on disk at each step of the save sequence.
	331	- Personal org files are a second, private corpus run locally; phase gates use the public corpus.
	332
	333	## Phases
	334
	335	Seven phases, each usable on its own. Phase 1 is read-only, so it can run next to Emacs from the first build.
	336
	337	1. Core and viewer: `OrgCore` parser, workspace, index, read-only rendering, folding, search.
	338	2. Editor: text editing, structure and heading commands, timestamps, `M-q`, table alignment, command palette, Emacs and Mac keymaps, remaining tree-sitter grammars.
	339	3. Agenda: day/week views, tag and TODO search, saved custom views, notifications from SCHEDULED/DEADLINE.
	340	4. Computation: native table formulas with Emacs fallback, Babel execution and results.
	341	5. Export and capture: native HTML and Markdown export, pandoc or Emacs for the rest, capture templates, global capture hotkey.
	342	6. Beyond Emacs: clock reports, habit charts, table and kanban views over properties, Doom keymap.
	343	7. iOS: same core; touch input, capture and agenda first.
	344
	345	Phase 1 scope
	346
	347	- [ ] `OrgCore` package: lossless parser with UTF-16 ranges and the bytes-and-encoding contract; element-aware settings pass; inline rules ported from OrgSwift
	348	- [ ] Incremental reparse with old/new classification and synchronization points; differential tests
	349	- [ ] Document session: original bytes, revisions, merge base, per-view state
	350	- [ ] Save path with merge, recovery copies and fault-injection tests (used by the editing spike, not exposed in the viewer)
	351	- [ ] Editable TextKit 2 spike: folding and org-indent display with correct caret movement, selection, IME composition, VoiceOver and copy/paste across folded text
	352	- [ ] Workspace: folders, bookmarks, discovery/agenda/link scopes, FSEvents with reconciliation scans, Syncthing temp/conflict handling, iCloud placeholders
	353	- [ ] Index: schema, settings version, dirty-buffer overlay, FTS5 search
	354	- [ ] Mac app (read-only): sidebar of folders and files, outline of the current file, quick open (`⌘P`), full-text search
	355	- [ ] Rendering over the text: heading styles, TODO/priority/tag styling, links, emphasis, tables, src blocks highlighted for org, emacs-lisp, sh and python only (the rest move to phase 2)
	356	- [ ] Folding: local and global cycling, `#+STARTUP` visibility
	357	- [ ] Benchmarks recorded for the files listed in the exit gates
	358
	359	Phase 1 exit gates (measured on an M1 MacBook Air, 8 GB, the slowest supported reference machine):
	360
	361	\| Gate \| Target \|
	362	\| --- \| --- \|
	363	\| Round trip \| 100% byte-identical on the public corpus and the private corpus \|
	364	\| Incremental equals full \| 0 mismatches over 100,000 fuzzed edits \|
	365	\| Open to first render \| p95 under 100 ms for a 1 MB file; under 500 ms for a 10 MB file \|
	366	\| Keystroke to restyled frame (spike) \| p95 under 16 ms for a 1 MB file, including one file whose content is a single top-level section and one with a 5,000-row table \|
	367	\| Peak memory \| Under 10x file size for an open document \|
	368	\| Index \| Full rebuild of 10,000 files under 30 s; reconciliation after restart under 2 s with no changes \|
	369	\| Save safety \| All fault-injection cases end with both versions recoverable \|
	370	\| Editing spike \| Caret, selection, IME, VoiceOver and copy/paste checks pass on folded and indented text \|
	371
	372	Phase 2 (writable editor) does not start until every gate passes.
	373
	374	## iOS considerations
	375
	376	\| Area \| Mac v1 \| iOS impact \|
	377	\| --- \| --- \| --- \|
	378	\| Core \| `OrgCore`, `OrgIndex` with no AppKit imports \| Reused unchanged. CI builds them for iOS from phase 1 to catch accidental AppKit use. \|
	379	\| Editor view \| TextKit 2 in an `NSTextView` wrapper \| Shared layout policy (what folds, what is indented, how org markup is styled) in a platform-free module; separate AppKit and UIKit adapters for caret movement, selection, input methods and accessibility across hidden text. The phase 1 editing spike keeps that adapter boundary explicit. \|
	380	\| Input \| Keymaps call commands \| Same commands from a keyboard accessory bar, palette and gestures; iPad hardware keyboards use the keymaps. \|
	381	\| Storage \| Bookmarks, FSEvents, `NSFilePresenter` \| Bookmarks and `NSFilePresenter` carry over; Syncthing through Möbius Sync's File Provider. \|
	382	\| Babel \| `Process` \| Maybe. Results blocks display either way. \|
	383	\| Table formulas \| Native plus Emacs fallback \| Native only; Emacs-only tables show a marker. \|
	384	\| Notifications \| Local notifications from the index \| iOS caps pending local notifications at 64 and doesn't guarantee background time to refill them. Schedule the next 64 events, reconcile on every launch, foreground and sync, and show in the agenda how far ahead reminders are scheduled. If the app isn't opened, reminders stop after the last scheduled one, and the UI says so. Repeating requests are used only where they match org repeater semantics. \|
	385
	386	## Open questions
	387
	388	- [ ] App name: Orgstar (chosen).
	389	- [x] Save model: a user setting (see Saving).
	390	- [x] Minimum OS: macOS 26 and iOS 26. One release behind current (macOS 27).
	391	- [x] License: 0BSD.
	392	- [x] Oracle pins: Emacs 31.1 and Org 9.8.7.
	393	- [x] OrgSwift moves onto `OrgCore`'s tree later, tracked in krz/org-swift#2. Out of scope for now.

docs/plans/2026-10-04-orgcore-parser.md added +1600

@@ -0,0 +1,1600 @@
	1	# OrgCore Parser Foundation Implementation Plan
	2
	3	> For agentic workers: REQUIRED SUB-SKILL: Use superpowers:subagent-driven-development (recommended) or superpowers:executing-plans to implement this plan task-by-task. Steps use checkbox (`- [ ]`) syntax for tracking.
	4
	5	Goal: A Swift package, `OrgCore`, that turns org file bytes into a lossless syntax tree of block-level elements, with the byte and encoding contract from the design.
	6
	7	Architecture: Bytes decode into `SourceText` (UTF-8 only, BOM kept, invalid input read-only). The parser splits lines without normalizing endings, classifies each line once, matches block and drawer ends, scans in-buffer TODO settings outside blocks, then builds an immutable green tree through `GreenBuilder`. `SyntaxNode` gives offset-aware red views. Every input byte ends up in exactly one token, so the tree text always equals the source.
	8
	9	Tech Stack: Swift 6.2 tools, Swift Testing, Foundation only. No third-party dependencies.
	10
	11	Spec: `docs/design.md` (sections "OrgCore data model", "Testing", "Phases").
	12
	13	## Global Constraints
	14
	15	- Platforms: macOS 26, iOS 26.
	16	- `OrgCore` imports Foundation only; never AppKit, UIKit or SwiftUI.
	17	- Ranges and offsets exposed by the tree are UTF-16 code units.
	18	- Supported encoding: UTF-8 with or without BOM. Anything else is read-only and never converted.
	19	- Round trip: tree text equals source text for every input, and encoding unchanged text returns the original bytes.
	20	- License 0BSD. No attribution lines in code, commits or docs.
	21
	22	## Out of scope for this plan
	23
	24	Inline objects (emphasis, links, timestamps), the semantic layer, incremental reparse, conformance rendering, and the private-corpus benchmarks. Each gets its own plan.
	25
	26	## File structure
	27
	28	\| File \| Responsibility \|
	29	\| --- \| --- \|
	30	\| `Package.swift` \| Package with `OrgCore` library and `OrgCoreTests` \|
	31	\| `Sources/OrgCore/SourceText.swift` \| Byte decoding, BOM, validity, encoding back to bytes \|
	32	\| `Sources/OrgCore/Syntax/SyntaxKind.swift` \| Token and node kinds \|
	33	\| `Sources/OrgCore/Syntax/GreenTree.swift` \| `GreenToken`, `GreenNode`, `GreenElement`, `GreenBuilder` \|
	34	\| `Sources/OrgCore/Syntax/SyntaxNode.swift` \| Red nodes, tokens, `OrgTree` \|
	35	\| `Sources/OrgCore/Parser/Lines.swift` \| Line splitting and classification \|
	36	\| `Sources/OrgCore/Parser/Settings.swift` \| TODO sequences, priorities, settings scan \|
	37	\| `Sources/OrgCore/Parser/Parser.swift` \| Tree construction \|
	38	\| `Tests/OrgCoreTests/*.swift` \| One test file per source file, plus round-trip fuzz and corpus tests \|
	39
	40	---
	41
	42	### Task 1: Package and SourceText
	43
	44	Files:
	45	- Create: `Package.swift`
	46	- Create: `Sources/OrgCore/SourceText.swift`
	47	- Test: `Tests/OrgCoreTests/SourceTextTests.swift`
	48
	49	Interfaces:
	50	- Produces: `SourceText(bytes: [UInt8])`, `SourceText(_ text: String)`, properties `originalBytes`, `hasBOM`, `isValidUTF8`, `isEditable`, `text`; `encode(_ newText: String) -> [UInt8]`.
	51
	52	- [ ] Step 1: Create the package
	53
	54	```swift
	55	// swift-tools-version: 6.2
	56	import PackageDescription
	57
	58	let package = Package(
	59	name: "Orgstar",
	60	platforms: [.macOS(.v26), .iOS(.v26)],
	61	products: [
	62	.library(name: "OrgCore", targets: ["OrgCore"])
	63	],
	64	targets: [
	65	.target(name: "OrgCore"),
	66	.testTarget(name: "OrgCoreTests", dependencies: ["OrgCore"])
	67	]
	68	)
	69	```
	70
	71	- [ ] Step 2: Write the failing tests
	72
	73	```swift
	74	import Testing
	75	@testable import OrgCore
	76
	77	struct SourceTextTests {
	78	@Test func plainUTF8() {
	79	let source = SourceText(bytes: Array("* a\n".utf8))
	80	#expect(source.text == "* a\n")
	81	#expect(!source.hasBOM)
	82	#expect(source.isEditable)
	83	}
	84
	85	@Test func bomIsStrippedAndRestored() {
	86	let bytes: [UInt8] = [0xEF, 0xBB, 0xBF] + Array("x\n".utf8)
	87	let source = SourceText(bytes: bytes)
	88	#expect(source.hasBOM)
	89	#expect(source.text == "x\n")
	90	#expect(source.encode(source.text) == bytes)
	91	#expect(source.encode("y\n") == [0xEF, 0xBB, 0xBF] + Array("y\n".utf8))
	92	}
	93
	94	@Test func crlfAndMixedEndingsSurvive() {
	95	let bytes = Array("a\r\nb\nc\r\n".utf8)
	96	let source = SourceText(bytes: bytes)
	97	#expect(source.encode(source.text) == bytes)
	98	}
	99
	100	@Test func invalidUTF8IsReadOnlyAndUnchanged() {
	101	let bytes: [UInt8] = [0x61, 0xFF, 0x0A]
	102	let source = SourceText(bytes: bytes)
	103	#expect(!source.isValidUTF8)
	104	#expect(!source.isEditable)
	105	#expect(source.encode(source.text) == bytes)
	106	}
	107
	108	@Test func nonBMPAndCombiningSurvive() {
	109	let bytes = Array("😀 e\u{301}\n".utf8)
	110	let source = SourceText(bytes: bytes)
	111	#expect(source.encode(source.text) == bytes)
	112	}
	113	}
	114	```
	115
	116	- [ ] Step 3: Run tests to verify they fail
	117
	118	Run: `swift test --filter SourceTextTests`
	119	Expected: build failure, `cannot find 'SourceText' in scope`.
	120
	121	- [ ] Step 4: Implement
	122
	123	```swift
	124	import Foundation
	125
	126	/// A file's bytes and their decoded text. Only UTF-8 (with or without a BOM) is editable;
	127	/// anything else decodes with replacement characters for display and is never written back.
	128	public struct SourceText: Sendable {
	129	public let originalBytes: [UInt8]
	130	public let hasBOM: Bool
	131	public let isValidUTF8: Bool
	132	/// Decoded text without the BOM.
	133	public let text: String
	134
	135	private static let bom: [UInt8] = [0xEF, 0xBB, 0xBF]
	136
	137	public init(bytes: [UInt8]) {
	138	originalBytes = bytes
	139	hasBOM = bytes.starts(with: Self.bom)
	140	let body = hasBOM ? Array(bytes.dropFirst(3)) : bytes
	141	if let decoded = String(validating: body, as: UTF8.self) {
	142	text = decoded
	143	isValidUTF8 = true
	144	} else {
	145	text = String(decoding: body, as: UTF8.self)
	146	isValidUTF8 = false
	147	}
	148	}
	149
	150	public init(_ text: String) {
	151	self.init(bytes: Array(text.utf8))
	152	}
	153
	154	public var isEditable: Bool { isValidUTF8 }
	155
	156	/// Bytes to write for `newText`. Unchanged text returns the original bytes. Valid UTF-8
	157	/// round-trips through `String` unchanged, so untouched spans keep their exact bytes.
	158	public func encode(_ newText: String) -> [UInt8] {
	159	if newText == text { return originalBytes }
	160	precondition(isEditable, "a read-only document cannot be re-encoded")
	161	return (hasBOM ? Self.bom : []) + Array(newText.utf8)
	162	}
	163	}
	164	```
	165
	166	- [ ] Step 5: Run tests to verify they pass
	167
	168	Run: `swift test --filter SourceTextTests`
	169	Expected: 5 tests pass.
	170
	171	- [ ] Step 6: Commit
	172
	173	```bash
	174	git add Package.swift Sources Tests
	175	git commit -m "Add OrgCore package and SourceText byte contract"
	176	```
	177
	178	---
	179
	180	### Task 2: Green and red syntax tree
	181
	182	Files:
	183	- Create: `Sources/OrgCore/Syntax/SyntaxKind.swift`
	184	- Create: `Sources/OrgCore/Syntax/GreenTree.swift`
	185	- Create: `Sources/OrgCore/Syntax/SyntaxNode.swift`
	186	- Test: `Tests/OrgCoreTests/SyntaxTreeTests.swift`
	187
	188	Interfaces:
	189	- Produces: `SyntaxKind` (enum, `String` raw values), `GreenToken(kind:text:)`, `GreenNode(kind:children:)` with `.length`, `.text`; `GreenElement` (`.node`, `.token`); internal `GreenBuilder` with `start(_:)`, `token(_:_:)`, `finish()`, `build()`; `SyntaxNode` with `kind`, `range`, `text`, `children`, `tokens`, `descendants()`; `SyntaxToken`. `OrgTree` depends on `OrgSettings` (Task 4), so it is added in Task 5.
	190
	191	- [ ] Step 1: Write the failing tests
	192
	193	```swift
	194	import Testing
	195	@testable import OrgCore
	196
	197	struct SyntaxTreeTests {
	198	func sample() -> GreenNode {
	199	var b = GreenBuilder()
	200	b.start(.document)
	201	b.start(.paragraph)
	202	b.token(.text, "hé😀")
	203	b.token(.newline, "\n")
	204	b.finish()
	205	b.token(.newline, "\r\n")
	206	b.finish()
	207	return b.build()
	208	}
	209
	210	@Test func lengthsAreUTF16() {
	211	let green = sample()
	212	#expect(green.length == 4 + 1 + 2)
	213	#expect(green.text == "hé😀\n\r\n")
	214	}
	215
	216	@Test func redNodesCarryOffsets() {
	217	let root = SyntaxNode(green: sample(), offset: 0, parent: nil)
	218	let paragraph = root.children[0]
	219	#expect(paragraph.kind == .paragraph)
	220	#expect(paragraph.range == 0..<5)
	221	#expect(paragraph.parent === root)
	222	#expect(root.tokens.map(\.range) == [5..<7])
	223	#expect(paragraph.tokens.map(\.kind) == [.text, .newline])
	224	}
	225
	226	@Test func builderSkipsEmptyTokens() {
	227	var b = GreenBuilder()
	228	b.start(.document)
	229	b.token(.whitespace, "")
	230	b.finish()
	231	#expect(b.build().children.isEmpty)
	232	}
	233
	234	@Test func descendantsArePreorder() {
	235	let root = SyntaxNode(green: sample(), offset: 0, parent: nil)
	236	#expect(root.descendants().map(\.kind) == [.document, .paragraph])
	237	}
	238	}
	239	```
	240
	241	- [ ] Step 2: Run tests to verify they fail
	242
	243	Run: `swift test --filter SyntaxTreeTests`
	244	Expected: build failure, `cannot find 'GreenBuilder' in scope`.
	245
	246	- [ ] Step 3: Implement `SyntaxKind.swift`
	247
	248	```swift
	249	public enum SyntaxKind: String, Sendable {
	250	// Tokens
	251	case text, newline, whitespace
	252	case stars, todoKeyword, priority, title, tags
	253
	254	// Nodes
	255	case document, zerothSection, section, heading
	256	case planning, propertyDrawer, nodeProperty, drawer, clock
	257	case paragraph, plainList, item, table, tableRow, tableFormula
	258	case block, dynamicBlock, keyword, affiliatedKeyword
	259	case comment, fixedWidth, horizontalRule, footnoteDefinition
	260	}
	261	```
	262
	263	- [ ] Step 4: Implement `GreenTree.swift`
	264
	265	```swift
	266	public struct GreenToken: Sendable, Equatable {
	267	public let kind: SyntaxKind
	268	public let text: String
	269	/// Length in UTF-16 code units.
	270	public let length: Int
	271
	272	public init(kind: SyntaxKind, text: String) {
	273	self.kind = kind
	274	self.text = text
	275	self.length = text.utf16.count
	276	}
	277	}
	278
	279	/// An immutable node. Stores only kind, children and length, so unchanged subtrees can be
	280	/// shared between versions of a document.
	281	public final class GreenNode: Sendable, Equatable {
	282	public let kind: SyntaxKind
	283	public let children: [GreenElement]
	284	/// Length in UTF-16 code units.
	285	public let length: Int
	286
	287	public init(kind: SyntaxKind, children: [GreenElement]) {
	288	self.kind = kind
	289	self.children = children
	290	self.length = children.reduce(0) { $0 + $1.length }
	291	}
	292
	293	public static func == (lhs: GreenNode, rhs: GreenNode) -> Bool {
	294	lhs === rhs \|\| (lhs.kind == rhs.kind && lhs.children == rhs.children)
	295	}
	296
	297	public var text: String {
	298	var out = ""
	299	write(to: &out)
	300	return out
	301	}
	302
	303	func write(to out: inout String) {
	304	for child in children {
	305	switch child {
	306	case .node(let node): node.write(to: &out)
	307	case .token(let token): out += token.text
	308	}
	309	}
	310	}
	311	}
	312
	313	public enum GreenElement: Sendable, Equatable {
	314	case node(GreenNode)
	315	case token(GreenToken)
	316
	317	public var length: Int {
	318	switch self {
	319	case .node(let node): node.length
	320	case .token(let token): token.length
	321	}
	322	}
	323	}
	324
	325	struct GreenBuilder {
	326	private var stack: [(kind: SyntaxKind, children: [GreenElement])] = []
	327	private var root: GreenNode?
	328
	329	mutating func start(_ kind: SyntaxKind) {
	330	stack.append((kind, []))
	331	}
	332
	333	mutating func token(_ kind: SyntaxKind, _ text: some StringProtocol) {
	334	guard !text.isEmpty else { return }
	335	stack[stack.count - 1].children.append(.token(GreenToken(kind: kind, text: String(text))))
	336	}
	337
	338	mutating func finish() {
	339	let (kind, children) = stack.removeLast()
	340	let node = GreenNode(kind: kind, children: children)
	341	if stack.isEmpty {
	342	root = node
	343	} else {
	344	stack[stack.count - 1].children.append(.node(node))
	345	}
	346	}
	347
	348	func build() -> GreenNode {
	349	precondition(stack.isEmpty, "unfinished nodes")
	350	return root!
	351	}
	352	}
	353	```
	354
	355	- [ ] Step 5: Implement `SyntaxNode.swift`
	356
	357	```swift
	358	/// A view of a green node at an absolute offset, with a parent link. Created on demand.
	359	public final class SyntaxNode: Sendable {
	360	public let green: GreenNode
	361	public let offset: Int
	362	public let parent: SyntaxNode?
	363
	364	init(green: GreenNode, offset: Int, parent: SyntaxNode?) {
	365	self.green = green
	366	self.offset = offset
	367	self.parent = parent
	368	}
	369
	370	public var kind: SyntaxKind { green.kind }
	371	public var range: Range<Int> { offset..<(offset + green.length) }
	372	public var text: String { green.text }
	373
	374	public var children: [SyntaxNode] {
	375	var result: [SyntaxNode] = []
	376	var at = offset
	377	for child in green.children {
	378	if case .node(let node) = child {
	379	result.append(SyntaxNode(green: node, offset: at, parent: self))
	380	}
	381	at += child.length
	382	}
	383	return result
	384	}
	385
	386	public var tokens: [SyntaxToken] {
	387	var result: [SyntaxToken] = []
	388	var at = offset
	389	for child in green.children {
	390	if case .token(let token) = child {
	391	result.append(SyntaxToken(kind: token.kind, text: token.text, range: at..<(at + token.length)))
	392	}
	393	at += child.length
	394	}
	395	return result
	396	}
	397
	398	/// This node and every node below it, in document order.
	399	public func descendants() -> [SyntaxNode] {
	400	[self] + children.flatMap { $0.descendants() }
	401	}
	402	}
	403
	404	public struct SyntaxToken: Sendable, Equatable {
	405	public let kind: SyntaxKind
	406	public let text: String
	407	public let range: Range<Int>
	408	}
	409	```
	410
	411	- [ ] Step 6: Run tests to verify they pass
	412
	413	Run: `swift test --filter SyntaxTreeTests`
	414	Expected: 4 tests pass.
	415
	416	- [ ] Step 7: Commit
	417
	418	```bash
	419	git add Sources Tests
	420	git commit -m "Add green and red syntax tree"
	421	```
	422
	423	---
	424
	425	### Task 3: Line splitting and classification
	426
	427	Files:
	428	- Create: `Sources/OrgCore/Parser/Lines.swift`
	429	- Test: `Tests/OrgCoreTests/LinesTests.swift`
	430
	431	Interfaces:
	432	- Produces (internal): `RawLine(content: Substring, ending: Substring)`, `splitRawLines(_ text: String) -> [RawLine]`, `LineClass` enum, `ClassifiedLine(cls:indent:)`, `classifyLine(_ line: Substring) -> ClassifiedLine`, `Substring.trimmingTrailingWhitespace`.
	433
	434	- [ ] Step 1: Write the failing tests
	435
	436	```swift
	437	import Testing
	438	@testable import OrgCore
	439
	440	struct LinesTests {
	441	@Test func splitKeepsEveryEnding() {
	442	let lines = splitRawLines("a\r\nb\n\nc")
	443	#expect(lines.map { String($0.content) } == ["a", "b", "", "c"])
	444	#expect(lines.map { String($0.ending) } == ["\r\n", "\n", "\n", ""])
	445	#expect(splitRawLines("").isEmpty)
	446	#expect(splitRawLines("x\n").count == 1)
	447	}
	448
	449	@Test func splitIsLossless() {
	450	let text = "\r\n\n a\r b\r\n😀\n"
	451	#expect(splitRawLines(text).map { String($0.content) + String($0.ending) }.joined() == text)
	452	}
	453
	454	@Test(arguments: [
	455	("", LineClass.blank),
	456	(" \t", .blank),
	457	("* a", .heading(level: 1)),
	458	("*** ", .heading(level: 3)),
	459	("*", .heading(level: 1)),
	460	("bold text", .plain),
	461	(" * a", .listItem),
	462	("#+BEGIN_SRC sh :results output", .blockBegin(name: "src")),
	463	("#+end_src", .blockEnd(name: "src")),
	464	("#+BEGIN: clocktable :scope file", .dynamicBegin),
	465	("#+END:", .dynamicEnd),
	466	("#+TITLE: x", .keyword(key: "TITLE")),
	467	("#+tblfm: $2=$1", .keyword(key: "TBLFM")),
	468	("# comment", .comment),
	469	("#", .comment),
	470	("#hashtag", .plain),
	471	(": fixed", .fixedWidth),
	472	(":", .fixedWidth),
	473	(":PROPERTIES:", .drawerBegin(name: "PROPERTIES")),
	474	(" :LOGBOOK:", .drawerBegin(name: "LOGBOOK")),
	475	(":END:", .drawerEnd),
	476	("\| a \| b \|", .tableRow),
	477	("-----", .horizontalRule),
	478	("----", .plain),
	479	("[fn:1] note", .footnoteDefinition),
	480	("CLOCK: [2026-10-04 Sun 10:00]", .clock),
	481	("SCHEDULED: <2026-10-04 Sun>", .planning),
	482	("- item", .listItem),
	483	("+ item", .listItem),
	484	("1. item", .listItem),
	485	("2) item", .listItem),
	486	("-", .listItem),
	487	("-x", .plain),
	488	("1.5 apples", .plain),
	489	("plain text", .plain),
	490	])
	491	func classify(line: String, expected: LineClass) {
	492	#expect(classifyLine(line[...]).cls == expected)
	493	}
	494
	495	@Test func indentCountsTabsToEight() {
	496	#expect(classifyLine("\t- a").indent == 8)
	497	#expect(classifyLine(" \t- a").indent == 8)
	498	#expect(classifyLine(" - a").indent == 3)
	499	}
	500	}
	501	```
	502
	503	- [ ] Step 2: Run tests to verify they fail
	504
	505	Run: `swift test --filter LinesTests`
	506	Expected: build failure, `cannot find 'splitRawLines' in scope`.
	507
	508	- [ ] Step 3: Implement
	509
	510	```swift
	511	/// One line of source: its content and its terminator ("", "\n" or "\r\n"), both as slices of
	512	/// the original text.
	513	struct RawLine {
	514	let content: Substring
	515	let ending: Substring
	516	}
	517
	518	/// Splits on "\n" without normalizing anything. Works on unicode scalars, because String
	519	/// treats "\r\n" as a single Character.
	520	func splitRawLines(_ text: String) -> [RawLine] {
	521	var lines: [RawLine] = []
	522	let scalars = text.unicodeScalars
	523	var lineStart = scalars.startIndex
	524	var i = lineStart
	525	while i != scalars.endIndex {
	526	if scalars[i] == "\n" {
	527	var contentEnd = i
	528	if contentEnd > lineStart, scalars[scalars.index(before: i)] == "\r" {
	529	contentEnd = scalars.index(before: i)
	530	}
	531	let next = scalars.index(after: i)
	532	lines.append(RawLine(content: text[lineStart..<contentEnd], ending: text[contentEnd..<next]))
	533	lineStart = next
	534	i = next
	535	} else {
	536	i = scalars.index(after: i)
	537	}
	538	}
	539	if lineStart != scalars.endIndex {
	540	lines.append(RawLine(content: text[lineStart...], ending: ""))
	541	}
	542	return lines
	543	}
	544
	545	enum LineClass: Equatable {
	546	case blank
	547	case heading(level: Int)
	548	case blockBegin(name: String)
	549	case blockEnd(name: String)
	550	case dynamicBegin
	551	case dynamicEnd
	552	case drawerBegin(name: String)
	553	case drawerEnd
	554	case keyword(key: String)
	555	case comment
	556	case fixedWidth
	557	case horizontalRule
	558	case tableRow
	559	case footnoteDefinition
	560	case clock
	561	case planning
	562	case listItem
	563	case plain
	564	}
	565
	566	struct ClassifiedLine {
	567	let cls: LineClass
	568	/// Column of the first non-blank character, with tabs advancing to the next multiple of 8.
	569	let indent: Int
	570	}
	571
	572	func classifyLine(_ line: Substring) -> ClassifiedLine {
	573	var column = 0
	574	var rest = line
	575	while let c = rest.first, c == " " \|\| c == "\t" {
	576	column = c == "\t" ? (column / 8 + 1) * 8 : column + 1
	577	rest = rest.dropFirst()
	578	}
	579	if rest.isEmpty { return ClassifiedLine(cls: .blank, indent: column) }
	580	return ClassifiedLine(cls: lineClass(rest, columnZero: column == 0), indent: column)
	581	}
	582
	583	private func lineClass(_ rest: Substring, columnZero: Bool) -> LineClass {
	584	let trimmed = rest.trimmingTrailingWhitespace
	585
	586	if columnZero, rest.first == "*" {
	587	let stars = rest.prefix { $0 == "*" }
	588	let after = rest.dropFirst(stars.count)
	589	if after.isEmpty \|\| after.first == " " \|\| after.first == "\t" {
	590	return .heading(level: stars.count)
	591	}
	592	}
	593
	594	if rest.hasPrefix("#+") {
	595	let lower = trimmed.lowercased()
	596	if lower.hasPrefix("#+begin_") {
	597	let name = lower.dropFirst(8).prefix { !$0.isWhitespace }
	598	if !name.isEmpty { return .blockBegin(name: String(name)) }
	599	}
	600	if lower.hasPrefix("#+end_") {
	601	let name = lower.dropFirst(6)
	602	if !name.isEmpty, !name.contains(where: \.isWhitespace) { return .blockEnd(name: String(name)) }
	603	}
	604	if lower.hasPrefix("#+begin:") { return .dynamicBegin }
	605	if lower == "#+end:" { return .dynamicEnd }
	606	if let colon = rest.firstIndex(of: ":") {
	607	let key = rest[rest.index(rest.startIndex, offsetBy: 2)..<colon]
	608	if !key.isEmpty, !key.contains(where: \.isWhitespace) { return .keyword(key: key.uppercased()) }
	609	}
	610	}
	611
	612	if trimmed == "#" \|\| rest.hasPrefix("# ") \|\| rest.hasPrefix("#\t") { return .comment }
	613
	614	if rest.first == ":" {
	615	if trimmed == ":" \|\| rest.hasPrefix(": ") \|\| rest.hasPrefix(":\t") { return .fixedWidth }
	616	if trimmed.uppercased() == ":END:" { return .drawerEnd }
	617	if trimmed.count >= 3, trimmed.last == ":" {
	618	let name = trimmed.dropFirst().dropLast()
	619	if name.allSatisfy({ $0.isLetter \|\| $0.isNumber \|\| $0 == "_" \|\| $0 == "-" }) {
	620	return .drawerBegin(name: String(name))
	621	}
	622	}
	623	}
	624
	625	if rest.first == "\|" { return .tableRow }
	626	if trimmed.count >= 5, trimmed.allSatisfy({ $0 == "-" }) { return .horizontalRule }
	627
	628	if columnZero, rest.hasPrefix("[fn:"), let close = rest.firstIndex(of: "]"),
	629	close > rest.index(rest.startIndex, offsetBy: 4) {
	630	return .footnoteDefinition
	631	}
	632
	633	if rest.hasPrefix("CLOCK:") { return .clock }
	634	if rest.hasPrefix("SCHEDULED:") \|\| rest.hasPrefix("DEADLINE:") \|\| rest.hasPrefix("CLOSED:") { return .planning }
	635	if isListBullet(rest, indented: !columnZero) { return .listItem }
	636	return .plain
	637	}
	638
	639	/// `-`, `+`, `*` (indented only), `1.` or `1)`, followed by whitespace or end of line.
	640	/// Alphabetical bullets are off, as in org's default.
	641	private func isListBullet(_ rest: Substring, indented: Bool) -> Bool {
	642	guard let first = rest.first else { return false }
	643	let afterBullet: Substring
	644	if first == "-" \|\| first == "+" \|\| (first == "*" && indented) {
	645	afterBullet = rest.dropFirst()
	646	} else if first.isASCII, first.isNumber {
	647	let digits = rest.prefix { $0.isASCII && $0.isNumber }
	648	let tail = rest.dropFirst(digits.count)
	649	guard let separator = tail.first, separator == "." \|\| separator == ")" else { return false }
	650	afterBullet = tail.dropFirst()
	651	} else {
	652	return false
	653	}
	654	return afterBullet.isEmpty \|\| afterBullet.first == " " \|\| afterBullet.first == "\t"
	655	}
	656
	657	extension Substring {
	658	var trimmingTrailingWhitespace: Substring {
	659	var s = self
	660	while let last = s.last, last == " " \|\| last == "\t" { s = s.dropLast() }
	661	return s
	662	}
	663	}
	664	```
	665
	666	- [ ] Step 4: Run tests to verify they pass
	667
	668	Run: `swift test --filter LinesTests`
	669	Expected: all pass.
	670
	671	- [ ] Step 5: Commit
	672
	673	```bash
	674	git add Sources Tests
	675	git commit -m "Add line splitting and classification"
	676	```
	677
	678	---
	679
	680	### Task 4: In-buffer settings
	681
	682	Files:
	683	- Create: `Sources/OrgCore/Parser/Settings.swift`
	684	- Test: `Tests/OrgCoreTests/SettingsTests.swift`
	685
	686	Interfaces:
	687	- Consumes: `RawLine`, `ClassifiedLine`, `LineClass` (Task 3).
	688	- Produces: `TodoKeyword`, `TodoSequence`, `Priorities`, `OrgSettings` with `.default`, `todoKeywordNames: Set<String>`, `isDone(_:)`; internal `SettingsScanner.scan(lines:info:blockEnds:defaults:) -> OrgSettings`, where `blockEnds: [Int: Int]` maps a block's begin line index to its end line index.
	689
	690	- [ ] Step 1: Write the failing tests
	691
	692	```swift
	693	import Testing
	694	@testable import OrgCore
	695
	696	struct SettingsTests {
	697	func scan(_ text: String, blockEnds: [Int: Int] = [:]) -> OrgSettings {
	698	let lines = splitRawLines(text)
	699	let info = lines.map { classifyLine($0.content) }
	700	return SettingsScanner.scan(lines: lines, info: info, blockEnds: blockEnds, defaults: .default)
	701	}
	702
	703	@Test func defaultsWithoutKeywords() {
	704	let settings = scan("* TODO a\n")
	705	#expect(settings.todoKeywordNames == ["TODO", "DONE"])
	706	#expect(settings.isDone("DONE"))
	707	}
	708
	709	@Test func fileKeywordsReplaceDefaults() {
	710	let settings = scan("#+TODO: NEXT(n) WAIT(w@/!) \| DONE(d!) CANCELED(c@)\n")
	711	#expect(settings.todoKeywordNames == ["NEXT", "WAIT", "DONE", "CANCELED"])
	712	let sequence = settings.todoSequences[0]
	713	#expect(sequence.active.map(\.name) == ["NEXT", "WAIT"])
	714	#expect(sequence.done.map(\.name) == ["DONE", "CANCELED"])
	715	#expect(sequence.active[1] == TodoKeyword(name: "WAIT", fastKey: "w", logOnEnter: "@", logOnLeave: "!"))
	716	#expect(sequence.done[0] == TodoKeyword(name: "DONE", fastKey: "d", logOnEnter: "!", logOnLeave: nil))
	717	}
	718
	719	@Test func lastWordIsDoneWithoutSeparator() {
	720	let settings = scan("#+SEQ_TODO: A B C\n")
	721	#expect(settings.todoSequences[0].active.map(\.name) == ["A", "B"])
	722	#expect(settings.todoSequences[0].done.map(\.name) == ["C"])
	723	}
	724
	725	@Test func severalLinesMakeSeveralSequences() {
	726	let settings = scan("#+TODO: A \| B\n#+TYP_TODO: X \| Y\n")
	727	#expect(settings.todoSequences.count == 2)
	728	#expect(settings.todoSequences[1].kind == .type)
	729	}
	730
	731	@Test func keywordsInsideBlocksAreIgnored() {
	732	let text = "#+begin_example\n#+TODO: X \| Y\n#+end_example\n"
	733	#expect(scan(text, blockEnds: [0: 2]).todoKeywordNames == ["TODO", "DONE"])
	734	}
	735
	736	@Test func priorities() {
	737	#expect(scan("#+PRIORITIES: 1 10 5\n").priorities == Priorities(highest: "1", lowest: "10", default: "5"))
	738	#expect(scan("").priorities == Priorities(highest: "A", lowest: "C", default: "B"))
	739	}
	740	}
	741	```
	742
	743	- [ ] Step 2: Run tests to verify they fail
	744
	745	Run: `swift test --filter SettingsTests`
	746	Expected: build failure, `cannot find 'SettingsScanner' in scope`.
	747
	748	- [ ] Step 3: Implement
	749
	750	```swift
	751	public struct TodoKeyword: Sendable, Hashable {
	752	public var name: String
	753	public var fastKey: Character?
	754	/// Logging flag when entering the state (`!` or `@`), from `NAME(k!/@)`.
	755	public var logOnEnter: String?
	756	/// Logging flag when leaving the state.
	757	public var logOnLeave: String?
	758
	759	public init(name: String, fastKey: Character? = nil, logOnEnter: String? = nil, logOnLeave: String? = nil) {
	760	self.name = name
	761	self.fastKey = fastKey
	762	self.logOnEnter = logOnEnter
	763	self.logOnLeave = logOnLeave
	764	}
	765	}
	766
	767	public struct TodoSequence: Sendable, Equatable {
	768	public enum Kind: Sendable, Equatable { case sequence, type }
	769
	770	public var kind: Kind
	771	public var active: [TodoKeyword]
	772	public var done: [TodoKeyword]
	773
	774	public init(kind: Kind, active: [TodoKeyword], done: [TodoKeyword]) {
	775	self.kind = kind
	776	self.active = active
	777	self.done = done
	778	}
	779	}
	780
	781	public struct Priorities: Sendable, Equatable {
	782	public var highest: String
	783	public var lowest: String
	784	public var `default`: String
	785
	786	public init(highest: String, lowest: String, default: String) {
	787	self.highest = highest
	788	self.lowest = lowest
	789	self.default = `default`
	790	}
	791	}
	792
	793	public struct OrgSettings: Sendable, Equatable {
	794	public var todoSequences: [TodoSequence]
	795	public var priorities: Priorities
	796
	797	public init(todoSequences: [TodoSequence], priorities: Priorities) {
	798	self.todoSequences = todoSequences
	799	self.priorities = priorities
	800	}
	801
	802	public static let `default` = OrgSettings(
	803	todoSequences: [TodoSequence(kind: .sequence, active: [TodoKeyword(name: "TODO")], done: [TodoKeyword(name: "DONE")])],
	804	priorities: Priorities(highest: "A", lowest: "C", default: "B")
	805	)
	806
	807	public var todoKeywordNames: Set<String> {
	808	Set(todoSequences.flatMap { ($0.active + $0.done).map(\.name) })
	809	}
	810
	811	public func isDone(_ name: String) -> Bool {
	812	todoSequences.contains { $0.done.contains { $0.name == name } }
	813	}
	814	}
	815
	816	enum SettingsScanner {
	817	/// Reads `#+TODO`, `#+SEQ_TODO`, `#+TYP_TODO` and `#+PRIORITIES` outside blocks. Any TODO
	818	/// line replaces the default sequences, as in org.
	819	static func scan(lines: [RawLine], info: [ClassifiedLine], blockEnds: [Int: Int], defaults: OrgSettings) -> OrgSettings {
	820	var sequences: [TodoSequence] = []
	821	var priorities = defaults.priorities
	822	var k = 0
	823	while k < lines.count {
	824	switch info[k].cls {
	825	case .blockBegin, .dynamicBegin:
	826	if let end = blockEnds[k] { k = end }
	827	case .keyword(let key):
	828	let value = keywordValue(lines[k].content)
	829	switch key {
	830	case "TODO", "SEQ_TODO":
	831	if let s = todoSequence(value, kind: .sequence) { sequences.append(s) }
	832	case "TYP_TODO":
	833	if let s = todoSequence(value, kind: .type) { sequences.append(s) }
	834	case "PRIORITIES":
	835	let words = value.split(whereSeparator: \.isWhitespace)
	836	if words.count == 3 {
	837	priorities = Priorities(highest: String(words[0]), lowest: String(words[1]), default: String(words[2]))
	838	}
	839	default:
	840	break
	841	}
	842	default:
	843	break
	844	}
	845	k += 1
	846	}
	847	return OrgSettings(todoSequences: sequences.isEmpty ? defaults.todoSequences : sequences, priorities: priorities)
	848	}
	849
	850	static func keywordValue(_ line: Substring) -> Substring {
	851	guard let colon = line.firstIndex(of: ":") else { return "" }
	852	return line[line.index(after: colon)...]
	853	}
	854
	855	static func todoSequence(_ value: Substring, kind: TodoSequence.Kind) -> TodoSequence? {
	856	let words = value.split(whereSeparator: \.isWhitespace)
	857	guard !words.isEmpty else { return nil }
	858	if let bar = words.firstIndex(of: "\|") {
	859	return TodoSequence(kind: kind, active: words[..<bar].map(todoKeyword), done: words[(bar + 1)...].map(todoKeyword))
	860	}
	861	return TodoSequence(kind: kind, active: words.dropLast().map(todoKeyword), done: [todoKeyword(words.last!)])
	862	}
	863
	864	/// `NAME`, or `NAME(spec)` where spec is an optional fast key followed by `enter/leave`
	865	/// logging flags.
	866	static func todoKeyword(_ word: Substring) -> TodoKeyword {
	867	guard let open = word.firstIndex(of: "("), word.last == ")" else { return TodoKeyword(name: String(word)) }
	868	var spec = word[word.index(after: open)..<word.index(before: word.endIndex)]
	869	var fastKey: Character?
	870	if let first = spec.first, first != "!", first != "@", first != "/" {
	871	fastKey = first
	872	spec = spec.dropFirst()
	873	}
	874	let parts = spec.split(separator: "/", omittingEmptySubsequences: false)
	875	let enter = parts.first.flatMap { $0.isEmpty ? nil : String($0) }
	876	let leave = parts.count > 1 && !parts[1].isEmpty ? String(parts[1]) : nil
	877	return TodoKeyword(name: String(word[..<open]), fastKey: fastKey, logOnEnter: enter, logOnLeave: leave)
	878	}
	879	}
	880	```
	881
	882	- [ ] Step 4: Run tests to verify they pass
	883
	884	Run: `swift test --filter SettingsTests`
	885	Expected: all pass.
	886
	887	- [ ] Step 5: Commit
	888
	889	```bash
	890	git add Sources Tests
	891	git commit -m "Add in-buffer TODO and priority settings"
	892	```
	893
	894	---
	895
	896	### Task 5: Parser — document, sections and headings
	897
	898	Files:
	899	- Create: `Sources/OrgCore/Parser/Parser.swift`
	900	- Modify: `Sources/OrgCore/Syntax/SyntaxNode.swift` (append `OrgTree`)
	901	- Test: `Tests/OrgCoreTests/ParserSectionTests.swift`
	902
	903	Interfaces:
	904	- Consumes: Tasks 2–4.
	905	- Produces: `public enum OrgParser { static func parse(_ text: String, defaults: OrgSettings = .default) -> OrgTree }`; `public struct OrgTree { green: GreenNode; settings: OrgSettings; root: SyntaxNode; text: String }`; internal `struct Parser` with `element(limit:floor:)` that Task 6 fills in. In this task `element` handles every class as a paragraph or a blank line.
	906
	907	- [ ] Step 1: Write the failing tests
	908
	909	```swift
	910	import Testing
	911	@testable import OrgCore
	912
	913	func nodeKinds(_ text: String) -> [SyntaxKind] {
	914	OrgParser.parse(text).root.descendants().map(\.kind)
	915	}
	916
	917	func tokens(of kind: SyntaxKind, in text: String) -> [SyntaxToken] {
	918	OrgParser.parse(text).root.descendants().filter { $0.kind == kind }.flatMap(\.tokens)
	919	}
	920
	921	struct ParserSectionTests {
	922	@Test func emptyDocument() {
	923	let tree = OrgParser.parse("")
	924	#expect(tree.text == "")
	925	#expect(nodeKinds("") == [.document])
	926	}
	927
	928	@Test func zerothSectionHoldsPreamble() {
	929	#expect(nodeKinds("text\n* a\n") == [.document, .zerothSection, .paragraph, .section, .heading])
	930	}
	931
	932	@Test func sectionsNestByLevel() {
	933	let text = "* a\n b\n* c\n** d\n* e\n"
	934	let root = OrgParser.parse(text).root
	935	let top = root.children
	936	#expect(top.map(\.kind) == [.section, .section])
	937	#expect(top[0].children.map(\.kind) == [.heading, .section, .section])
	938	#expect(top[0].children[1].children.map(\.kind) == [.heading, .section])
	939	}
	940
	941	@Test func headingTokens() {
	942	let parts = tokens(of: .heading, in: "** TODO [#A] Write the plan :work:urgent: \n")
	943	#expect(parts.map(\.kind) == [.stars, .whitespace, .todoKeyword, .whitespace, .priority, .whitespace, .title, .whitespace, .tags, .whitespace, .newline])
	944	#expect(parts.first { $0.kind == .tags }?.text == ":work:urgent:")
	945	#expect(parts.first { $0.kind == .title }?.text == "Write the plan")
	946	}
	947
	948	@Test func todoKeywordsComeFromSettings() {
	949	let text = "#+TODO: NEXT \| DONE\n* NEXT a\n* TODO b\n"
	950	let todo = tokens(of: .heading, in: text).filter { $0.kind == .todoKeyword }.map(\.text)
	951	#expect(todo == ["NEXT"])
	952	}
	953
	954	@Test func priorityNeedsValidValueAndSpace() {
	955	#expect(tokens(of: .heading, in: "* [#B] x\n").contains { $0.kind == .priority })
	956	#expect(tokens(of: .heading, in: "* [#10] x\n").contains { $0.kind == .priority })
	957	#expect(!tokens(of: .heading, in: "* [#AB] x\n").contains { $0.kind == .priority })
	958	#expect(!tokens(of: .heading, in: "* [#A]x\n").contains { $0.kind == .priority })
	959	}
	960
	961	@Test func tagsNeedValidCharacters() {
	962	#expect(tokens(of: .heading, in: "* a :b@c_1:\n").contains { $0.kind == .tags })
	963	#expect(!tokens(of: .heading, in: "* a :b c:\n").contains { $0.kind == .tags })
	964	#expect(!tokens(of: .heading, in: "* a :b:c\n").contains { $0.kind == .tags })
	965	}
	966
	967	@Test func headingWithoutNewlineAtEnd() {
	968	#expect(OrgParser.parse("* a").text == "* a")
	969	}
	970
	971	@Test(arguments: ["* a\n", "\n", " TODO\n", "text\r\n* a\r\n** b\r\n", "\n\n* a\n\n"])
	972	func roundTrip(text: String) {
	973	#expect(OrgParser.parse(text).text == text)
	974	}
	975	}
	976	```
	977
	978	- [ ] Step 2: Run tests to verify they fail
	979
	980	Run: `swift test --filter ParserSectionTests`
	981	Expected: build failure, `cannot find 'OrgParser' in scope`.
	982
	983	- [ ] Step 3: Append `OrgTree` to `SyntaxNode.swift`
	984
	985	```swift
	986	public struct OrgTree: Sendable {
	987	public let green: GreenNode
	988	public let settings: OrgSettings
	989
	990	public var root: SyntaxNode { SyntaxNode(green: green, offset: 0, parent: nil) }
	991	public var text: String { green.text }
	992	}
	993	```
	994
	995	- [ ] Step 4: Implement `Parser.swift`
	996
	997	```swift
	998	public enum OrgParser {
	999	public static func parse(_ text: String, defaults: OrgSettings = .default) -> OrgTree {
	1000	var parser = Parser(text: text, defaults: defaults)
	1001	return parser.run()
	1002	}
	1003	}
	1004
	1005	struct Parser {
	1006	let lines: [RawLine]
	1007	let info: [ClassifiedLine]
	1008	/// Begin line → end line, for blocks, dynamic blocks and drawers that are closed before the
	1009	/// next heading.
	1010	let blockEnds: [Int: Int]
	1011	let settings: OrgSettings
	1012	var builder = GreenBuilder()
	1013	var i = 0
	1014
	1015	init(text: String, defaults: OrgSettings) {
	1016	lines = splitRawLines(text)
	1017	info = lines.map { classifyLine($0.content) }
	1018	blockEnds = Parser.matchEnds(info)
	1019	settings = SettingsScanner.scan(lines: lines, info: info, blockEnds: blockEnds, defaults: defaults)
	1020	}
	1021
	1022	static func matchEnds(_ info: [ClassifiedLine]) -> [Int: Int] {
	1023	var ends: [Int: Int] = [:]
	1024	var k = 0
	1025	while k < info.count {
	1026	let isEnd: ((LineClass) -> Bool)?
	1027	switch info[k].cls {
	1028	case .blockBegin(let name): isEnd = { $0 == .blockEnd(name: name) }
	1029	case .dynamicBegin: isEnd = { $0 == .dynamicEnd }
	1030	case .drawerBegin: isEnd = { $0 == .drawerEnd }
	1031	default: isEnd = nil
	1032	}
	1033	if let isEnd {
	1034	var j = k + 1
	1035	while j < info.count {
	1036	if case .heading = info[j].cls { break }
	1037	if isEnd(info[j].cls) { ends[k] = j; break }
	1038	j += 1
	1039	}
	1040	// Block contents are verbatim, so nothing inside starts another element.
	1041	if let end = ends[k], !isDrawer(info[k].cls) { k = end }
	1042	}
	1043	k += 1
	1044	}
	1045	return ends
	1046	}
	1047
	1048	static func isDrawer(_ cls: LineClass) -> Bool {
	1049	if case .drawerBegin = cls { return true }
	1050	return false
	1051	}
	1052
	1053	mutating func run() -> OrgTree {
	1054	builder.start(.document)
	1055	if !lines.isEmpty, !isHeading(0) {
	1056	builder.start(.zerothSection)
	1057	parseContent(limit: lines.count)
	1058	builder.finish()
	1059	}
	1060	while i < lines.count, case .heading(let level) = info[i].cls {
	1061	parseSection(level: level)
	1062	}
	1063	builder.finish()
	1064	return OrgTree(green: builder.build(), settings: settings)
	1065	}
	1066
	1067	func isHeading(_ k: Int) -> Bool {
	1068	if case .heading = info[k].cls { return true }
	1069	return false
	1070	}
	1071
	1072	mutating func parseSection(level: Int) {
	1073	builder.start(.section)
	1074	headingLine(lines[i])
	1075	i += 1
	1076	if i < lines.count, info[i].cls == .planning {
	1077	builder.start(.planning)
	1078	line(i)
	1079	i += 1
	1080	builder.finish()
	1081	}
	1082	if i < lines.count, case .drawerBegin(let name) = info[i].cls, name.uppercased() == "PROPERTIES",
	1083	let end = blockEnds[i] {
	1084	propertyDrawer(end: end)
	1085	}
	1086	parseContent(limit: lines.count)
	1087	while i < lines.count, case .heading(let child) = info[i].cls, child > level {
	1088	parseSection(level: child)
	1089	}
	1090	builder.finish()
	1091	}
	1092
	1093	mutating func propertyDrawer(end: Int) {
	1094	builder.start(.propertyDrawer)
	1095	line(i)
	1096	i += 1
	1097	while i < end {
	1098	if info[i].cls == .blank {
	1099	line(i)
	1100	} else {
	1101	builder.start(.nodeProperty)
	1102	line(i)
	1103	builder.finish()
	1104	}
	1105	i += 1
	1106	}
	1107	line(i)
	1108	i += 1
	1109	builder.finish()
	1110	}
	1111
	1112	/// Elements until `limit` or the next heading.
	1113	mutating func parseContent(limit: Int) {
	1114	while i < limit, !isHeading(i) {
	1115	element(limit: limit, floor: nil)
	1116	}
	1117	}
	1118
	1119	/// One element starting at `i`. `floor` is the indent of the enclosing list item, if any:
	1120	/// non-blank lines at or left of it end the element.
	1121	mutating func element(limit: Int, floor: Int?) {
	1122	if info[i].cls == .blank {
	1123	line(i)
	1124	i += 1
	1125	} else {
	1126	paragraph(limit: limit, floor: floor)
	1127	}
	1128	}
	1129
	1130	mutating func paragraph(limit: Int, floor: Int?) {
	1131	builder.start(.paragraph)
	1132	line(i)
	1133	i += 1
	1134	while i < limit, within(floor, i), continuesParagraph(i) {
	1135	line(i)
	1136	i += 1
	1137	}
	1138	builder.finish()
	1139	}
	1140
	1141	func continuesParagraph(_ k: Int) -> Bool {
	1142	switch info[k].cls {
	1143	case .blank, .heading: return false
	1144	default: return true
	1145	}
	1146	}
	1147
	1148	func within(_ floor: Int?, _ k: Int) -> Bool {
	1149	guard let floor else { return true }
	1150	return info[k].indent > floor
	1151	}
	1152
	1153	// MARK: - Tokens
	1154
	1155	/// A whole line as leading whitespace, content and line ending.
	1156	mutating func line(_ k: Int) {
	1157	let content = lines[k].content
	1158	let rest = whitespace(content)
	1159	builder.token(.text, rest)
	1160	builder.token(.newline, lines[k].ending)
	1161	}
	1162
	1163	mutating func whitespace(_ s: Substring) -> Substring {
	1164	let ws = s.prefix { $0 == " " \|\| $0 == "\t" }
	1165	builder.token(.whitespace, ws)
	1166	return s.dropFirst(ws.count)
	1167	}
	1168
	1169	mutating func headingLine(_ raw: RawLine) {
	1170	builder.start(.heading)
	1171	var rest = raw.content
	1172	let stars = rest.prefix { $0 == "*" }
	1173	builder.token(.stars, stars)
	1174	rest = whitespace(rest.dropFirst(stars.count))
	1175
	1176	let word = rest.prefix { $0 != " " && $0 != "\t" }
	1177	if !word.isEmpty, settings.todoKeywordNames.contains(String(word)) {
	1178	builder.token(.todoKeyword, word)
	1179	rest = whitespace(rest.dropFirst(word.count))
	1180	}
	1181
	1182	if let cookie = priorityCookie(rest) {
	1183	builder.token(.priority, cookie)
	1184	rest = whitespace(rest.dropFirst(cookie.count))
	1185	}
	1186
	1187	let parts = splitTags(rest)
	1188	builder.token(.title, parts.title)
	1189	builder.token(.whitespace, parts.gap)
	1190	builder.token(.tags, parts.tags)
	1191	builder.token(.whitespace, parts.trailing)
	1192	builder.token(.newline, raw.ending)
	1193	builder.finish()
	1194	}
	1195
	1196	/// `[#A]` or `[#10]`, followed by whitespace or end of line.
	1197	func priorityCookie(_ s: Substring) -> Substring? {
	1198	guard s.hasPrefix("[#"), let close = s.firstIndex(of: "]") else { return nil }
	1199	let value = s[s.index(s.startIndex, offsetBy: 2)..<close]
	1200	let valid = (value.count == 1 && value.first!.isLetter && value.first!.isUppercase)
	1201	\|\| (!value.isEmpty && value.allSatisfy { $0.isASCII && $0.isNumber })
	1202	guard valid else { return nil }
	1203	let after = s[s.index(after: close)...]
	1204	guard after.isEmpty \|\| after.first == " " \|\| after.first == "\t" else { return nil }
	1205	return s[...close]
	1206	}
	1207
	1208	func splitTags(_ s: Substring) -> (title: Substring, gap: Substring, tags: Substring, trailing: Substring) {
	1209	let trimmed = s.trimmingTrailingWhitespace
	1210	let trailing = s[trimmed.endIndex...]
	1211	let none = (title: trimmed, gap: Substring(), tags: Substring(), trailing: trailing)
	1212	guard trimmed.last == ":" else { return none }
	1213	let tagStart = trimmed.lastIndex { $0 == " " \|\| $0 == "\t" }.map { trimmed.index(after: $0) } ?? trimmed.startIndex
	1214	let tags = trimmed[tagStart...]
	1215	guard tags.count >= 3, tags.first == ":", isTagString(tags) else { return none }
	1216	let before = trimmed[..<tagStart]
	1217	let title = before.trimmingTrailingWhitespace
	1218	return (title, before[title.endIndex...], tags, trailing)
	1219	}
	1220
	1221	func isTagString(_ tags: Substring) -> Bool {
	1222	tags.dropFirst().dropLast().split(separator: ":", omittingEmptySubsequences: false).allSatisfy { tag in
	1223	!tag.isEmpty && tag.allSatisfy { $0.isLetter \|\| $0.isNumber \|\| "_@#%".contains($0) }
	1224	}
	1225	}
	1226	}
	1227	```
	1228
	1229	- [ ] Step 5: Run tests to verify they pass
	1230
	1231	Run: `swift test --filter ParserSectionTests`
	1232	Expected: all pass.
	1233
	1234	- [ ] Step 6: Commit
	1235
	1236	```bash
	1237	git add Sources Tests
	1238	git commit -m "Parse document, sections and headings"
	1239	```
	1240
	1241	---
	1242
	1243	### Task 6: Parser — elements
	1244
	1245	Files:
	1246	- Modify: `Sources/OrgCore/Parser/Parser.swift` (replace `element`, `continuesParagraph`; add `list`, `item`, `table`, `consecutive`, `footnoteDefinition`)
	1247	- Test: `Tests/OrgCoreTests/ParserElementTests.swift`
	1248
	1249	Interfaces:
	1250	- Consumes: Task 5's `Parser`.
	1251	- Produces: nodes `block`, `dynamicBlock`, `drawer`, `keyword`, `affiliatedKeyword`, `comment`, `fixedWidth`, `horizontalRule`, `table`/`tableRow`/`tableFormula`, `footnoteDefinition`, `clock`, `plainList`/`item`, `planning` (after heading only), `propertyDrawer`/`nodeProperty`.
	1252
	1253	- [ ] Step 1: Write the failing tests
	1254
	1255	```swift
	1256	import Testing
	1257	@testable import OrgCore
	1258
	1259	func childKinds(_ text: String) -> [SyntaxKind] {
	1260	let root = OrgParser.parse(text).root
	1261	let container = root.children.first { $0.kind == .zerothSection \|\| $0.kind == .section }!
	1262	return container.children.map(\.kind)
	1263	}
	1264
	1265	struct ParserElementTests {
	1266	@Test func planningAndPropertiesFollowHeading() {
	1267	let text = "* a\nSCHEDULED: <2026-10-04 Sun>\n:PROPERTIES:\n:ID: x\n:END:\nbody\n"
	1268	#expect(childKinds(text) == [.heading, .planning, .propertyDrawer, .paragraph])
	1269	}
	1270
	1271	@Test func planningElsewhereIsText() {
	1272	#expect(childKinds("SCHEDULED: <2026-10-04 Sun>\n") == [.paragraph])
	1273	}
	1274
	1275	@Test func blocks() {
	1276	#expect(childKinds("#+begin_src sh\n,* escaped\n:END:\n#+end_src\nafter\n") == [.block, .paragraph])
	1277	#expect(childKinds("#+BEGIN: clocktable\n#+END:\n") == [.dynamicBlock])
	1278	}
	1279
	1280	@Test func headingsEndBlocks() {
	1281	#expect(childKinds("#+begin_src sh\n* heading\n#+end_src\n") == [.paragraph])
	1282	}
	1283
	1284	@Test func unclosedBlockIsParagraph() {
	1285	#expect(childKinds("#+begin_src sh\necho\n") == [.paragraph])
	1286	}
	1287
	1288	@Test func drawersHoldElements() {
	1289	let root = OrgParser.parse(":LOGBOOK:\nCLOCK: [2026-10-04 Sun 10:00]\n:END:\n").root
	1290	let drawer = root.children[0].children[0]
	1291	#expect(drawer.kind == .drawer)
	1292	#expect(drawer.children.map(\.kind) == [.clock])
	1293	}
	1294
	1295	@Test func keywords() {
	1296	#expect(childKinds("#+TITLE: x\n#+NAME: t\n#+ATTR_HTML: :width 10\n") == [.keyword, .affiliatedKeyword, .affiliatedKeyword])
	1297	}
	1298
	1299	@Test func commentsAndFixedWidthGroup() {
	1300	#expect(childKinds("# a\n# b\n: c\n: d\n-----\n") == [.comment, .fixedWidth, .horizontalRule])
	1301	}
	1302
	1303	@Test func tableWithFormulas() {
	1304	let root = OrgParser.parse("\| a \|\n\|---\|\n\| 1 \|\n#+TBLFM: $1=2\n#+TBLFM: $1=3\n").root
	1305	let table = root.children[0].children[0]
	1306	#expect(table.kind == .table)
	1307	#expect(table.children.map(\.kind) == [.tableRow, .tableRow, .tableRow, .tableFormula, .tableFormula])
	1308	}
	1309
	1310	@Test func footnoteDefinition() {
	1311	#expect(childKinds("[fn:1] note\ncontinued\n\nafter\n") == [.footnoteDefinition, .paragraph])
	1312	}
	1313
	1314	@Test func listsNestByIndent() {
	1315	let text = "- a\n more\n - b\n- c\n\nafter\n"
	1316	let root = OrgParser.parse(text).root
	1317	let list = root.children[0].children[0]
	1318	#expect(list.kind == .plainList)
	1319	#expect(list.children.map(\.kind) == [.item, .item])
	1320	#expect(list.children[0].children.map(\.kind) == [.paragraph, .plainList])
	1321	#expect(childKinds(text) == [.plainList, .paragraph])
	1322	}
	1323
	1324	@Test func twoBlankLinesEndAList() {
	1325	#expect(childKinds("- a\n\n\n- b\n") == [.plainList, .plainList])
	1326	#expect(childKinds("- a\n\n- b\n") == [.plainList])
	1327	}
	1328
	1329	@Test func paragraphStopsAtElementStart() {
	1330	#expect(childKinds("text\n\| a \|\n") == [.paragraph, .table])
	1331	#expect(childKinds("text\n#+begin_quote\nq\n#+end_quote\n") == [.paragraph, .block])
	1332	#expect(childKinds("text\n#+begin_quote\nq\n") == [.paragraph])
	1333	}
	1334	}
	1335	```
	1336
	1337	- [ ] Step 2: Run tests to verify they fail
	1338
	1339	Run: `swift test --filter ParserElementTests`
	1340	Expected: failures; every element parses as a paragraph.
	1341
	1342	- [ ] Step 3: Replace `element` and `continuesParagraph`, add the element parsers
	1343
	1344	```swift
	1345	static let affiliatedKeys: Set<String> = ["NAME", "CAPTION", "RESULTS", "HEADER", "PLOT"]
	1346
	1347	mutating func element(limit: Int, floor: Int?) {
	1348	switch info[i].cls {
	1349	case .blank:
	1350	line(i)
	1351	i += 1
	1352	case .blockBegin, .dynamicBegin:
	1353	if let end = blockEnds[i], end < limit {
	1354	builder.start(info[i].cls == .dynamicBegin ? .dynamicBlock : .block)
	1355	while i <= end {
	1356	line(i)
	1357	i += 1
	1358	}
	1359	builder.finish()
	1360	} else {
	1361	paragraph(limit: limit, floor: floor)
	1362	}
	1363	case .drawerBegin:
	1364	if let end = blockEnds[i], end < limit {
	1365	builder.start(.drawer)
	1366	line(i)
	1367	i += 1
	1368	parseContent(limit: end)
	1369	line(i)
	1370	i += 1
	1371	builder.finish()
	1372	} else {
	1373	paragraph(limit: limit, floor: floor)
	1374	}
	1375	case .keyword(let key):
	1376	let affiliated = Self.affiliatedKeys.contains(key) \|\| key.hasPrefix("ATTR_")
	1377	single(affiliated ? .affiliatedKeyword : .keyword)
	1378	case .comment:
	1379	consecutive(.comment, limit: limit, floor: floor) { $0 == .comment }
	1380	case .fixedWidth:
	1381	consecutive(.fixedWidth, limit: limit, floor: floor) { $0 == .fixedWidth }
	1382	case .horizontalRule:
	1383	single(.horizontalRule)
	1384	case .clock:
	1385	single(.clock)
	1386	case .tableRow:
	1387	table(limit: limit, floor: floor)
	1388	case .footnoteDefinition:
	1389	footnoteDefinition(limit: limit)
	1390	case .listItem:
	1391	list(limit: limit, floor: floor)
	1392	default:
	1393	paragraph(limit: limit, floor: floor)
	1394	}
	1395	}
	1396
	1397	/// Lines that don't start an element of their own.
	1398	func continuesParagraph(_ k: Int) -> Bool {
	1399	switch info[k].cls {
	1400	case .plain, .planning, .blockEnd, .dynamicEnd, .drawerEnd:
	1401	return true
	1402	case .blockBegin, .dynamicBegin, .drawerBegin:
	1403	return blockEnds[k] == nil
	1404	default:
	1405	return false
	1406	}
	1407	}
	1408
	1409	mutating func single(_ kind: SyntaxKind) {
	1410	builder.start(kind)
	1411	line(i)
	1412	i += 1
	1413	builder.finish()
	1414	}
	1415
	1416	mutating func consecutive(_ kind: SyntaxKind, limit: Int, floor: Int?, matching: (LineClass) -> Bool) {
	1417	builder.start(kind)
	1418	repeat {
	1419	line(i)
	1420	i += 1
	1421	} while i < limit && matching(info[i].cls) && within(floor, i)
	1422	builder.finish()
	1423	}
	1424
	1425	mutating func table(limit: Int, floor: Int?) {
	1426	builder.start(.table)
	1427	while i < limit, info[i].cls == .tableRow, within(floor, i) {
	1428	single(.tableRow)
	1429	}
	1430	while i < limit, info[i].cls == .keyword(key: "TBLFM"), within(floor, i) {
	1431	single(.tableFormula)
	1432	}
	1433	builder.finish()
	1434	}
	1435
	1436	mutating func footnoteDefinition(limit: Int) {
	1437	builder.start(.footnoteDefinition)
	1438	line(i)
	1439	i += 1
	1440	while i < limit, info[i].cls == .plain {
	1441	line(i)
	1442	i += 1
	1443	}
	1444	builder.finish()
	1445	}
	1446
	1447	mutating func list(limit: Int, floor: Int?) {
	1448	let base = info[i].indent
	1449	builder.start(.plainList)
	1450	while i < limit, info[i].cls == .listItem, info[i].indent == base, within(floor, i) {
	1451	item(base: base, limit: limit)
	1452	}
	1453	builder.finish()
	1454	}
	1455
	1456	/// An item's first line, then everything indented past its bullet. One blank line stays
	1457	/// inside the item when the item or list continues after it; two end the list.
	1458	mutating func item(base: Int, limit: Int) {
	1459	builder.start(.item)
	1460	line(i)
	1461	i += 1
	1462	while i < limit, !isHeading(i) {
	1463	if info[i].cls == .blank {
	1464	var j = i
	1465	while j < limit, info[j].cls == .blank { j += 1 }
	1466	guard j - i < 2, j < limit else { break }
	1467	let continuesItem = info[j].indent > base
	1468	let nextSibling = info[j].cls == .listItem && info[j].indent == base
	1469	guard continuesItem \|\| nextSibling else { break }
	1470	line(i)
	1471	i += 1
	1472	if nextSibling { break }
	1473	continue
	1474	}
	1475	guard info[i].indent > base else { break }
	1476	element(limit: limit, floor: base)
	1477	}
	1478	builder.finish()
	1479	}
	1480	```
	1481
	1482	- [ ] Step 4: Run all tests
	1483
	1484	Run: `swift test`
	1485	Expected: all pass, including Task 5's tests.
	1486
	1487	- [ ] Step 5: Commit
	1488
	1489	```bash
	1490	git add Sources Tests
	1491	git commit -m "Parse block-level elements"
	1492	```
	1493
	1494	---
	1495
	1496	### Task 7: Round-trip fuzz, encoding fixtures and corpus test
	1497
	1498	Files:
	1499	- Test: `Tests/OrgCoreTests/RoundTripTests.swift`
	1500
	1501	Interfaces:
	1502	- Consumes: `SourceText`, `OrgParser`, `SyntaxNode`.
	1503	- Produces: a seeded fuzz test (2,000 documents), a structural invariant check, and an opt-in corpus test driven by `ORGSTAR_CORPUS`.
	1504
	1505	- [ ] Step 1: Write the tests
	1506
	1507	```swift
	1508	import Foundation
	1509	import Testing
	1510	@testable import OrgCore
	1511
	1512	/// SplitMix64, so failures reproduce from the seed.
	1513	struct SeededGenerator: RandomNumberGenerator {
	1514	var state: UInt64
	1515	mutating func next() -> UInt64 {
	1516	state &+= 0x9E37_79B9_7F4A_7C15
	1517	var z = state
	1518	z = (z ^ (z >> 30)) &* 0xBF58_476D_1CE4_E5B9
	1519	z = (z ^ (z >> 27)) &* 0x94D0_49BB_1331_11EB
	1520	return z ^ (z >> 31)
	1521	}
	1522	}
	1523
	1524	let fragments = [
	1525	"* ", " TODO [#A] title :a:b:", "* DONE", "#+TODO: NEXT \| DONE", "#+begin_src sh", "#+end_src",
	1526	"#+BEGIN_QUOTE", "#+end_quote", "#+BEGIN: clocktable", "#+END:", ":PROPERTIES:", ":ID: x", ":END:",
	1527	":LOGBOOK:", "CLOCK: [2026-10-04 Sun 10:00]", "SCHEDULED: <2026-10-04 Sun>", "- item", " - nested",
	1528	"\t+ tab", "1. one", "\| a \| b \|", "\|---+---\|", "#+TBLFM: $2=$1", "# comment", ": fixed", "-----",
	1529	"[fn:1] note", "#+NAME: x", "plain text", "é", "😀", "e\u{301}", " ", "\t", "\n", "\n", "\r\n", "\r", "",
	1530	]
	1531
	1532	func randomDocument(_ rng: inout SeededGenerator) -> String {
	1533	(0..<Int.random(in: 0...40, using: &rng)).map { _ in
	1534	fragments.randomElement(using: &rng)! + (Bool.random(using: &rng) ? "\n" : "")
	1535	}.joined()
	1536	}
	1537
	1538	func checkLengths(_ node: SyntaxNode) -> Bool {
	1539	let sum = node.green.children.reduce(0) { $0 + $1.length }
	1540	return sum == node.green.length && node.children.allSatisfy(checkLengths)
	1541	}
	1542
	1543	struct RoundTripTests {
	1544	@Test func fuzzedDocumentsRoundTrip() {
	1545	var rng = SeededGenerator(state: 20261004)
	1546	for n in 0..<2_000 {
	1547	let text = randomDocument(&rng)
	1548	let tree = OrgParser.parse(text)
	1549	#expect(tree.text == text, "document \(n)")
	1550	#expect(checkLengths(tree.root), "document \(n)")
	1551	}
	1552	}
	1553
	1554	@Test(arguments: [
	1555	[0xEF, 0xBB, 0xBF] + Array("* a\r\n".utf8),
	1556	Array("* a\r\n- b\n\tc".utf8),
	1557	Array("* a".utf8),
	1558	Array("😀 e\u{301}\n".utf8),
	1559	[0x2A, 0x20, 0xFF, 0x0A] as [UInt8],
	1560	])
	1561	func encodingFixturesRoundTrip(bytes: [UInt8]) {
	1562	let source = SourceText(bytes: bytes)
	1563	let tree = OrgParser.parse(source.text)
	1564	#expect(source.encode(tree.text) == bytes)
	1565	}
	1566
	1567	/// Private corpus: `ORGSTAR_CORPUS=/path/to/org swift test --filter corpus`.
	1568	@Test(.enabled(if: ProcessInfo.processInfo.environment["ORGSTAR_CORPUS"] != nil))
	1569	func corpusRoundTrips() throws {
	1570	let root = URL(fileURLWithPath: ProcessInfo.processInfo.environment["ORGSTAR_CORPUS"]!)
	1571	let files = FileManager.default.enumerator(at: root, includingPropertiesForKeys: nil)!
	1572	.compactMap { $0 as? URL }
	1573	.filter { ["org", "org_archive"].contains($0.pathExtension) }
	1574	#expect(!files.isEmpty)
	1575	for file in files {
	1576	let bytes = try [UInt8](Data(contentsOf: file))
	1577	let source = SourceText(bytes: bytes)
	1578	let tree = OrgParser.parse(source.text)
	1579	#expect(source.encode(tree.text) == bytes, "\(file.path)")
	1580	}
	1581	}
	1582	}
	1583	```
	1584
	1585	- [ ] Step 2: Run the tests
	1586
	1587	Run: `swift test`
	1588	Expected: all pass. The corpus test is skipped.
	1589
	1590	- [ ] Step 3: Run the private corpus locally
	1591
	1592	Run: `ORGSTAR_CORPUS=~/org swift test --filter corpusRoundTrips`
	1593	Expected: pass. Any failure names the file; reduce it to a fuzz fragment or a unit test before fixing.
	1594
	1595	- [ ] Step 4: Commit
	1596
	1597	```bash
	1598	git add Tests
	1599	git commit -m "Add round-trip fuzz, encoding and corpus tests"
	1600	```

Commit f4591a4d3b

Commit `f4591a4d3b`