Commit ce36aceaee
Verified · cmc
Layout: unified · split
docs/plans/2026-10-04-semantic-layer.md added +844
| @@ -0,0 +1,844 @@ | |||
| 1 | # Semantic Layer 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:** Typed facts over the parsed tree, so the index and commands never read raw text: headings with their planning, properties, clocks, timestamps and links; tag and property inheritance with provenance; table models; source blocks with resolved header arguments. | ||
| 6 | |||
| 7 | **Architecture:** `DocumentModel(tree:settings:)` walks the tree once. Each heading's own facts go into `HeadingInfo`; content of child sections stays with the child. Inheritance is computed on request from local values, so `HeadingInfo` stores only what is written under the heading. Every resolved value is a `Resolved<Value>` carrying a `ValueSource`. Source blocks resolve after the walk because their header arguments depend on file and heading properties. | ||
| 8 | |||
| 9 | **Tech Stack:** Swift 6.2 tools, Swift Testing, Foundation only. | ||
| 10 | |||
| 11 | **Spec:** `docs/design.md`, "Semantic layer" and the inheritance rules under it. | ||
| 12 | |||
| 13 | ## Global Constraints | ||
| 14 | |||
| 15 | - Local values and resolved values stay separate; inheritance never writes into `HeadingInfo`. | ||
| 16 | - Tags inherit by default minus an exclusion list; `#+FILETAGS` count as inherited. | ||
| 17 | - Properties don't inherit by default; `CATEGORY`, `ARCHIVE`, `COLUMNS`, `LOGGING` and `header-args*` always do; `ID` and `CUSTOM_ID` never do. `KEY+` appends with a space. | ||
| 18 | - Header arguments: org-babel defaults, then `header-args`, then `header-args:LANG` (each from `#+PROPERTY` down to the heading, a plain property replacing what is above it), then `#+HEADER` lines, then the begin line. | ||
| 19 | |||
| 20 | ## Out of scope | ||
| 21 | |||
| 22 | Merging `:results` values by group (collection, type, format, handling), language default header arguments, `CATEGORY` defaulting to the file name (needs the workspace), and source body unescaping (`,*`). Each lands with the phase that needs it. | ||
| 23 | |||
| 24 | ## File structure | ||
| 25 | |||
| 26 | | File | Responsibility | | ||
| 27 | | --- | --- | | ||
| 28 | | `Sources/OrgCore/Semantic/SemanticSettings.swift` | Inheritance settings, `ValueSource`, `Resolved` | | ||
| 29 | | `Sources/OrgCore/Semantic/HeadingInfo.swift` | `HeadingInfo`, `Property`, `Clock` | | ||
| 30 | | `Sources/OrgCore/Semantic/DisplayWidth.swift` | Monospace column width of text | | ||
| 31 | | `Sources/OrgCore/Semantic/TableModel.swift` | Rows, rules, formulas, column widths | | ||
| 32 | | `Sources/OrgCore/Semantic/SrcBlockInfo.swift` | `SrcBlockInfo`, header argument parsing and defaults | | ||
| 33 | | `Sources/OrgCore/Semantic/DocumentModel.swift` | The walk, inheritance queries, header argument resolution | | ||
| 34 | |||
| 35 | --- | ||
| 36 | |||
| 37 | ### Task 1: Value types | ||
| 38 | |||
| 39 | **Files:** | ||
| 40 | - Create: `Sources/OrgCore/Semantic/SemanticSettings.swift`, `HeadingInfo.swift`, `DisplayWidth.swift`, `TableModel.swift`, `SrcBlockInfo.swift` | ||
| 41 | - Test: `Tests/OrgCoreTests/SemanticValueTests.swift` | ||
| 42 | |||
| 43 | **Interfaces:** | ||
| 44 | - Produces: `SemanticSettings` (`.default`, `PropertyInheritance` `.none`/`.all`/`.only(Set<String>)` of upper-cased keys), `ValueSource` (`.defaults`, `.file`, `.heading(Int)`, `.element`), `Resolved<Value>(_:_:)`, `Property(key:value:additive:)` and internal `Property(line:)`, `Clock(start:end:minutes:)`, `HeadingInfo`, `displayWidth(_:)`, `TableModel` (internal `init(node:heading:)`), `SrcBlockInfo`, `HeaderArguments.defaults`, `HeaderArguments.parse(_:) -> [(key: String, value: String)]`. | ||
| 45 | |||
| 46 | - [ ] **Step 1: Write the failing tests** | ||
| 47 | |||
| 48 | ```swift | ||
| 49 | import Testing | ||
| 50 | @testable import OrgCore | ||
| 51 | |||
| 52 | struct SemanticValueTests { | ||
| 53 | @Test func parseHeaderArguments() { | ||
| 54 | let pairs = HeaderArguments.parse("ignored :results output silent :dir /tmp :VAR x=1") | ||
| 55 | #expect(pairs.map { "\($0.key)=\($0.value)" } == [":results=output silent", ":dir=/tmp", ":var=x=1"]) | ||
| 56 | } | ||
| 57 | |||
| 58 | @Test func widths() { | ||
| 59 | #expect(displayWidth("abc") == 3) | ||
| 60 | #expect(displayWidth("日本") == 4) | ||
| 61 | #expect(displayWidth("e\u{301}") == 1) | ||
| 62 | #expect(displayWidth("😀") == 2) | ||
| 63 | } | ||
| 64 | } | ||
| 65 | ``` | ||
| 66 | |||
| 67 | - [ ] **Step 2: Run to verify failure** | ||
| 68 | |||
| 69 | Run: `swift test --filter SemanticValueTests` | ||
| 70 | Expected: build failure, `cannot find 'HeaderArguments' in scope`. | ||
| 71 | |||
| 72 | - [ ] **Step 3: Implement `SemanticSettings.swift`** | ||
| 73 | |||
| 74 | ```swift | ||
| 75 | /// App-wide settings that change what a document means without changing how it parses. | ||
| 76 | public struct SemanticSettings: Sendable, Equatable { | ||
| 77 | public enum PropertyInheritance: Sendable, Equatable { | ||
| 78 | /// org's default: properties apply only to their own heading. | ||
| 79 | case none | ||
| 80 | case all | ||
| 81 | /// Upper-cased property keys that inherit. | ||
| 82 | case only(Set<String>) | ||
| 83 | } | ||
| 84 | |||
| 85 | public var tagInheritance: Bool | ||
| 86 | public var tagsExcludedFromInheritance: Set<String> | ||
| 87 | public var propertyInheritance: PropertyInheritance | ||
| 88 | |||
| 89 | public init( | ||
| 90 | tagInheritance: Bool = true, | ||
| 91 | tagsExcludedFromInheritance: Set<String> = [], | ||
| 92 | propertyInheritance: PropertyInheritance = .none | ||
| 93 | ) { | ||
| 94 | self.tagInheritance = tagInheritance | ||
| 95 | self.tagsExcludedFromInheritance = tagsExcludedFromInheritance | ||
| 96 | self.propertyInheritance = propertyInheritance | ||
| 97 | } | ||
| 98 | |||
| 99 | public static let `default` = SemanticSettings() | ||
| 100 | } | ||
| 101 | |||
| 102 | /// Where a resolved value came from. | ||
| 103 | public enum ValueSource: Sendable, Equatable { | ||
| 104 | case defaults | ||
| 105 | /// A file-level keyword such as `#+FILETAGS` or `#+PROPERTY`. | ||
| 106 | case file | ||
| 107 | /// A heading, by index into `DocumentModel.headings`. | ||
| 108 | case heading(Int) | ||
| 109 | /// The element itself: a `#+HEADER` line or a block's begin line. | ||
| 110 | case element | ||
| 111 | } | ||
| 112 | |||
| 113 | public struct Resolved<Value: Sendable & Equatable>: Sendable, Equatable { | ||
| 114 | public let value: Value | ||
| 115 | public let source: ValueSource | ||
| 116 | |||
| 117 | public init(_ value: Value, _ source: ValueSource) { | ||
| 118 | self.value = value | ||
| 119 | self.source = source | ||
| 120 | } | ||
| 121 | } | ||
| 122 | ``` | ||
| 123 | |||
| 124 | - [ ] **Step 4: Implement `HeadingInfo.swift`** | ||
| 125 | |||
| 126 | ```swift | ||
| 127 | import Foundation | ||
| 128 | |||
| 129 | public struct Property: Sendable, Equatable { | ||
| 130 | public let key: String | ||
| 131 | public let value: String | ||
| 132 | /// `:KEY+:` appends to the value instead of replacing it. | ||
| 133 | public let additive: Bool | ||
| 134 | |||
| 135 | public init(key: String, value: String, additive: Bool = false) { | ||
| 136 | self.key = key | ||
| 137 | self.value = value | ||
| 138 | self.additive = additive | ||
| 139 | } | ||
| 140 | |||
| 141 | /// A `:KEY: value` or `:KEY+: value` drawer line. | ||
| 142 | init?(line: String) { | ||
| 143 | let trimmed = line.trimmingCharacters(in: .whitespacesAndNewlines) | ||
| 144 | guard trimmed.hasPrefix(":") else { return nil } | ||
| 145 | let rest = trimmed.dropFirst() | ||
| 146 | guard let colon = rest.firstIndex(of: ":") else { return nil } | ||
| 147 | var key = rest[..<colon] | ||
| 148 | guard !key.isEmpty else { return nil } | ||
| 149 | let additive = key.hasSuffix("+") | ||
| 150 | if additive { key = key.dropLast() } | ||
| 151 | let value = rest[rest.index(after: colon)...].trimmingCharacters(in: .whitespaces) | ||
| 152 | self.init(key: String(key), value: value, additive: additive) | ||
| 153 | } | ||
| 154 | } | ||
| 155 | |||
| 156 | public struct Clock: Sendable, Equatable { | ||
| 157 | public let start: Timestamp.Point | ||
| 158 | /// Nil while the clock is running. | ||
| 159 | public let end: Timestamp.Point? | ||
| 160 | /// From `=> H:MM`, when present. | ||
| 161 | public let minutes: Int? | ||
| 162 | } | ||
| 163 | |||
| 164 | /// One heading and the facts that belong to it directly, not to its children. | ||
| 165 | public struct HeadingInfo: Sendable, Equatable { | ||
| 166 | public internal(set) var sectionRange: Range<Int> | ||
| 167 | public internal(set) var headingRange: Range<Int> | ||
| 168 | public internal(set) var level: Int | ||
| 169 | /// Index of the parent heading in `DocumentModel.headings`. | ||
| 170 | public internal(set) var parent: Int? | ||
| 171 | public internal(set) var todo: String? | ||
| 172 | public internal(set) var isDone: Bool | ||
| 173 | /// The cookie's value (`A`, `10`), when the heading has one. | ||
| 174 | public internal(set) var priority: String? | ||
| 175 | /// The title as written, markup included. | ||
| 176 | public internal(set) var title: String | ||
| 177 | public internal(set) var localTags: [String] | ||
| 178 | /// Property drawer lines in order. | ||
| 179 | public internal(set) var properties: [Property] | ||
| 180 | public internal(set) var scheduled: Timestamp? | ||
| 181 | public internal(set) var deadline: Timestamp? | ||
| 182 | public internal(set) var closed: Timestamp? | ||
| 183 | public internal(set) var clocks: [Clock] | ||
| 184 | /// Timestamps in the title and body, excluding planning and clock lines. | ||
| 185 | public internal(set) var timestamps: [Timestamp] | ||
| 186 | /// Link targets in the title and body. | ||
| 187 | public internal(set) var links: [String] | ||
| 188 | |||
| 189 | /// The heading's own `ID`. Never inherited. | ||
| 190 | public var id: String? { | ||
| 191 | properties.last { $0.key.uppercased() == "ID" && !$0.additive }?.value | ||
| 192 | } | ||
| 193 | } | ||
| 194 | ``` | ||
| 195 | |||
| 196 | - [ ] **Step 5: Implement `DisplayWidth.swift`** | ||
| 197 | |||
| 198 | ```swift | ||
| 199 | /// Columns a string occupies in a monospaced table: East Asian wide and fullwidth characters | ||
| 200 | /// and emoji take two, everything else one. Combining marks belong to their Character. | ||
| 201 | public func displayWidth(_ text: some StringProtocol) -> Int { | ||
| 202 | text.reduce(0) { $0 + displayWidth(of: $1) } | ||
| 203 | } | ||
| 204 | |||
| 205 | func displayWidth(of character: Character) -> Int { | ||
| 206 | guard let scalar = character.unicodeScalars.first else { return 0 } | ||
| 207 | if scalar.properties.isEmojiPresentation { return 2 } | ||
| 208 | switch scalar.value { | ||
| 209 | case 0x1100...0x115F, 0x2E80...0x303E, 0x3041...0xA4CF, 0xAC00...0xD7A3, 0xF900...0xFAFF, | ||
| 210 | 0xFE30...0xFE4F, 0xFF00...0xFF60, 0xFFE0...0xFFE6, 0x20000...0x3FFFD: | ||
| 211 | return 2 | ||
| 212 | default: | ||
| 213 | return 1 | ||
| 214 | } | ||
| 215 | } | ||
| 216 | ``` | ||
| 217 | |||
| 218 | - [ ] **Step 6: Implement `TableModel.swift`** | ||
| 219 | |||
| 220 | ```swift | ||
| 221 | import Foundation | ||
| 222 | |||
| 223 | public struct TableModel: Sendable, Equatable { | ||
| 224 | public enum Row: Sendable, Equatable { | ||
| 225 | case rule | ||
| 226 | /// Cell text with surrounding blanks trimmed. | ||
| 227 | case cells([String]) | ||
| 228 | } | ||
| 229 | |||
| 230 | public let range: Range<Int> | ||
| 231 | public let heading: Int? | ||
| 232 | public let rows: [Row] | ||
| 233 | /// `#+TBLFM:` values in order. | ||
| 234 | public let formulas: [String] | ||
| 235 | /// Widest cell per column, in display columns. | ||
| 236 | public let columnWidths: [Int] | ||
| 237 | |||
| 238 | public var columnCount: Int { columnWidths.count } | ||
| 239 | |||
| 240 | init(node: SyntaxNode, heading: Int?) { | ||
| 241 | range = node.range | ||
| 242 | self.heading = heading | ||
| 243 | rows = node.children.filter { $0.kind == .tableRow }.map { row in | ||
| 244 | let cells = row.children.filter { $0.kind == .tableCell } | ||
| 245 | if cells.isEmpty, row.text.trimmingCharacters(in: .whitespaces).hasPrefix("|-") { return .rule } | ||
| 246 | return .cells(cells.map { $0.text.trimmingCharacters(in: .whitespaces) }) | ||
| 247 | } | ||
| 248 | formulas = node.children.filter { $0.kind == .tableFormula }.map { | ||
| 249 | SettingsScanner.keywordValue($0.text.trimmingCharacters(in: .whitespacesAndNewlines)[...]) | ||
| 250 | .trimmingCharacters(in: .whitespaces) | ||
| 251 | } | ||
| 252 | var widths: [Int] = [] | ||
| 253 | for case .cells(let cells) in rows { | ||
| 254 | for (column, cell) in cells.enumerated() { | ||
| 255 | if column == widths.count { widths.append(0) } | ||
| 256 | widths[column] = max(widths[column], displayWidth(cell)) | ||
| 257 | } | ||
| 258 | } | ||
| 259 | columnWidths = widths | ||
| 260 | } | ||
| 261 | } | ||
| 262 | ``` | ||
| 263 | |||
| 264 | - [ ] **Step 7: Implement `SrcBlockInfo.swift`** | ||
| 265 | |||
| 266 | ```swift | ||
| 267 | public struct SrcBlockInfo: Sendable, Equatable { | ||
| 268 | public let range: Range<Int> | ||
| 269 | public let heading: Int? | ||
| 270 | /// From an attached `#+NAME:` line. | ||
| 271 | public let name: String? | ||
| 272 | public let language: String? | ||
| 273 | /// Words like `-n` or `-r` between the language and the first header argument. | ||
| 274 | public let switches: [String] | ||
| 275 | /// The lines between the begin and end lines, unchanged. | ||
| 276 | public let body: String | ||
| 277 | /// Keys keep their leading colon (`:results`). | ||
| 278 | public let headerArguments: [String: Resolved<String>] | ||
| 279 | } | ||
| 280 | |||
| 281 | public enum HeaderArguments { | ||
| 282 | /// org-babel's global defaults. | ||
| 283 | public static let defaults: [String: String] = [ | ||
| 284 | ":session": "none", ":results": "replace", ":exports": "code", ":cache": "no", | ||
| 285 | ":noweb": "no", ":hlines": "no", ":tangle": "no", | ||
| 286 | ] | ||
| 287 | |||
| 288 | /// `:key value :other value two` as ordered pairs. Keys are lower-cased and keep their colon; | ||
| 289 | /// words before the first key are dropped. | ||
| 290 | public static func parse(_ text: some StringProtocol) -> [(key: String, value: String)] { | ||
| 291 | var pairs: [(key: String, value: String)] = [] | ||
| 292 | for word in text.split(whereSeparator: \.isWhitespace) { | ||
| 293 | if word.hasPrefix(":"), word.count > 1 { | ||
| 294 | pairs.append((word.lowercased(), "")) | ||
| 295 | } else if !pairs.isEmpty { | ||
| 296 | let current = pairs[pairs.count - 1].value | ||
| 297 | pairs[pairs.count - 1].value = current.isEmpty ? String(word) : current + " " + word | ||
| 298 | } | ||
| 299 | } | ||
| 300 | return pairs | ||
| 301 | } | ||
| 302 | } | ||
| 303 | ``` | ||
| 304 | |||
| 305 | - [ ] **Step 8: Run to verify pass** | ||
| 306 | |||
| 307 | Run: `swift test --filter SemanticValueTests` | ||
| 308 | Expected: all pass. | ||
| 309 | |||
| 310 | - [ ] **Step 9: Commit** | ||
| 311 | |||
| 312 | ```bash | ||
| 313 | git add Sources/OrgCore/Semantic Tests/OrgCoreTests/SemanticValueTests.swift | ||
| 314 | git commit -m "Add semantic value types" | ||
| 315 | ``` | ||
| 316 | |||
| 317 | --- | ||
| 318 | |||
| 319 | ### Task 2: Document model | ||
| 320 | |||
| 321 | **Files:** | ||
| 322 | - Create: `Sources/OrgCore/Semantic/DocumentModel.swift` | ||
| 323 | - Test: `Tests/OrgCoreTests/SemanticTests.swift` | ||
| 324 | |||
| 325 | **Interfaces:** | ||
| 326 | - Consumes: Task 1 types, `OrgTree`, `SyntaxNode`, `Timestamp.parse`, `SettingsScanner.keywordValue`, `splitRawLines`. | ||
| 327 | - Produces: `DocumentModel(tree:settings:)` with `headings`, `fileTags`, `fileProperties`, `tables`, `srcBlocks`, `ancestors(of:)`, `outlinePath(of:)`, `inherits(_:)`, `property(_:of:) -> Resolved<String>?`, `tags(of:) -> [Resolved<String>]`. | ||
| 328 | |||
| 329 | - [ ] **Step 1: Write the failing tests** | ||
| 330 | |||
| 331 | ```swift | ||
| 332 | import Testing | ||
| 333 | @testable import OrgCore | ||
| 334 | |||
| 335 | func model(_ text: String, _ settings: SemanticSettings = .default) -> DocumentModel { | ||
| 336 | DocumentModel(tree: OrgParser.parse(text), settings: settings) | ||
| 337 | } | ||
| 338 | |||
| 339 | struct HeadingInfoTests { | ||
| 340 | @Test func basics() { | ||
| 341 | let m = model("* TODO [#A] Write *plan* :work:\n** DONE Sub\n") | ||
| 342 | #expect(m.headings.count == 2) | ||
| 343 | let top = m.headings[0] | ||
| 344 | #expect(top.level == 1) | ||
| 345 | #expect(top.todo == "TODO") | ||
| 346 | #expect(!top.isDone) | ||
| 347 | #expect(top.priority == "A") | ||
| 348 | #expect(top.title == "Write *plan*") | ||
| 349 | #expect(top.localTags == ["work"]) | ||
| 350 | #expect(top.parent == nil) | ||
| 351 | #expect(m.headings[1].parent == 0) | ||
| 352 | #expect(m.headings[1].isDone) | ||
| 353 | #expect(m.outlinePath(of: 1) == ["Write *plan*", "Sub"]) | ||
| 354 | } | ||
| 355 | |||
| 356 | @Test func planningPropertiesClocksAndBody() throws { | ||
| 357 | let text = """ | ||
| 358 | * a | ||
| 359 | SCHEDULED: <2026-10-04 Sun> DEADLINE: <2026-10-10 Sat -2d> | ||
| 360 | :PROPERTIES: | ||
| 361 | :ID: abc | ||
| 362 | :Effort: 1:00 | ||
| 363 | :END: | ||
| 364 | :LOGBOOK: | ||
| 365 | CLOCK: [2026-10-04 Sun 10:00]--[2026-10-04 Sun 11:30] => 1:30 | ||
| 366 | CLOCK: [2026-10-05 Mon 09:00] | ||
| 367 | :END: | ||
| 368 | See <2026-10-07 Wed> and [[https://a.b][x]]. | ||
| 369 | |||
| 370 | """ | ||
| 371 | let h = model(text).headings[0] | ||
| 372 | #expect(h.scheduled?.start.day == 4) | ||
| 373 | #expect(h.deadline?.warning?.interval.value == 2) | ||
| 374 | #expect(h.closed == nil) | ||
| 375 | #expect(h.id == "abc") | ||
| 376 | #expect(h.properties == [Property(key: "ID", value: "abc"), Property(key: "Effort", value: "1:00")]) | ||
| 377 | #expect(h.clocks.count == 2) | ||
| 378 | #expect(h.clocks[0].minutes == 90) | ||
| 379 | #expect(h.clocks[0].end?.hour == 11) | ||
| 380 | #expect(h.clocks[1].end == nil) | ||
| 381 | #expect(h.timestamps == [try #require(Timestamp.parse("<2026-10-07 Wed>"))]) | ||
| 382 | #expect(h.links == ["https://a.b"]) | ||
| 383 | } | ||
| 384 | |||
| 385 | @Test func childContentStaysWithChild() { | ||
| 386 | let m = model("* a\n** b\n<2026-10-04 Sun>\n") | ||
| 387 | #expect(m.headings[0].timestamps.isEmpty) | ||
| 388 | #expect(m.headings[1].timestamps.count == 1) | ||
| 389 | } | ||
| 390 | |||
| 391 | @Test func titleTimestampsAndLinks() { | ||
| 392 | let h = model("* Call <2026-10-04 Sun> [[id:x][y]]\n").headings[0] | ||
| 393 | #expect(h.timestamps.count == 1) | ||
| 394 | #expect(h.links == ["id:x"]) | ||
| 395 | } | ||
| 396 | |||
| 397 | @Test func doneStatesComeFromFileSettings() { | ||
| 398 | let m = model("#+TODO: NEXT | SHIPPED\n* SHIPPED x\n") | ||
| 399 | #expect(m.headings[0].isDone) | ||
| 400 | } | ||
| 401 | } | ||
| 402 | |||
| 403 | struct InheritanceTests { | ||
| 404 | @Test func tagInheritance() { | ||
| 405 | let text = "#+FILETAGS: :f:\n* a :x:noinherit:\n** b :y:\n" | ||
| 406 | let m = model(text, SemanticSettings(tagsExcludedFromInheritance: ["noinherit"])) | ||
| 407 | #expect(m.tags(of: 1).map(\.value) == ["f", "x", "y"]) | ||
| 408 | #expect(m.tags(of: 1).map(\.source) == [.file, .heading(0), .heading(1)]) | ||
| 409 | #expect(model(text, SemanticSettings(tagInheritance: false)).tags(of: 1).map(\.value) == ["y"]) | ||
| 410 | } | ||
| 411 | |||
| 412 | @Test func ownTagWinsOverInherited() { | ||
| 413 | #expect(model("* a :x:\n** b :x:\n").tags(of: 1) == [Resolved("x", .heading(1))]) | ||
| 414 | } | ||
| 415 | |||
| 416 | @Test func propertyPolicies() { | ||
| 417 | let text = "#+PROPERTY: owner team\n* a\n:PROPERTIES:\n:OWNER: alice\n:CATEGORY: work\n:ID: p\n:END:\n** b\n" | ||
| 418 | let plain = model(text) | ||
| 419 | #expect(plain.property("OWNER", of: 1) == nil) | ||
| 420 | #expect(plain.property("CATEGORY", of: 1) == Resolved("work", .heading(0))) | ||
| 421 | #expect(plain.property("OWNER", of: nil) == Resolved("team", .file)) | ||
| 422 | |||
| 423 | let all = model(text, SemanticSettings(propertyInheritance: .all)) | ||
| 424 | #expect(all.property("owner", of: 1) == Resolved("alice", .heading(0))) | ||
| 425 | #expect(all.property("ID", of: 1) == nil) | ||
| 426 | |||
| 427 | let some = model(text, SemanticSettings(propertyInheritance: .only(["OWNER"]))) | ||
| 428 | #expect(some.property("OWNER", of: 1) == Resolved("alice", .heading(0))) | ||
| 429 | } | ||
| 430 | |||
| 431 | @Test func additiveProperties() { | ||
| 432 | let text = "* a\n:PROPERTIES:\n:VAR: x=1\n:END:\n** b\n:PROPERTIES:\n:VAR+: y=2\n:END:\n" | ||
| 433 | #expect(model(text, SemanticSettings(propertyInheritance: .all)).property("VAR", of: 1) == Resolved("x=1 y=2", .heading(1))) | ||
| 434 | #expect(model(text).property("VAR", of: 1) == Resolved("y=2", .heading(1))) | ||
| 435 | } | ||
| 436 | } | ||
| 437 | |||
| 438 | struct SrcBlockTests { | ||
| 439 | @Test func resolution() throws { | ||
| 440 | let text = """ | ||
| 441 | #+PROPERTY: header-args :dir /a :cache yes | ||
| 442 | #+PROPERTY: header-args:sh :results output | ||
| 443 | * a | ||
| 444 | :PROPERTIES: | ||
| 445 | :header-args+: :dir /b | ||
| 446 | :END: | ||
| 447 | #+NAME: hello | ||
| 448 | #+HEADER: :var x=1 | ||
| 449 | #+begin_src sh -n :results silent | ||
| 450 | echo $x | ||
| 451 | #+end_src | ||
| 452 | |||
| 453 | """ | ||
| 454 | let block = try #require(model(text).srcBlocks.first) | ||
| 455 | #expect(block.name == "hello") | ||
| 456 | #expect(block.language == "sh") | ||
| 457 | #expect(block.switches == ["-n"]) | ||
| 458 | #expect(block.body == "echo $x\n") | ||
| 459 | #expect(block.heading == 0) | ||
| 460 | #expect(block.headerArguments[":dir"] == Resolved("/b", .heading(0))) | ||
| 461 | #expect(block.headerArguments[":cache"] == Resolved("yes", .file)) | ||
| 462 | #expect(block.headerArguments[":results"] == Resolved("silent", .element)) | ||
| 463 | #expect(block.headerArguments[":var"] == Resolved("x=1", .element)) | ||
| 464 | #expect(block.headerArguments[":exports"] == Resolved("code", .defaults)) | ||
| 465 | } | ||
| 466 | |||
| 467 | @Test func plainPropertyReplacesInheritedHeaderArguments() { | ||
| 468 | let text = "#+PROPERTY: header-args :dir /a\n* a\n:PROPERTIES:\n:header-args: :cache yes\n:END:\n#+begin_src sh\n#+end_src\n" | ||
| 469 | let block = model(text).srcBlocks[0] | ||
| 470 | #expect(block.headerArguments[":dir"] == nil) | ||
| 471 | #expect(block.headerArguments[":cache"] == Resolved("yes", .heading(0))) | ||
| 472 | } | ||
| 473 | |||
| 474 | @Test func blankLineDetachesAffiliatedKeywords() { | ||
| 475 | #expect(model("#+NAME: x\n\n#+begin_src sh\n#+end_src\n").srcBlocks[0].name == nil) | ||
| 476 | } | ||
| 477 | |||
| 478 | @Test func otherBlocksAreNotSourceBlocks() { | ||
| 479 | #expect(model("#+begin_example\nx\n#+end_example\n").srcBlocks.isEmpty) | ||
| 480 | } | ||
| 481 | } | ||
| 482 | |||
| 483 | struct TableModelTests { | ||
| 484 | @Test func rowsFormulasAndWidths() { | ||
| 485 | let table = model("| a | 日本 |\n|---+---|\n| 😀 | b |\n#+TBLFM: $2=$1\n").tables[0] | ||
| 486 | #expect(table.rows == [.cells(["a", "日本"]), .rule, .cells(["😀", "b"])]) | ||
| 487 | #expect(table.formulas == ["$2=$1"]) | ||
| 488 | #expect(table.columnWidths == [2, 4]) | ||
| 489 | } | ||
| 490 | } | ||
| 491 | ``` | ||
| 492 | |||
| 493 | - [ ] **Step 2: Run to verify failure** | ||
| 494 | |||
| 495 | Run: `swift test --filter HeadingInfoTests` | ||
| 496 | Expected: build failure, `cannot find 'DocumentModel' in scope`. | ||
| 497 | |||
| 498 | - [ ] **Step 3: Implement** | ||
| 499 | |||
| 500 | ```swift | ||
| 501 | import Foundation | ||
| 502 | |||
| 503 | /// Typed facts read from one parsed document. Built once per tree; commands and the index read | ||
| 504 | /// this rather than raw text. | ||
| 505 | public struct DocumentModel: Sendable { | ||
| 506 | public let settings: SemanticSettings | ||
| 507 | public let orgSettings: OrgSettings | ||
| 508 | public private(set) var headings: [HeadingInfo] = [] | ||
| 509 | public private(set) var fileTags: [String] = [] | ||
| 510 | /// `#+PROPERTY:` lines in order. | ||
| 511 | public private(set) var fileProperties: [Property] = [] | ||
| 512 | public private(set) var tables: [TableModel] = [] | ||
| 513 | public private(set) var srcBlocks: [SrcBlockInfo] = [] | ||
| 514 | |||
| 515 | /// Always inherited, whatever the inheritance setting. | ||
| 516 | static let alwaysInherited: Set<String> = ["CATEGORY", "ARCHIVE", "COLUMNS", "LOGGING"] | ||
| 517 | /// Identify one heading, so never inherited. | ||
| 518 | static let neverInherited: Set<String> = ["ID", "CUSTOM_ID"] | ||
| 519 | |||
| 520 | private typealias PendingBlock = (node: SyntaxNode, heading: Int?, affiliated: [SyntaxNode]) | ||
| 521 | |||
| 522 | public init(tree: OrgTree, settings: SemanticSettings = .default) { | ||
| 523 | self.settings = settings | ||
| 524 | self.orgSettings = tree.settings | ||
| 525 | var pending: [PendingBlock] = [] | ||
| 526 | walk(tree.root, heading: nil, pending: &pending) | ||
| 527 | // Header arguments depend on file and heading properties, so blocks resolve last. | ||
| 528 | srcBlocks = pending.map { srcBlock($0.node, heading: $0.heading, affiliated: $0.affiliated) } | ||
| 529 | } | ||
| 530 | |||
| 531 | // MARK: - Queries | ||
| 532 | |||
| 533 | /// Ancestor indexes of a heading, outermost first. | ||
| 534 | public func ancestors(of index: Int) -> [Int] { | ||
| 535 | var result: [Int] = [] | ||
| 536 | var current = headings[index].parent | ||
| 537 | while let parent = current { | ||
| 538 | result.insert(parent, at: 0) | ||
| 539 | current = headings[parent].parent | ||
| 540 | } | ||
| 541 | return result | ||
| 542 | } | ||
| 543 | |||
| 544 | /// Titles from the outermost ancestor down to the heading itself. | ||
| 545 | public func outlinePath(of index: Int) -> [String] { | ||
| 546 | (ancestors(of: index) + [index]).map { headings[$0].title } | ||
| 547 | } | ||
| 548 | |||
| 549 | public func inherits(_ key: String) -> Bool { | ||
| 550 | let key = key.uppercased() | ||
| 551 | if Self.neverInherited.contains(key) { return false } | ||
| 552 | if Self.alwaysInherited.contains(key) || key.hasPrefix("HEADER-ARGS") { return true } | ||
| 553 | switch settings.propertyInheritance { | ||
| 554 | case .none: return false | ||
| 555 | case .all: return true | ||
| 556 | case .only(let keys): return keys.contains(key) | ||
| 557 | } | ||
| 558 | } | ||
| 559 | |||
| 560 | /// The value of `key` for a heading, or for the file when `index` is nil. Inherited keys | ||
| 561 | /// start from `#+PROPERTY` and walk down the ancestors; `KEY+` lines append with a space. | ||
| 562 | public func property(_ key: String, of index: Int?) -> Resolved<String>? { | ||
| 563 | let key = key.uppercased() | ||
| 564 | var layers: [(entries: [Property], source: ValueSource)] = [] | ||
| 565 | if let index { | ||
| 566 | if inherits(key) { | ||
| 567 | layers.append((fileProperties, .file)) | ||
| 568 | layers += ancestors(of: index).map { (headings[$0].properties, .heading($0)) } | ||
| 569 | } | ||
| 570 | layers.append((headings[index].properties, .heading(index))) | ||
| 571 | } else { | ||
| 572 | layers.append((fileProperties, .file)) | ||
| 573 | } | ||
| 574 | var result: Resolved<String>? | ||
| 575 | for layer in layers { | ||
| 576 | for entry in layer.entries where entry.key.uppercased() == key { | ||
| 577 | if entry.additive, let current = result { | ||
| 578 | result = Resolved(current.value + " " + entry.value, layer.source) | ||
| 579 | } else { | ||
| 580 | result = Resolved(entry.value, layer.source) | ||
| 581 | } | ||
| 582 | } | ||
| 583 | } | ||
| 584 | return result | ||
| 585 | } | ||
| 586 | |||
| 587 | /// Inherited tags first (file tags, then ancestors outermost first), then the heading's own. | ||
| 588 | /// A tag the heading has itself counts as its own. | ||
| 589 | public func tags(of index: Int) -> [Resolved<String>] { | ||
| 590 | let local = headings[index].localTags | ||
| 591 | var inherited: [Resolved<String>] = [] | ||
| 592 | if settings.tagInheritance { | ||
| 593 | let layers = [(fileTags, ValueSource.file)] + ancestors(of: index).map { (headings[$0].localTags, ValueSource.heading($0)) } | ||
| 594 | for (tags, source) in layers { | ||
| 595 | for tag in tags where !settings.tagsExcludedFromInheritance.contains(tag) | ||
| 596 | && !local.contains(tag) && !inherited.contains(where: { $0.value == tag }) { | ||
| 597 | inherited.append(Resolved(tag, source)) | ||
| 598 | } | ||
| 599 | } | ||
| 600 | } | ||
| 601 | var own: [Resolved<String>] = [] | ||
| 602 | for tag in local where !own.contains(where: { $0.value == tag }) { | ||
| 603 | own.append(Resolved(tag, .heading(index))) | ||
| 604 | } | ||
| 605 | return inherited + own | ||
| 606 | } | ||
| 607 | |||
| 608 | // MARK: - Building | ||
| 609 | |||
| 610 | private mutating func walk(_ node: SyntaxNode, heading: Int?, pending: inout [PendingBlock]) { | ||
| 611 | var previous: SyntaxNode? | ||
| 612 | var affiliated: [SyntaxNode] = [] | ||
| 613 | for child in node.children { | ||
| 614 | // Blank lines between nodes detach affiliated keywords. | ||
| 615 | if let previous, previous.range.upperBound != child.range.lowerBound { affiliated = [] } | ||
| 616 | switch child.kind { | ||
| 617 | case .section: | ||
| 618 | section(child, parent: heading, pending: &pending) | ||
| 619 | case .heading, .affiliatedKeyword: | ||
| 620 | break | ||
| 621 | case .planning: | ||
| 622 | if let heading { planning(child, heading) } | ||
| 623 | case .propertyDrawer: | ||
| 624 | if let heading { | ||
| 625 | headings[heading].properties += child.children.filter { $0.kind == .nodeProperty }.compactMap { Property(line: $0.text) } | ||
| 626 | } | ||
| 627 | case .clock: | ||
| 628 | if let heading, let clock = clock(child) { headings[heading].clocks.append(clock) } | ||
| 629 | case .timestamp: | ||
| 630 | if let heading, let stamp = Timestamp.parse(child.text) { headings[heading].timestamps.append(stamp) } | ||
| 631 | case .link: | ||
| 632 | if let heading { headings[heading].links.append(linkTarget(child)) } | ||
| 633 | case .keyword: | ||
| 634 | fileKeyword(child) | ||
| 635 | case .table: | ||
| 636 | tables.append(TableModel(node: child, heading: heading)) | ||
| 637 | walk(child, heading: heading, pending: &pending) | ||
| 638 | case .block: | ||
| 639 | if Self.isSrcBlock(child) { pending.append((child, heading, affiliated)) } | ||
| 640 | default: | ||
| 641 | walk(child, heading: heading, pending: &pending) | ||
| 642 | } | ||
| 643 | affiliated = child.kind == .affiliatedKeyword ? affiliated + [child] : [] | ||
| 644 | previous = child | ||
| 645 | } | ||
| 646 | } | ||
| 647 | |||
| 648 | private mutating func section(_ node: SyntaxNode, parent: Int?, pending: inout [PendingBlock]) { | ||
| 649 | guard let headingNode = node.children.first(where: { $0.kind == .heading }) else { return } | ||
| 650 | let index = headings.count | ||
| 651 | let tokens = headingNode.tokens | ||
| 652 | let todo = tokens.first { $0.kind == .todoKeyword }?.text | ||
| 653 | let title = headingNode.children.first { $0.kind == .title } | ||
| 654 | headings.append(HeadingInfo( | ||
| 655 | sectionRange: node.range, | ||
| 656 | headingRange: headingNode.range, | ||
| 657 | level: tokens.first { $0.kind == .stars }?.text.count ?? 0, | ||
| 658 | parent: parent, | ||
| 659 | todo: todo, | ||
| 660 | isDone: todo.map(orgSettings.isDone) ?? false, | ||
| 661 | priority: tokens.first { $0.kind == .priority }.map { String($0.text.dropFirst(2).dropLast()) }, | ||
| 662 | title: title?.text ?? "", | ||
| 663 | localTags: tokens.first { $0.kind == .tags }.map { $0.text.split(separator: ":").map(String.init) } ?? [], | ||
| 664 | properties: [], scheduled: nil, deadline: nil, closed: nil, clocks: [], timestamps: [], links: [] | ||
| 665 | )) | ||
| 666 | if let title { walk(title, heading: index, pending: &pending) } | ||
| 667 | walk(node, heading: index, pending: &pending) | ||
| 668 | } | ||
| 669 | |||
| 670 | /// Each timestamp belongs to the keyword just before it. | ||
| 671 | private mutating func planning(_ node: SyntaxNode, _ heading: Int) { | ||
| 672 | let text = node.text | ||
| 673 | for stampNode in node.children where stampNode.kind == .timestamp { | ||
| 674 | guard let stamp = Timestamp.parse(stampNode.text) else { continue } | ||
| 675 | let end = String.Index(utf16Offset: stampNode.offset - node.offset, in: text) | ||
| 676 | let before = text[..<end].trimmingCharacters(in: .whitespaces) | ||
| 677 | if before.hasSuffix("SCHEDULED:") { | ||
| 678 | headings[heading].scheduled = stamp | ||
| 679 | } else if before.hasSuffix("DEADLINE:") { | ||
| 680 | headings[heading].deadline = stamp | ||
| 681 | } else if before.hasSuffix("CLOSED:") { | ||
| 682 | headings[heading].closed = stamp | ||
| 683 | } | ||
| 684 | } | ||
| 685 | } | ||
| 686 | |||
| 687 | private func clock(_ node: SyntaxNode) -> Clock? { | ||
| 688 | guard let stampNode = node.children.first(where: { $0.kind == .timestamp }), | ||
| 689 | let stamp = Timestamp.parse(stampNode.text) else { return nil } | ||
| 690 | var minutes: Int? | ||
| 691 | if let arrow = node.text.range(of: "=>") { | ||
| 692 | let parts = node.text[arrow.upperBound...].trimmingCharacters(in: .whitespacesAndNewlines).split(separator: ":") | ||
| 693 | if parts.count == 2, let hours = Int(parts[0]), let mins = Int(parts[1]) { minutes = hours * 60 + mins } | ||
| 694 | } | ||
| 695 | return Clock(start: stamp.start, end: stamp.end, minutes: minutes) | ||
| 696 | } | ||
| 697 | |||
| 698 | private func linkTarget(_ node: SyntaxNode) -> String { | ||
| 699 | if let path = node.tokens.first(where: { $0.kind == .linkPath }) { return path.text } | ||
| 700 | let text = node.text | ||
| 701 | if text.hasPrefix("<"), text.hasSuffix(">") { return String(text.dropFirst().dropLast()) } | ||
| 702 | return text | ||
| 703 | } | ||
| 704 | |||
| 705 | private mutating func fileKeyword(_ node: SyntaxNode) { | ||
| 706 | let text = node.text.trimmingCharacters(in: .whitespacesAndNewlines) | ||
| 707 | guard let colon = text.firstIndex(of: ":") else { return } | ||
| 708 | let key = text[text.index(text.startIndex, offsetBy: 2)..<colon].uppercased() | ||
| 709 | let value = text[text.index(after: colon)...].trimmingCharacters(in: .whitespaces) | ||
| 710 | switch key { | ||
| 711 | case "FILETAGS": | ||
| 712 | fileTags += value.split(whereSeparator: { $0 == ":" || $0.isWhitespace }).map(String.init) | ||
| 713 | case "PROPERTY": | ||
| 714 | let parts = value.split(maxSplits: 1, whereSeparator: \.isWhitespace) | ||
| 715 | guard var name = parts.first.map(String.init) else { return } | ||
| 716 | let additive = name.hasSuffix("+") | ||
| 717 | if additive { name.removeLast() } | ||
| 718 | fileProperties.append(Property(key: name, value: parts.count > 1 ? String(parts[1]) : "", additive: additive)) | ||
| 719 | default: | ||
| 720 | break | ||
| 721 | } | ||
| 722 | } | ||
| 723 | |||
| 724 | // MARK: - Source blocks | ||
| 725 | |||
| 726 | static func isSrcBlock(_ node: SyntaxNode) -> Bool { | ||
| 727 | node.text.drop { $0 == " " || $0 == "\t" }.lowercased().hasPrefix("#+begin_src") | ||
| 728 | } | ||
| 729 | |||
| 730 | private func srcBlock(_ node: SyntaxNode, heading: Int?, affiliated: [SyntaxNode]) -> SrcBlockInfo { | ||
| 731 | let raw = splitRawLines(node.text) | ||
| 732 | let beginLine = raw[0].content.drop { $0 == " " || $0 == "\t" }.dropFirst("#+begin_src".count) | ||
| 733 | let words = beginLine.split(whereSeparator: \.isWhitespace) | ||
| 734 | var language: String? | ||
| 735 | var switches: [String] = [] | ||
| 736 | var rest = words[...] | ||
| 737 | if let first = rest.first, !first.hasPrefix("-"), !first.hasPrefix("+"), !first.hasPrefix(":") { | ||
| 738 | language = String(first) | ||
| 739 | rest = rest.dropFirst() | ||
| 740 | } | ||
| 741 | while let word = rest.first, !word.hasPrefix(":") { | ||
| 742 | switches.append(String(word)) | ||
| 743 | rest = rest.dropFirst() | ||
| 744 | } | ||
| 745 | let body = raw.dropFirst().dropLast().map { String($0.content) + String($0.ending) }.joined() | ||
| 746 | |||
| 747 | var name: String? | ||
| 748 | var headerLines: [String] = [] | ||
| 749 | for keyword in affiliated { | ||
| 750 | let text = keyword.text.trimmingCharacters(in: .whitespacesAndNewlines) | ||
| 751 | let value = SettingsScanner.keywordValue(text[...]).trimmingCharacters(in: .whitespaces) | ||
| 752 | if text.uppercased().hasPrefix("#+NAME:") { name = value } | ||
| 753 | if text.uppercased().hasPrefix("#+HEADER:") { headerLines.append(value) } | ||
| 754 | } | ||
| 755 | |||
| 756 | var arguments = HeaderArguments.defaults.mapValues { Resolved($0, .defaults) } | ||
| 757 | var keys = ["header-args"] | ||
| 758 | if let language { keys.append("header-args:\(language)") } | ||
| 759 | for key in keys { | ||
| 760 | for pair in headerArgumentLayers(key, heading: heading) { | ||
| 761 | arguments[pair.key] = Resolved(pair.value, pair.source) | ||
| 762 | } | ||
| 763 | } | ||
| 764 | for line in headerLines + [rest.joined(separator: " ")] { | ||
| 765 | for pair in HeaderArguments.parse(line) { | ||
| 766 | arguments[pair.key] = Resolved(pair.value, .element) | ||
| 767 | } | ||
| 768 | } | ||
| 769 | return SrcBlockInfo( | ||
| 770 | range: node.range, heading: heading, name: name, language: language, | ||
| 771 | switches: switches, body: body, headerArguments: arguments | ||
| 772 | ) | ||
| 773 | } | ||
| 774 | |||
| 775 | /// `header-args` pairs from `#+PROPERTY` down to the heading. As in org, a plain property | ||
| 776 | /// replaces everything above it; `header-args+` adds to it. | ||
| 777 | private func headerArgumentLayers(_ key: String, heading: Int?) -> [(key: String, value: String, source: ValueSource)] { | ||
| 778 | var layers: [(entries: [Property], source: ValueSource)] = [(fileProperties, .file)] | ||
| 779 | if let heading { | ||
| 780 | layers += (ancestors(of: heading) + [heading]).map { (headings[$0].properties, .heading($0)) } | ||
| 781 | } | ||
| 782 | var pairs: [(key: String, value: String, source: ValueSource)] = [] | ||
| 783 | for layer in layers { | ||
| 784 | for entry in layer.entries where entry.key.lowercased() == key { | ||
| 785 | let parsed = HeaderArguments.parse(entry.value).map { (key: $0.key, value: $0.value, source: layer.source) } | ||
| 786 | pairs = entry.additive ? pairs + parsed : parsed | ||
| 787 | } | ||
| 788 | } | ||
| 789 | return pairs | ||
| 790 | } | ||
| 791 | } | ||
| 792 | ``` | ||
| 793 | |||
| 794 | - [ ] **Step 4: Run all tests** | ||
| 795 | |||
| 796 | Run: `swift test` | ||
| 797 | Expected: all pass. | ||
| 798 | |||
| 799 | - [ ] **Step 5: Commit** | ||
| 800 | |||
| 801 | ```bash | ||
| 802 | git add Sources/OrgCore/Semantic/DocumentModel.swift Tests/OrgCoreTests/SemanticTests.swift | ||
| 803 | git commit -m "Add document model with inheritance and header arguments" | ||
| 804 | ``` | ||
| 805 | |||
| 806 | --- | ||
| 807 | |||
| 808 | ### Task 3: Build the model in fuzz and corpus tests | ||
| 809 | |||
| 810 | **Files:** | ||
| 811 | - Modify: `Tests/OrgCoreTests/RoundTripTests.swift` | ||
| 812 | |||
| 813 | - [ ] **Step 1: Build a model for every parsed document** | ||
| 814 | |||
| 815 | ```diff | ||
| 816 | @@ -45,6 +45,7 @@ struct RoundTripTests { | ||
| 817 | let tree = OrgParser.parse(text) | ||
| 818 | #expect(tree.text == text, "document \(n)") | ||
| 819 | #expect(checkLengths(tree.root), "document \(n)") | ||
| 820 | + _ = DocumentModel(tree: tree) | ||
| 821 | } | ||
| 822 | } | ||
| 823 | |||
| 824 | @@ -74,6 +75,7 @@ struct RoundTripTests { | ||
| 825 | let source = SourceText(bytes: bytes) | ||
| 826 | let tree = OrgParser.parse(source.text) | ||
| 827 | #expect(source.encode(tree.text) == bytes, "\(file.path)") | ||
| 828 | + _ = DocumentModel(tree: tree) | ||
| 829 | } | ||
| 830 | } | ||
| 831 | } | ||
| 832 | ``` | ||
| 833 | |||
| 834 | - [ ] **Step 2: Run** | ||
| 835 | |||
| 836 | Run: `swift test` then `ORGSTAR_CORPUS=~/Documents/notes swift test --filter corpusRoundTrips` | ||
| 837 | Expected: all pass, no crashes. | ||
| 838 | |||
| 839 | - [ ] **Step 3: Commit** | ||
| 840 | |||
| 841 | ```bash | ||
| 842 | git add Tests/OrgCoreTests/RoundTripTests.swift | ||
| 843 | git commit -m "Build document models in fuzz and corpus tests" | ||
| 844 | ``` | ||