Press n or j to go to the next uncovered block, b, p or k for the previous block.
| 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 | 4832240x 4793412x 4472902x 4375072x 2904750x 2012979x 1710905x 1523867x 529532x 320510x 320510x 5313036x 5301500x 5297034x 5296780x 5296360x 5296360x 5296360x 303834x 34676x 7062x 7062x 13312x 24x 17431x 7278x 7278x 105994x 7170x 7170x 17431x 247765x 247765x 232669x 224620x 144370x 144370x 7693x 7693x 7693x 7693x 7693x 136677x 6770x 6770x 136677x 17531x 17531x 2158350x 2158350x 2158350x 18234x 18234x 2140116x 2038651x 144370x 144370x 1894281x 52248x 17531x 561x 52248x 631x 631x 631x 631x 631x 51617x 2524x 1336x 1188x 1188x 781x 781x 17145x 17145x 1580837x 1580837x 1580837x 11368x 11368x 1569469x 1506432x 103395x 103395x 1403037x 29249x 17145x 103395x 103395x 103395x 103395x 29249x 13114x 16135x 16135x 6281x 6281x 8781x 8781x 1093053x 1093053x 1093053x 9226x 9226x 1083827x 1032390x 72745x 72745x 72745x 72745x 959645x 16259x 8781x 16259x 16151x 105x 105x 16151x 105x 108x 3x 3x 3x 3x 3x | // SPDX-FileCopyrightText: 2024-2026 Hack23 AB
// SPDX-License-Identifier: Apache-2.0
/**
* @module Aggregator/Metadata/BriefingHighlightSections
* @description Section walker primitives for {@link briefing-highlight.ts}.
* Extracted as a sibling leaf module so the public extractor surface stays
* below the 600-line drift-guard cap (see
* `test/unit/source-file-size.test.js`). The walkers are pure — every
* function takes (markdown body, section-needle list, language) and returns
* a structural fragment with no I/O or upward imports.
*
* Three walker shapes are exported:
*
* - {@link extractFirstSubsectionUnderSection} — first `### …` heading +
* paragraph inside the first matched `## …` block (used by the
* Strategic-synthesis and Top-findings extractors).
* - {@link extractFirstParagraphUnderSection} — first prose paragraph
* inside the matched `## …` block, ignoring `### …` sub-headings
* (used by Reader-Briefing paragraph fallbacks and Strategic-paragraph
* derivation).
* - {@link extractFirstNumberedItemUnderSection} — first `1. **Label**:
* …` list item inside the matched `## …` block (used by the
* May-2026 Reader-Briefing priority-list style).
*
* The {@link stripTradecraftLabels} helper is exported too — both this
* module and `briefing-highlight.ts` need it to normalize sub-headings
* before they reach the headline / description surface.
*/
import type { LanguageCode } from '../../types/languages.js';
import { resolveBoilerplatePatterns } from './briefing-highlight-i18n.js';
import {
EXTENDED_DESCRIPTION_MAX_LENGTH,
shouldSkipDescriptionLine,
stripInlineMarkdown,
stripLeadingBoldLabel,
stripLeadingProseLabel,
} from './text-utils.js';
/** Sentinel returned by {@link classifyLine} when the line is benign prose. */
type LineKind = 'fence' | 'h2' | 'h3' | 'blank' | 'structural' | 'numbered' | 'bullet' | 'prose';
/**
* Classify a trimmed Markdown line into one of the structural buckets
* the section walker cares about. Extracted from the inline walker
* loops to keep their cognitive complexity below the 15-point limit.
*
* @param line - Trimmed Markdown line
* @returns Line kind sentinel
*/
function classifyLine(line: string): LineKind {
if (line.startsWith('```') || line.startsWith('~~~')) return 'fence';
if (line.startsWith('## ')) return 'h2';
if (line.startsWith('### ')) return 'h3';
if (line === '') return 'blank';
if (line.startsWith('|') || line.startsWith('>') || line.startsWith('<')) return 'structural';
if (line.startsWith('---') || line.startsWith('===')) return 'structural';
if (/^\d+\.\s+/u.test(line)) return 'numbered';
if (line.startsWith('-') || line.startsWith('*')) return 'bullet';
return 'prose';
}
/**
* Compare a raw `## …` heading line against a whitelist of expected
* section names. The comparison strips inline Markdown decorations and
* leading non-alphanumeric characters (emoji, punctuation) so a brief
* that writes the heading as `## 🧭 Strategic Intelligence Summary`
* still matches.
*
* @param raw - Heading text without the leading `#`s
* @param needles - Lower-case whitelist entries
* @returns `true` when the heading text matches any whitelist entry
*/
function headingMatches(raw: string, needles: readonly string[]): boolean {
const normalized = stripInlineMarkdown(raw)
.replace(/[*_`#]+/g, '')
// Strip leading non-letter / non-digit characters across all
// Unicode scripts (emoji, ASCII punctuation, fullwidth punctuation
// such as `（`, `：`). Using `\p{L}\p{N}` keeps Arabic, Hebrew,
// CJK and other non-Latin headings intact — the previous
// `[A-Za-z0-9]` form silently stripped the entire heading text
// for those scripts.
.replace(/^[^\p{L}\p{N}]+/u, '')
// Strip leading numeric list prefix (e.g. "1. ", "7. ")
.replace(/^\d+\.\s+/, '')
// Normalise fullwidth ASCII variants commonly used in CJK headings
// so a heading written as `主要判断（要約）` matches the lower-cased
// ASCII needle `主要判断 (要約)` shape.
.replace(/[\uFF08]/g, '(')
.replace(/[\uFF09]/g, ')')
.replace(/[\uFF1A]/g, ':')
.replace(/\u3000/g, ' ')
.trim()
.toLowerCase();
for (const needle of needles) {
if (normalized === needle) return true;
if (normalized.startsWith(`${needle} `)) return true;
if (normalized.startsWith(`${needle}:`)) return true;
if (normalized.startsWith(`${needle}(`)) return true;
// Em-dash, en-dash, hyphen separators.
Iif (normalized.startsWith(`${needle} —`)) return true;
Iif (normalized.startsWith(`${needle} –`)) return true;
Iif (normalized.startsWith(`${needle} -`)) return true;
}
return false;
}
/**
* Mutable iterator state used by the section walkers. The fence-state
* flag tracks ``` / ~~~ pairs so we never enter a section that lives
* inside a code block; `inSection` flips on/off as we cross `## …`
* boundaries; `subHeading` captures the first `### …` we see inside
* the matched section (only used by the synthesis extractor).
*/
interface WalkerState {
inFence: boolean;
inSection: boolean;
subHeading: string;
lines: string[];
byteCount: number;
}
/**
* Build an empty walker state.
*
* @returns Fresh, fence-aware {@link WalkerState} with empty buffers.
*/
function newState(): WalkerState {
return { inFence: false, inSection: false, subHeading: '', lines: [], byteCount: 0 };
}
/**
* Push a prose line into the walker's collected buffer.
*
* @param state - Walker state (mutated)
* @param line - Cleaned line to append
*/
function appendLine(state: WalkerState, line: string): void {
state.lines.push(line);
state.byteCount += line.length + 1;
}
/**
* Strip intelligence tradecraft labels (WEP probability bands, KJ-N
* prefixes, Admiralty grades) from a paragraph so they don't pollute
* reader-facing headlines. These are analyst-internal markers that
* readers find confusing.
*
* @param text - Raw paragraph text
* @returns Text with tradecraft labels removed
*/
export function stripTradecraftLabels(text: string): string {
return (
text
// "KJ-1 [WEP: HIGHLY LIKELY, 90–95%]: " prefix
.replace(/^KJ-?\d+\s*\[.*?\]:\s*/iu, '')
// "[WEP: LIKELY (60-75%)]" or "(WEP Probable, 60–75% confidence)" inline
.replace(/\[WEP:?\s*[^\]]+\]\s*/giu, '')
.replace(/\(WEP\s+[^)]+\)\s*/giu, '')
// "Admiralty Grade: B2" or similar
// eslint-disable-next-line security/detect-unsafe-regex
.replace(/Admiralty\s+(?:Source\s+)?Grade:?\s*[A-Z]\d\s*/giu, '')
// Leading numbered list prefix "1. ", "2. " etc.
.replace(/^\d+\.\s+/, '')
// "(Admiralty B2)" or "*(Admiralty B2)*" trailing references
.replace(/\*?\(Admiralty\s+[A-Z]\d\)\*?\s*$/giu, '')
.trim()
);
}
/**
* Boilerplate sentence patterns that should never surface as headlines
* or descriptions. These are self-referential meta-prose that describes
* the brief itself rather than the substantive intelligence content.
*/
const BOILERPLATE_PATTERNS: readonly RegExp[] = [
/^this (?:executive )?brief (?:synthesi[sz]es|provides|covers|summariz|presents|contains|offers)/iu,
/^this (?:report|document|analysis|assessment) (?:synthesi[sz]es|provides|covers|summariz|presents)/iu,
/^the brief is designed to be read/iu,
/^all forward-looking assessments are/iu,
/^confidence:/iu,
/^admiralty (?:source )?grade/iu,
/^classification:/iu,
/^data[\s-]?mode/iu,
/^key assessments?\s*\(/iu,
/^overall (?:admiralty|assessment) grade/iu,
/^sats? applied/iu,
/^generated by (?:eu parliament|automated)/iu,
/^subject:/iu,
];
function normalizeBriefingLine(
line: string,
preserveLeadingLabel = false,
lang: LanguageCode = 'en'
): string {
if (shouldSkipDescriptionLine(line)) return '';
const withoutMarkdown = stripInlineMarkdown(line);
const patterns = resolveBoilerplatePatterns(BOILERPLATE_PATTERNS, lang);
if (patterns.some((re) => re.test(withoutMarkdown.trim()))) return '';
const stripped = stripTradecraftLabels(withoutMarkdown);
const normalized = preserveLeadingLabel
? stripped
: stripLeadingProseLabel(stripLeadingBoldLabel(stripped));
return normalized.replace(/^[:;—–-]\s+/u, '').trim();
}
/**
* Decide what to do when the walker sees a `## …` heading.
*
* @param state - Walker state
* @param raw - Raw heading line (already trimmed)
* @param needles - Lower-case section whitelist
* @returns `'enter'` when the heading opens the target section,
* `'leave'` when it closes an already-open target section,
* `'skip'` otherwise.
*/
function transitionForH2(
state: WalkerState,
raw: string,
needles: readonly string[]
): 'enter' | 'leave' | 'skip' {
const headingText = raw.replace(/^##\s+/, '');
if (headingMatches(headingText, needles)) return 'enter';
if (state.inSection) return 'leave';
return 'skip';
}
/**
* Handle a `## …` line for the sub-section walker. Returns `true`
* when the caller should stop walking.
*
* @param state - Walker state (mutated)
* @param line - Trimmed `## …` line
* @param needles - Section whitelist
* @returns `true` to stop walking
*/
function handleH2ForSubsection(
state: WalkerState,
line: string,
needles: readonly string[]
): boolean {
const t = transitionForH2(state, line, needles);
if (t === 'enter') {
state.inSection = true;
state.subHeading = '';
state.lines.length = 0;
state.byteCount = 0;
return false;
}
if (t === 'leave') {
Iif (state.subHeading && state.lines.length > 0) return true;
state.inSection = false;
}
return false;
}
/**
* Walk the brief body line-by-line and return the first `### …`
* heading + its first prose paragraph that occur **inside** the
* matched `## …` block. Returns `null` when the matched block does
* not contain a `### …` sub-heading.
*
* @param markdown - Brief body (SPDX preamble already stripped)
* @param sectionNeedles - Lower-case `## …` whitelist
* @param lang - Locale (drives boilerplate stem patterns)
* @returns First `{subHeading, paragraph}` pair under the matched
* section, or `null` when no sub-heading exists
*/
export function extractFirstSubsectionUnderSection(
markdown: string,
sectionNeedles: readonly string[],
lang: LanguageCode = 'en'
): { readonly subHeading: string; readonly paragraph: string } | null {
const state = newState();
for (const raw of markdown.split('\n')) {
const line = raw.trim();
const kind = classifyLine(line);
if (kind === 'fence') {
state.inFence = !state.inFence;
continue;
}
if (state.inFence) continue;
if (kind === 'h2') {
Iif (handleH2ForSubsection(state, line, sectionNeedles)) break;
continue;
}
if (!state.inSection) continue;
if (collectSubsectionLine(state, line, kind, lang)) break;
}
if (!state.subHeading || state.lines.length === 0) return null;
return {
subHeading: state.subHeading.trim(),
paragraph: state.lines.join(' ').trim(),
};
}
/**
* Process one non-heading line inside the matched section for the
* sub-section extractor. Returns `true` to signal the caller should
* stop walking (paragraph boundary reached or budget exceeded).
*
* @param state - Walker state (mutated)
* @param line - Trimmed line being processed
* @param kind - Pre-classified line kind from {@link classifyLine}
* @param lang - Locale (drives boilerplate stem patterns)
* @returns `true` to stop walking
*/
function collectSubsectionLine(
state: WalkerState,
line: string,
kind: LineKind,
lang: LanguageCode = 'en'
): boolean {
if (kind === 'h3') {
Iif (state.subHeading && state.lines.length > 0) return true;
state.subHeading = stripInlineMarkdown(line.replace(/^###\s+/, ''));
state.lines.length = 0;
state.byteCount = 0;
return false;
}
if (!state.subHeading) return false;
if (kind === 'blank' || kind === 'structural') {
return state.lines.length > 0;
}
const clean = normalizeBriefingLine(line, false, lang);
if (!clean) return state.lines.length > 0;
appendLine(state, clean);
return state.byteCount >= EXTENDED_DESCRIPTION_MAX_LENGTH;
}
/**
* Walk the brief body and return the first prose paragraph that occurs
* **inside** the matched `## …` block (ignoring any `### …`
* sub-headings). Used as the fallback extractor when the section is a
* single-paragraph block (the term-outlook Reader Briefing style).
*
* @param markdown - Brief body (SPDX preamble already stripped)
* @param sectionNeedles - Lower-case `## …` whitelist
* @param lang - Locale (drives boilerplate stem patterns)
* @returns First prose paragraph, or empty string when absent
*/
export function extractFirstParagraphUnderSection(
markdown: string,
sectionNeedles: readonly string[],
lang: LanguageCode = 'en'
): string {
const state = newState();
for (const raw of markdown.split('\n')) {
const line = raw.trim();
const kind = classifyLine(line);
if (kind === 'fence') {
state.inFence = !state.inFence;
continue;
}
if (state.inFence) continue;
if (kind === 'h2') {
Iif (handleH2ForParagraph(state, line, sectionNeedles)) break;
continue;
}
if (!state.inSection || kind === 'h3') continue;
if (collectParagraphLine(state, line, kind, lang)) break;
}
return state.lines.length === 0 ? '' : state.lines.join(' ').trim();
}
/**
* Handle a `## …` line for the first-paragraph walker. Returns `true`
* when the caller should stop walking (a complete paragraph was
* already captured in a prior matched section).
*
* @param state - Walker state (mutated)
* @param line - Trimmed `## …` line
* @param needles - Section whitelist
* @returns `true` to stop walking
*/
function handleH2ForParagraph(
state: WalkerState,
line: string,
needles: readonly string[]
): boolean {
Iif (state.inSection && state.lines.length > 0) return true;
const t = transitionForH2(state, line, needles);
state.inSection = t === 'enter';
return false;
}
/**
* Process one non-heading line inside the matched section for the
* first-paragraph extractor. Returns `true` when the caller should
* stop walking.
*
* @param state - Walker state (mutated)
* @param line - Trimmed line being processed
* @param kind - Pre-classified line kind from {@link classifyLine}
* @param lang - Locale (drives boilerplate stem patterns)
* @returns `true` to stop walking
*/
function collectParagraphLine(
state: WalkerState,
line: string,
kind: LineKind,
lang: LanguageCode = 'en'
): boolean {
if (kind === 'blank' || kind === 'structural') {
return state.lines.length > 0;
}
const clean = normalizeBriefingLine(line, false, lang);
if (!clean) return state.lines.length > 0;
appendLine(state, clean);
return state.byteCount >= EXTENDED_DESCRIPTION_MAX_LENGTH;
}
/** Mutable accumulator for {@link extractFirstNumberedItemUnderSection}. */
interface NumberedItemState {
inFence: boolean;
inSection: boolean;
item: string[];
}
/**
* Walk the brief body and return the first numbered-list item that
* appears **inside** the matched `## …` block. Recognises the
* `1. **Immediate priority**: …` shape used by the May-2026
* Reader Briefing style guide. The bold label and tail are joined into
* a single headline-shaped string.
*
* @param markdown - Brief body
* @param sectionNeedles - `## …` whitelist
* @param lang - Locale (drives boilerplate stem patterns)
* @returns Flattened first list item, or empty string when absent
*/
export function extractFirstNumberedItemUnderSection(
markdown: string,
sectionNeedles: readonly string[],
lang: LanguageCode = 'en'
): string {
const state: NumberedItemState = { inFence: false, inSection: false, item: [] };
for (const raw of markdown.split('\n')) {
const line = raw.trim();
const kind = classifyLine(line);
if (kind === 'fence') {
state.inFence = !state.inFence;
continue;
}
if (state.inFence) continue;
if (kind === 'h2') {
Iif (state.inSection && state.item.length > 0) break;
const headingText = line.replace(/^##\s+/, '');
state.inSection = headingMatches(headingText, sectionNeedles);
continue;
}
if (!state.inSection) continue;
if (handleNumberedLine(state, line, kind, lang)) break;
}
return state.item.join(' ').trim();
}
/**
* Process one line inside the matched section for the numbered-item
* extractor. Returns `true` when the caller should stop walking.
*
* @param state - Numbered-item walker state (mutated)
* @param line - Trimmed line being processed
* @param kind - Pre-classified line kind from {@link classifyLine}
* @param lang - Locale (drives boilerplate stem patterns)
* @returns `true` to stop walking
*/
function handleNumberedLine(
state: NumberedItemState,
line: string,
kind: LineKind,
lang: LanguageCode = 'en'
): boolean {
if (state.item.length === 0) {
if (kind !== 'numbered') return false;
const m = /^1\.\s+(.*)$/u.exec(line);
const clean = m?.[1] ? normalizeBriefingLine(m[1], true, lang) : '';
if (clean) state.item.push(clean);
return false;
}
if (kind === 'blank' || kind === 'numbered' || kind === 'bullet') return true;
Iif (kind === 'h2' || kind === 'h3') return true;
const clean = normalizeBriefingLine(line, false, lang);
Iif (!clean) return state.item.length > 0;
state.item.push(clean);
return false;
}
|