|
|
@@ -5,7 +5,7 @@ import { gfmFromMarkdown } from 'mdast-util-gfm'
|
|
|
import { gfm } from 'micromark-extension-gfm'
|
|
|
import type { Nodes } from 'mdast'
|
|
|
|
|
|
-/** One authored Markdown line outside fenced code. */
|
|
|
+/** One authored Markdown line outside fenced code and rendered-away HTML comments. */
|
|
|
export interface MarkdownProseLine {
|
|
|
/** 1-based source line number. */
|
|
|
index: number
|
|
|
@@ -13,6 +13,14 @@ export interface MarkdownProseLine {
|
|
|
raw: string
|
|
|
}
|
|
|
|
|
|
+/** One parsed Markdown heading, retaining its authored first line and rendered text. */
|
|
|
+export interface MarkdownHeadingLine extends MarkdownProseLine {
|
|
|
+ /** Parsed ATX or Setext heading depth. */
|
|
|
+ depth: 1 | 2 | 3 | 4 | 5 | 6
|
|
|
+ /** Rendered heading text, excluding raw HTML such as comments. */
|
|
|
+ text: string
|
|
|
+}
|
|
|
+
|
|
|
/** Parse GitHub-flavored Markdown with the repository's standard extensions. */
|
|
|
export function parseMarkdown(source: string): Nodes {
|
|
|
return fromMarkdown(source, { extensions: [gfm()], mdastExtensions: [gfmFromMarkdown()] })
|
|
|
@@ -30,15 +38,93 @@ export function visitMarkdown(node: Nodes, visitor: (node: Nodes) => boolean | v
|
|
|
}
|
|
|
}
|
|
|
|
|
|
+/** Text a reader sees from one Markdown node; raw HTML itself contributes none. */
|
|
|
+function renderedText(node: Nodes): string {
|
|
|
+ if (node.type === 'text' || node.type === 'inlineCode') return node.value
|
|
|
+ if (node.type === 'image' || node.type === 'imageReference') return node.alt ?? ''
|
|
|
+ if (node.type === 'break') return ' '
|
|
|
+ if ('children' in node) return node.children.map(child => renderedText(child)).join('')
|
|
|
+ return ''
|
|
|
+}
|
|
|
+
|
|
|
+/** Return every parsed Markdown heading with its rendered text and source line. */
|
|
|
+export function markdownHeadingLines(source: string): MarkdownHeadingLine[] {
|
|
|
+ const rawLines = source.split('\n')
|
|
|
+ const headings: MarkdownHeadingLine[] = []
|
|
|
+ visitMarkdown(parseMarkdown(source), (node) => {
|
|
|
+ if (node.type !== 'heading' || node.position === undefined) return
|
|
|
+ headings.push({
|
|
|
+ depth: node.depth,
|
|
|
+ index: node.position.start.line,
|
|
|
+ raw: rawLines[node.position.start.line - 1] ?? '',
|
|
|
+ text: renderedText(node),
|
|
|
+ })
|
|
|
+ })
|
|
|
+ return headings
|
|
|
+}
|
|
|
+
|
|
|
+type ColumnRange = readonly [start: number, end: number]
|
|
|
+type OffsetRange = readonly [start: number, end: number]
|
|
|
+
|
|
|
+/** Source-column ranges occupied by parsed HTML comments, keyed by source line. */
|
|
|
+function htmlCommentRanges(source: string, rawLines: readonly string[]): Map<number, ColumnRange[]> {
|
|
|
+ const comments: OffsetRange[] = []
|
|
|
+ visitMarkdown(parseMarkdown(source), (node) => {
|
|
|
+ if (node.type !== 'html' || node.position?.start.offset === undefined) return
|
|
|
+ let cursor = 0
|
|
|
+ while (true) {
|
|
|
+ const start = node.value.indexOf('<!--', cursor)
|
|
|
+ if (start < 0) break
|
|
|
+ const close = node.value.indexOf('-->', start + '<!--'.length)
|
|
|
+ const end = close < 0 ? node.value.length : close + '-->'.length
|
|
|
+ comments.push([node.position.start.offset + start, node.position.start.offset + end])
|
|
|
+ cursor = end
|
|
|
+ }
|
|
|
+ })
|
|
|
+
|
|
|
+ const ranges = new Map<number, ColumnRange[]>()
|
|
|
+ let lineOffset = 0
|
|
|
+ rawLines.forEach((raw, index) => {
|
|
|
+ const lineEnd = lineOffset + raw.length
|
|
|
+ for (const [start, end] of comments) {
|
|
|
+ const from = Math.max(start, lineOffset)
|
|
|
+ const to = Math.min(end, lineEnd)
|
|
|
+ const coversEmptyLine = raw.length === 0 && start <= lineOffset && end > lineOffset
|
|
|
+ if (from < to || coversEmptyLine) {
|
|
|
+ const lineRanges = ranges.get(index + 1) ?? []
|
|
|
+ lineRanges.push([from - lineOffset, to - lineOffset])
|
|
|
+ ranges.set(index + 1, lineRanges)
|
|
|
+ }
|
|
|
+ }
|
|
|
+ lineOffset = lineEnd + 1
|
|
|
+ })
|
|
|
+ return ranges
|
|
|
+}
|
|
|
+
|
|
|
+/** Whether a source line retains non-whitespace text after HTML comments disappear. */
|
|
|
+function hasRenderedTextOutsideComments(raw: string, ranges: readonly ColumnRange[] | undefined): boolean {
|
|
|
+ if (ranges === undefined) return true
|
|
|
+ let cursor = 0
|
|
|
+ let visible = ''
|
|
|
+ for (const [start, end] of [...ranges].sort((left, right) => left[0] - right[0])) {
|
|
|
+ visible += raw.slice(cursor, start)
|
|
|
+ cursor = Math.max(cursor, end)
|
|
|
+ }
|
|
|
+ visible += raw.slice(cursor)
|
|
|
+ return visible.trim().length > 0
|
|
|
+}
|
|
|
+
|
|
|
/**
|
|
|
- * Return source lines outside backtick or tilde fences.
|
|
|
+ * Return source lines outside backtick or tilde fences and HTML comments.
|
|
|
* @param source - Markdown source whose prose should be retained verbatim.
|
|
|
* @returns unfenced lines with their original 1-based locations.
|
|
|
*/
|
|
|
export function markdownProseLines(source: string): MarkdownProseLine[] {
|
|
|
let fence: { marker: '`' | '~'; length: number } | undefined
|
|
|
const kept: MarkdownProseLine[] = []
|
|
|
- source.split('\n').forEach((raw, i) => {
|
|
|
+ const rawLines = source.split('\n')
|
|
|
+ const comments = htmlCommentRanges(source, rawLines)
|
|
|
+ rawLines.forEach((raw, i) => {
|
|
|
const token = /^ {0,3}(`{3,}|~{3,})/.exec(raw)?.[1]
|
|
|
if (token !== undefined) {
|
|
|
const marker = token[0] as '`' | '~'
|
|
|
@@ -49,7 +135,9 @@ export function markdownProseLines(source: string): MarkdownProseLine[] {
|
|
|
}
|
|
|
return
|
|
|
}
|
|
|
- if (fence === undefined) kept.push({ index: i + 1, raw })
|
|
|
+ if (fence === undefined && hasRenderedTextOutsideComments(raw, comments.get(i + 1))) {
|
|
|
+ kept.push({ index: i + 1, raw })
|
|
|
+ }
|
|
|
})
|
|
|
return kept
|
|
|
}
|
Tianyi Cui