mirror of https://github.com/requarks/wiki
NGPixel
3 days ago
parent
7ca0c302f7
commit
f64d643123
@ -0,0 +1,168 @@
|
||||
// -> Relative, like the renderers' other in-repo imports: this module is reachable from the headless
|
||||
// renderer bundle, which is built on its own
|
||||
import { normalizePagePath } from '../../helpers/pagePaths'
|
||||
|
||||
/**
|
||||
* MediaWiki's link syntax: `[[Page Name]]`, `[[Page Name|shown text]]`, `[[Page Name#Section]]` and
|
||||
* `[[#Section]]`.
|
||||
*
|
||||
* A wikilink is a link to a page by its NAME, so the target is turned into the path that name would
|
||||
* be filed under -- `normalizePagePath`, the same spelling every path field in the app corrects to:
|
||||
* spaces become dashes and everything is lowercased. `[[Getting Started]]` is `/getting-started`, and
|
||||
* `[[Guides/Getting Started]]` is `/guides/getting-started`.
|
||||
*
|
||||
* Always from the root, as in MediaWiki, and never relative to the page it is written on. A site that
|
||||
* brackets its URLs by locale reads an unprefixed path as its primary locale, which is where these
|
||||
* land.
|
||||
*
|
||||
* The section is spelled the way `slugifyHeading` in `backend/models/rendering.ts` spells the id it
|
||||
* gives a heading, so `[[Page#Some Heading]]` finds the heading called "Some Heading". That makes this
|
||||
* one more copy of that rule, and the two have to agree.
|
||||
*
|
||||
* The tokens are an ordinary `link_open` / `link_close` pair, so everything that already reads links
|
||||
* reads these too -- the external-link class, the server's backlinks, the Visual editor. The target
|
||||
* as written is kept on the opening token's `meta.wikilink`, which is how the Visual editor knows to
|
||||
* write the link back in this form rather than as `[text](/path)`.
|
||||
*/
|
||||
|
||||
/**
|
||||
* A heading as an anchor fragment. Mirrors `slugifyHeading` in `backend/models/rendering.ts`.
|
||||
*/
|
||||
function slugifySection(text) {
|
||||
return (
|
||||
text
|
||||
.toLowerCase()
|
||||
.trim()
|
||||
.replaceAll(/[^\p{L}\p{N}\s-]/gu, '')
|
||||
.replaceAll(/\s+/g, '-')
|
||||
.replaceAll(/-{2,}/g, '-')
|
||||
.replace(/^-+|-+$/g, '')
|
||||
.slice(0, 100) || 'section'
|
||||
)
|
||||
}
|
||||
|
||||
/**
|
||||
* Where a wikilink target points, as an href.
|
||||
*
|
||||
* @param {string} target The target as written, backslash escapes already resolved.
|
||||
* @returns {string} A root-relative path, optionally with a fragment; or the fragment alone for a
|
||||
* link to a section of the page it is written on.
|
||||
*/
|
||||
export function wikiLinkHref(target) {
|
||||
const hash = target.indexOf('#')
|
||||
const page = hash < 0 ? target : target.slice(0, hash)
|
||||
const section = hash < 0 ? '' : target.slice(hash + 1)
|
||||
const fragment = section.trim() ? `#${slugifySection(section)}` : ''
|
||||
if (!page.trim()) {
|
||||
return fragment || '/'
|
||||
}
|
||||
return `/${normalizePagePath(page)}${fragment}`
|
||||
}
|
||||
|
||||
/**
|
||||
* What a target may not contain. MediaWiki refuses the same characters in a title, and every one of
|
||||
* them is something else to the parser here: `|` is the label, `[` / `]` the brackets, `{` / `}` MDC
|
||||
* and `markdown-it-attrs`, `<` / `>` HTML.
|
||||
*/
|
||||
const INVALID_TARGET = /[[\]{}<>|\n]/
|
||||
|
||||
/** A target that is already a URL, `[[https://…]]`, which is not a page name. */
|
||||
const URL_TARGET = /^[a-z][a-z\d+.-]*:\/\//i
|
||||
|
||||
/**
|
||||
* Find the `]]` that closes a wikilink opened at `start`, skipping backslash escapes.
|
||||
*
|
||||
* @returns {{ pipe: number, end: number } | null} The position of the first `|` (or -1) and of the
|
||||
* closing `]]`, or null when there is none before the end of what is being tokenized.
|
||||
*/
|
||||
function scan(src, start, max) {
|
||||
let pipe = -1
|
||||
for (let pos = start; pos < max; pos++) {
|
||||
const ch = src.charCodeAt(pos)
|
||||
if (ch === 0x5c /* \ */) {
|
||||
pos++
|
||||
} else if (ch === 0x0a /* \n */) {
|
||||
return null
|
||||
} else if (ch === 0x7c /* | */ && pipe < 0) {
|
||||
pipe = pos
|
||||
} else if (ch === 0x5d /* ] */ && src.charCodeAt(pos + 1) === 0x5d && pos + 1 < max) {
|
||||
return { pipe, end: pos }
|
||||
}
|
||||
}
|
||||
return null
|
||||
}
|
||||
|
||||
function wikilink(state, silent) {
|
||||
const { src, posMax } = state
|
||||
const start = state.pos
|
||||
if (src.charCodeAt(start) !== 0x5b /* [ */ || src.charCodeAt(start + 1) !== 0x5b) {
|
||||
return false
|
||||
}
|
||||
// -> A link inside a link is not markup, as with every other link syntax
|
||||
if (state.linkLevel > 0) {
|
||||
return false
|
||||
}
|
||||
|
||||
const found = scan(src, start + 2, posMax)
|
||||
if (!found) {
|
||||
return false
|
||||
}
|
||||
// -> `[[1]](https://…)` is an ordinary link whose text happens to be `[1]`, the way citations are
|
||||
// often written, and `[[x]][ref]` the same by reference
|
||||
const next = src.charCodeAt(found.end + 2)
|
||||
if (next === 0x28 /* ( */ || next === 0x5b /* [ */) {
|
||||
return false
|
||||
}
|
||||
|
||||
const targetEnd = found.pipe < 0 ? found.end : found.pipe
|
||||
const rawTarget = src.slice(start + 2, targetEnd)
|
||||
const target = state.md.utils.unescapeAll(rawTarget).trim()
|
||||
if (!target || INVALID_TARGET.test(target) || URL_TARGET.test(target)) {
|
||||
return false
|
||||
}
|
||||
|
||||
const href = state.md.normalizeLink(wikiLinkHref(target))
|
||||
if (!state.md.validateLink(href)) {
|
||||
return false
|
||||
}
|
||||
|
||||
const labelStart = found.pipe < 0 ? -1 : found.pipe + 1
|
||||
// -> `[[Page|]]` has nothing to show, and is shown as the page name rather than as nothing
|
||||
const hasLabel = labelStart >= 0 && src.slice(labelStart, found.end).trim().length > 0
|
||||
|
||||
if (!silent) {
|
||||
const open = state.push('link_open', 'a', 1)
|
||||
open.attrs = [['href', href]]
|
||||
open.markup = 'wikilink'
|
||||
open.meta = { wikilink: target }
|
||||
|
||||
if (hasLabel) {
|
||||
// -> The label is markdown, as a link's text is: `[[Page|**bold** text]]`
|
||||
const oldPosMax = state.posMax
|
||||
state.pos = labelStart
|
||||
state.posMax = found.end
|
||||
state.linkLevel++
|
||||
state.md.inline.tokenize(state)
|
||||
state.linkLevel--
|
||||
state.posMax = oldPosMax
|
||||
} else {
|
||||
// -> The target alone is shown as written, and is not parsed: it is a name, not markup
|
||||
const text = state.push('text', '', 0)
|
||||
text.content = target
|
||||
}
|
||||
|
||||
const close = state.push('link_close', 'a', -1)
|
||||
close.markup = 'wikilink'
|
||||
}
|
||||
|
||||
state.pos = found.end + 2
|
||||
return true
|
||||
}
|
||||
|
||||
export default (md) => {
|
||||
/*
|
||||
Ahead of MDC's inline span, which claims every `[` it meets and would otherwise read `[[Page]]` as
|
||||
a span holding `[Page]`. That rule is registered before `link`, so this is ahead of both.
|
||||
*/
|
||||
md.inline.ruler.before('mdc_inline_span', 'wikilink', wikilink)
|
||||
}
|
||||
Loading…
Reference in new issue