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 | 6x 6x 6x 6x 6x 6x 6x 576x 3204x 22x 6x 2x 6x 6x 6x 33x 2x 2x 2x 31x 31x 31x 17x 1226x 17x 15x 15x 10x 6x 2439x 17x 33x 6x 6x 6x 6x 6x 6x 5342x 5342x 6x 330x 330x 330x 330x 6x 1120x 2634x 6x 194x 194x 194x 2335x 194x 194x 1693x 1380x 1380x 1380x 1349x 194x 538x 556x 537x 555x 537x 537x 194x 194x 1691x 1691x 168x 168x 1523x 1523x 1523x 29x 29x 28x 27x 27x 29x 1494x 4x 4x 4x 4x 1490x 550x 550x 550x 550x 550x 550x 940x 2x 2x 2x 938x 938x 938x 938x 938x 938x 938x 1846x 1846x 1846x 955x 47x 908x 908x 908x 908x 908x 908x 740x 740x 740x 740x 738x 738x 738x 738x 2x 2x 740x 908x 908x 938x 938x 938x 926x 926x 926x 894x 893x 330x 330x 330x 330x 330x 330x 563x 563x 194x | // ---------------------------------------------------------------------------
// an HTML tree without a DOM
//
// precompile() has to read a component where there is no DOMParser - on node,
// for the Vite plugin, and in a service worker - and find every attribute and
// every text node the runtime's own parse finds. This builds that tree:
// elements with their attributes and children, and text decoded the way the
// HTML parser decodes it.
//
// It is not a conforming HTML parser and doesn't try to be. What precompile
// reads is *which* attributes and texts exist, and wherever the real parser
// would put one somewhere else - an implied end tag, a table's foster parent -
// it still exists. The places where it would read a different *value* are the
// ones that matter, and those it follows:
// - newlines normalized first (CRLF and a lone CR are LF), as the input stream is
// - character references decoded in text and attribute values, with the
// attribute rule for a legacy reference with no semicolon
// - <html>, <head> and <body> dropped, as a template drops them
// - attribute names lowercased, and the first of two duplicates kept
// - <script>/<style> (and the other raw text elements) read verbatim to their
// end tag, <textarea>/<title> decoded but never parsed for tags
// - comments dropped, and the text on either side kept as two texts
//
// tests/precompile.test.ts holds it against DOMParser, tree for tree
// ---------------------------------------------------------------------------
export type HTMLElementNode = { tag: string; attrs: Record<string, string>; children: HTMLNode[] }
export type HTMLNode = HTMLElementNode | string
const VOID_TAGS = new Set([
"area", "base", "br", "col", "embed", "hr", "img", "input",
"link", "meta", "param", "source", "track", "wbr",
])
// the document's own structure, which a template can't hold: inside one the
// parser drops these tags, start and end, and keeps what they wrapped
const DOCUMENT_TAGS = new Set(["html", "head", "body"])
// read verbatim up to their end tag - and, for the escapable two, decoded
const RAW_TEXT_TAGS = new Set(["script", "style", "xmp", "iframe", "noembed", "noframes"])
const ESCAPABLE_RAW_TEXT_TAGS = new Set(["textarea", "title"])
// the named references a component plausibly writes. The HTML table has over
// two thousand; one missing here leaves its text undecoded, and that is a
// precompiled function under a key the runtime never asks for - which safe
// mode reports by name, rather than a wrong function running
const NAMED_REFERENCES: Record<string, string> = {
apos: "'", trade: "\u2122", hellip: "\u2026", mdash: "\u2014", ndash: "\u2013",
lsquo: "\u2018", rsquo: "\u2019", ldquo: "\u201c", rdquo: "\u201d", bull: "\u2022", euro: "\u20ac",
larr: "\u2190", rarr: "\u2192", uarr: "\u2191", darr: "\u2193", check: "\u2713",
lbrace: "{", rbrace: "}", lcub: "{", rcub: "}", lpar: "(", rpar: ")", lsqb: "[", rsqb: "]",
colon: ":", comma: ",", period: ".", semi: ";", excl: "!", quest: "?", num: "#", dollar: "$",
percnt: "%", ast: "*", plus: "+", equals: "=", sol: "/", bsol: "\\", verbar: "|", vert: "|",
grave: "`", Hat: "^", lowbar: "_", commat: "@", Tab: "\t", NewLine: "\n",
}
// the references that also decode without their semicolon - the web before
// HTML required one - which is the whole of this list, from the standard: the
// five markup ones in both cases, and Latin-1's 96, U+00A0 to U+00FF in order.
// A name that isn't a reference is read as the longest of these it starts
// with (`¬areference;` is `¬areference;`), so all of them are needed to
// read what the parser reads
const LATIN_1 = (
"nbsp iexcl cent pound curren yen brvbar sect uml copy ordf laquo not shy reg macr deg plusmn sup2 sup3 " +
"acute micro para middot cedil sup1 ordm raquo frac14 frac12 frac34 iquest Agrave Aacute Acirc Atilde Auml " +
"Aring AElig Ccedil Egrave Eacute Ecirc Euml Igrave Iacute Icirc Iuml ETH Ntilde Ograve Oacute Ocirc Otilde " +
"Ouml times Oslash Ugrave Uacute Ucirc Uuml Yacute THORN szlig agrave aacute acirc atilde auml aring aelig " +
"ccedil egrave eacute ecirc euml igrave iacute icirc iuml eth ntilde ograve oacute ocirc otilde ouml divide " +
"oslash ugrave uacute ucirc uuml yacute thorn yuml"
).split(" ")
const LEGACY_REFERENCES: Record<string, string> = {
amp: "&", AMP: "&", lt: "<", LT: "<", gt: ">", GT: ">", quot: "\"", QUOT: "\"", COPY: "\u00a9", REG: "\u00ae",
...Object.fromEntries(LATIN_1.map((name, i) => [name, String.fromCharCode(0xa0 + i)])),
}
// longest first, so `²` is ² and not "sup" plus a 2
const LEGACY_NAMES = Object.keys(LEGACY_REFERENCES).sort((a, b) => b.length - a.length)
const has = (table: Record<string, string>, name: string) => Object.prototype.hasOwnProperty.call(table, name)
const fromCodePoint = (code: number): string =>
code === 0 || code > 0x10ffff || (code >= 0xd800 && code <= 0xdfff) ? "�" : String.fromCodePoint(code)
export const CHARACTER_REFERENCE_RE = /&(#[0-9]+;?|#[xX][0-9a-fA-F]+;?|[A-Za-z][A-Za-z0-9]*;?)/g
const ALPHANUMERIC_RE = /[A-Za-z0-9]/
// what one match of CHARACTER_REFERENCE_RE in `text`, at `offset`, decodes to.
// In an attribute a legacy reference with no semicolon stays literal when the
// next character is `=` or alphanumeric - `?a=1©=2` is a query string, not
// a copyright sign. In text it decodes, and what follows the name stays.
// Exported for the editor extension, which decodes an attribute value while
// keeping where each character came from (editors/vscode/src/entities.ts)
export const decodeReference = (text: string, whole: string, ref: string, offset: number, inAttribute: boolean): string => {
if (ref[0] === "#") {
const hex = ref[1] === "x" || ref[1] === "X"
const digits = ref.slice(hex ? 2 : 1).replace(/;$/, "")
return fromCodePoint(parseInt(digits, hex ? 16 : 10))
}
const semicolon = ref.endsWith(";")
const name = semicolon ? ref.slice(0, -1) : ref
if (semicolon && has(LEGACY_REFERENCES, name)) return LEGACY_REFERENCES[name]
Iif (semicolon && has(NAMED_REFERENCES, name)) return NAMED_REFERENCES[name]
const legacy = LEGACY_NAMES.find(candidate => name.startsWith(candidate))
if (legacy === undefined) return whole
const after = name.length > legacy.length ? name[legacy.length] : semicolon ? ";" : text[offset + whole.length]
if (inAttribute && after !== undefined && (after === "=" || ALPHANUMERIC_RE.test(after))) return whole
return LEGACY_REFERENCES[legacy] + ref.slice(legacy.length)
}
const decode = (text: string, inAttribute: boolean): string => {
if (!text.includes("&")) return text
return text.replace(CHARACTER_REFERENCE_RE, (whole: string, ref: string, offset: number) =>
decodeReference(text, whole, ref, offset, inAttribute))
}
const TAG_NAME_RE = /[^\s/>]*/y
const ATTRIBUTE_NAME_RE = /[^\s/>][^\s/>=]*/y
const WHITESPACE_RE = /\s*/y
const UNQUOTED_VALUE_RE = /[^\s>]*/y
const END_TAG_RE = /<\/([A-Za-z][^\s/>]*)[^>]*>/y
// runs a sticky pattern at `at` and answers with the match (or "")
const readAt = (re: RegExp, src: string, at: number): string => {
re.lastIndex = at
return re.exec(src)?.[0] ?? ""
}
// where a raw text element's content ends: its own end tag, whatever the case,
// followed by what can end a tag name
const rawTextEnd = (src: string, from: number, tag: string): number => {
const re = new RegExp(`</${tag}(?=[\\s/>])`, "gi")
re.lastIndex = from
const match = re.exec(src)
return match ? match.index : src.length
}
// the element children of a comment are nothing, and the texts on either side
// of it stay two texts - a comment is a node in the DOM, which is what keeps
// them apart. A marker holds its place while the tree is built
const COMMENT = null
type BuildNode = { tag: string; attrs: Record<string, string>; children: (BuildNode | string | null)[] }
const finish = (node: BuildNode): HTMLElementNode => ({
tag: node.tag,
attrs: node.attrs,
children: node.children.flatMap((child): HTMLNode[] =>
child === COMMENT ? [] : typeof child === "string" ? (child ? [child] : []) : [finish(child)]
),
})
export const parseHTML = (input: string): HTMLNode[] => {
const src = input.replace(/\r\n?/g, "\n")
const root: BuildNode = { tag: "#root", attrs: {}, children: [] }
const stack: BuildNode[] = [root]
const current = () => stack[stack.length - 1]
// inside <svg> or <math> a `/>` closes the element - in HTML it is ignored
let foreignDepth = 0
const appendText = (text: string) => {
if (!text) return
const children = current().children
const last = children[children.length - 1]
if (typeof last === "string") children[children.length - 1] = last + text
else children.push(text)
}
const close = (tag: string) => {
for (let depth = stack.length - 1; depth > 0; depth--) {
if (stack[depth].tag !== tag) continue
for (let popped = stack.length - 1; popped >= depth; popped--) {
if (stack[popped].tag === "svg" || stack[popped].tag === "math") foreignDepth--
}
stack.length = depth
return
}
}
let at = 0
while (at < src.length) {
const lt = src.indexOf("<", at)
if (lt === -1) {
appendText(decode(src.slice(at), false))
break
}
appendText(decode(src.slice(at, lt), false))
at = lt
// a comment - `<!-->` and `<!--->` are complete, empty ones
if (src.startsWith("<!--", at)) {
current().children.push(COMMENT)
if (src.startsWith(">", at + 4)) at += 5
else if (src.startsWith("->", at + 4)) at += 6
else {
const end = src.indexOf("-->", at + 4)
at = end === -1 ? src.length : end + 3
}
continue
}
// a doctype leaves no node at all - so the texts around it are one text -
// and a CDATA section or a processing instruction is a (bogus) comment
if (src[at + 1] === "!" || src[at + 1] === "?") {
const end = src.indexOf(">", at)
Iif (!/^<!doctype/i.test(src.slice(at, at + 9))) current().children.push(COMMENT)
at = end === -1 ? src.length : end + 1
continue
}
if (src[at + 1] === "/") {
END_TAG_RE.lastIndex = at
const end = END_TAG_RE.exec(src)
if (end) {
if (!DOCUMENT_TAGS.has(end[1].toLowerCase())) close(end[1].toLowerCase())
at = END_TAG_RE.lastIndex
} else E{
// `</>` is dropped, and `</ anything>` is a bogus comment
const gt = src.indexOf(">", at)
at = gt === -1 ? src.length : gt + 1
}
continue
}
if (!/[A-Za-z]/.test(src[at + 1] ?? "")) {
appendText("<")
at++
continue
}
// a start tag
const tag = readAt(TAG_NAME_RE, src, at + 1).toLowerCase()
const ignored = DOCUMENT_TAGS.has(tag)
let cursor = at + 1 + tag.length
const attrs: Record<string, string> = {}
let selfClosing = false
let complete = false
while (cursor < src.length) {
cursor += readAt(WHITESPACE_RE, src, cursor).length
const char = src[cursor]
if (char === ">") { cursor++; complete = true; break }
if (char === "/") {
Eif (src[cursor + 1] === ">") { selfClosing = true; cursor += 2; complete = true; break }
cursor++
continue
}
Iif (char === undefined) break
const name = readAt(ATTRIBUTE_NAME_RE, src, cursor)
cursor += name.length
let value = ""
const afterName = cursor + readAt(WHITESPACE_RE, src, cursor).length
if (src[afterName] === "=") {
cursor = afterName + 1
cursor += readAt(WHITESPACE_RE, src, cursor).length
const quote = src[cursor]
if (quote === "\"" || quote === "'") {
const close = src.indexOf(quote, cursor + 1)
const end = close === -1 ? src.length : close
value = src.slice(cursor + 1, end)
cursor = end + 1
} else {
value = readAt(UNQUOTED_VALUE_RE, src, cursor)
cursor += value.length
}
value = decode(value, true)
}
const key = name.toLowerCase()
if (!(key in attrs)) attrs[key] = value
}
// a tag cut off by the end of the input is dropped, as the parser drops it
Iif (!complete) break
at = cursor
if (ignored) continue
const node: BuildNode = { tag, attrs, children: [] }
current().children.push(node)
if (VOID_TAGS.has(tag)) continue
if (selfClosing && foreignDepth > 0) continue
if (foreignDepth === 0 && (RAW_TEXT_TAGS.has(tag) || ESCAPABLE_RAW_TEXT_TAGS.has(tag))) {
const end = rawTextEnd(src, at, tag)
const text = src.slice(at, end)
node.children.push(ESCAPABLE_RAW_TEXT_TAGS.has(tag) ? decode(text, false) : text)
const gt = src.indexOf(">", end)
at = end === src.length || gt === -1 ? src.length : gt + 1
continue
}
if (tag === "svg" || tag === "math") foreignDepth++
stack.push(node)
}
return finish(root).children
}
|