postcss
Version:
Tool for transforming styles with JS plugins
72 lines (63 loc) • 2.17 kB
JavaScript
let Stringifier = require('./stringifier')
const STYLE_TAG = /(<)(\/?style\b)/gi
const COMMENT_OPEN = /(<)(!--)/g
function escapeHTMLInCSS(str) {
if (!str.includes('<')) return str
return str.replace(STYLE_TAG, '\\3c $2').replace(COMMENT_OPEN, '\\3c $2')
}
class SafeStringifier extends Stringifier {
escapeHTML(node, print) {
let builder = this.builder
let chunks, css
this.builder = (str, child, type) => {
if (chunks) {
css += str
chunks.push(str, child, type)
} else if (str.includes('<')) {
// Dangerous sequences start with `<`, so everything before
// the first `<` is safe to pass through. The rest is buffered
// to escape it as a whole.
css = str
chunks = [str, child, type]
} else {
builder(str, child, type)
}
}
print()
this.builder = builder
if (!chunks) return
let escaped = escapeHTMLInCSS(css)
if (escaped === css) {
// Keep the chunks to get a precise source map
for (let i = 0; i < chunks.length; i += 3) {
builder(chunks[i], chunks[i + 1], chunks[i + 2])
}
} else {
// Escaping happens only on a hack attempt, source map is not important.
// Map only the node start: the end mapping expects the node’s last
// char at the end of the string, which is wrong for Root raws.after.
builder(escaped, node, 'start')
}
}
root(node) {
if (node.parent && node.parent.type === 'document') {
// Document raws around the Root hold markup and are kept as is since CSS in Document can be marked with <style>
this.escapeHTML(node, () => this.body(node))
if (node.raws.after) this.builder(node.raws.after)
} else {
this.escapeHTML(node, () => super.root(node))
}
}
}
function stringify(node, builder) {
let str = new SafeStringifier(builder)
// Root escapes its own output and keeps markup in Document raws
if (node.type === 'root' || node.type === 'document') {
str.stringify(node)
} else {
str.escapeHTML(node, () => str.stringify(node))
}
}
module.exports = stringify
stringify.default = stringify