bot_dev1
85df71fc28
- 新增OperationDashboard.vue全屏大屏组件 - 添加6个核心KPI指标卡片: 进水总量/出水总量/产销差率/营收额/平均水质/报警次数 - 实现6个ECharts图表: 供水趋势/水质分布/报警统计/管网空间/设备状态/营收分析 - 创建静态HTML版本operation-dashboard.html，使用CDN加载Vue3/ECharts/Element Plus - 更新路由配置，添加/operation路径支持 - 修复nextTick导入问题，优化build脚本 🚧 开发者: bot_dev1 📝 任务: #38 [BI] 运营仪表盘 + 供水专题大屏
2662 lines
91 KiB
TypeScript
2662 lines
91 KiB
TypeScript
/* eslint-disable @typescript-eslint/indent */
|
|
|
|
import type { Config, UseProfilesConfig } from './config';
|
|
import type { DOMPurify, HooksMap, HookFunction, WindowLike } from './types';
|
|
import * as TAGS from './tags.js';
|
|
import * as ATTRS from './attrs.js';
|
|
import * as EXPRESSIONS from './regexp.js';
|
|
import {
|
|
addToSet,
|
|
clone,
|
|
entries,
|
|
freeze,
|
|
seal,
|
|
arrayForEach,
|
|
arrayIsArray,
|
|
arrayLastIndexOf,
|
|
arrayPop,
|
|
arrayPush,
|
|
arraySplice,
|
|
stringMatch,
|
|
stringReplace,
|
|
stringToLowerCase,
|
|
stringToString,
|
|
stringIndexOf,
|
|
stringTrim,
|
|
regExpTest,
|
|
isRegex,
|
|
typeErrorCreate,
|
|
lookupGetter,
|
|
create,
|
|
objectHasOwnProperty,
|
|
stringifyValue,
|
|
} from './utils.js';
|
|
|
|
export type { Config } from './config';
|
|
|
|
export type {
|
|
DOMPurify,
|
|
RemovedElement,
|
|
RemovedAttribute,
|
|
HookName,
|
|
NodeHook,
|
|
ElementHook,
|
|
DocumentFragmentHook,
|
|
UponSanitizeElementHook,
|
|
UponSanitizeAttributeHook,
|
|
UponSanitizeElementHookEvent,
|
|
UponSanitizeAttributeHookEvent,
|
|
WindowLike,
|
|
} from './types';
|
|
|
|
declare const VERSION: string;
|
|
|
|
// https://developer.mozilla.org/en-US/docs/Web/API/Node/nodeType
|
|
const NODE_TYPE = {
|
|
element: 1,
|
|
attribute: 2,
|
|
text: 3,
|
|
cdataSection: 4,
|
|
entityReference: 5, // Deprecated
|
|
entityNode: 6, // Deprecated
|
|
processingInstruction: 7,
|
|
comment: 8,
|
|
document: 9,
|
|
documentType: 10,
|
|
documentFragment: 11,
|
|
notation: 12, // Deprecated
|
|
};
|
|
|
|
const getGlobal = function (): WindowLike {
|
|
return typeof window === 'undefined' ? null : window;
|
|
};
|
|
|
|
/**
|
|
* Creates a no-op policy for internal use only.
|
|
* Don't export this function outside this module!
|
|
* @param trustedTypes The policy factory.
|
|
* @param purifyHostElement The Script element used to load DOMPurify (to determine policy name suffix).
|
|
* @return The policy created (or null, if Trusted Types
|
|
* are not supported or creating the policy failed).
|
|
*/
|
|
const _createTrustedTypesPolicy = function (
|
|
trustedTypes: TrustedTypePolicyFactory,
|
|
purifyHostElement: HTMLScriptElement
|
|
) {
|
|
if (
|
|
typeof trustedTypes !== 'object' ||
|
|
typeof trustedTypes.createPolicy !== 'function'
|
|
) {
|
|
return null;
|
|
}
|
|
|
|
// Allow the callers to control the unique policy name
|
|
// by adding a data-tt-policy-suffix to the script element with the DOMPurify.
|
|
// Policy creation with duplicate names throws in Trusted Types.
|
|
let suffix = null;
|
|
const ATTR_NAME = 'data-tt-policy-suffix';
|
|
if (purifyHostElement && purifyHostElement.hasAttribute(ATTR_NAME)) {
|
|
suffix = purifyHostElement.getAttribute(ATTR_NAME);
|
|
}
|
|
|
|
const policyName = 'dompurify' + (suffix ? '#' + suffix : '');
|
|
|
|
try {
|
|
return trustedTypes.createPolicy(policyName, {
|
|
createHTML(html) {
|
|
return html;
|
|
},
|
|
createScriptURL(scriptUrl) {
|
|
return scriptUrl;
|
|
},
|
|
});
|
|
} catch (_) {
|
|
// Policy creation failed (most likely another DOMPurify script has
|
|
// already run). Skip creating the policy, as this will only cause errors
|
|
// if TT are enforced.
|
|
console.warn(
|
|
'TrustedTypes policy ' + policyName + ' could not be created.'
|
|
);
|
|
return null;
|
|
}
|
|
};
|
|
|
|
const _createHooksMap = function (): HooksMap {
|
|
return {
|
|
afterSanitizeAttributes: [],
|
|
afterSanitizeElements: [],
|
|
afterSanitizeShadowDOM: [],
|
|
beforeSanitizeAttributes: [],
|
|
beforeSanitizeElements: [],
|
|
beforeSanitizeShadowDOM: [],
|
|
uponSanitizeAttribute: [],
|
|
uponSanitizeElement: [],
|
|
uponSanitizeShadowNode: [],
|
|
};
|
|
};
|
|
|
|
/**
|
|
* Resolve a set-valued configuration option: a fresh set built from
|
|
* cfg[key] when it is an own array property (seeded with a clone of
|
|
* options.base when given, case-normalized via options.transform),
|
|
* the fallback set otherwise.
|
|
*
|
|
* @param cfg the cloned, prototype-free configuration object
|
|
* @param key the configuration property to read
|
|
* @param fallback the set to use when the option is absent or not an array
|
|
* @param options transform and optional base set to merge into
|
|
* @returns the resolved set
|
|
*/
|
|
const _resolveSetOption = function (
|
|
cfg: Config,
|
|
key: keyof Config,
|
|
fallback: Record<string, boolean>,
|
|
options: {
|
|
transform: Parameters<typeof addToSet>[2];
|
|
base?: Record<string, boolean>;
|
|
}
|
|
): Record<string, boolean> {
|
|
return objectHasOwnProperty(cfg, key) && arrayIsArray(cfg[key])
|
|
? addToSet(
|
|
options.base ? clone(options.base) : {},
|
|
cfg[key] as readonly unknown[],
|
|
options.transform
|
|
)
|
|
: fallback;
|
|
};
|
|
|
|
function createDOMPurify(window: WindowLike = getGlobal()): DOMPurify {
|
|
const DOMPurify: DOMPurify = (root: WindowLike) => createDOMPurify(root);
|
|
|
|
DOMPurify.version = VERSION;
|
|
|
|
DOMPurify.removed = [];
|
|
|
|
if (
|
|
!window ||
|
|
!window.document ||
|
|
window.document.nodeType !== NODE_TYPE.document ||
|
|
!window.Element
|
|
) {
|
|
// Not running in a browser, provide a factory function
|
|
// so that you can pass your own Window
|
|
DOMPurify.isSupported = false;
|
|
|
|
return DOMPurify;
|
|
}
|
|
|
|
let { document } = window;
|
|
|
|
const originalDocument = document;
|
|
const currentScript: HTMLScriptElement =
|
|
originalDocument.currentScript as HTMLScriptElement;
|
|
const {
|
|
DocumentFragment,
|
|
HTMLTemplateElement,
|
|
Node,
|
|
Element,
|
|
NodeFilter,
|
|
NamedNodeMap = window.NamedNodeMap || (window as any).MozNamedAttrMap,
|
|
HTMLFormElement,
|
|
DOMParser,
|
|
trustedTypes,
|
|
} = window;
|
|
|
|
const ElementPrototype = Element.prototype;
|
|
|
|
const cloneNode = lookupGetter(ElementPrototype, 'cloneNode');
|
|
const remove = lookupGetter(ElementPrototype, 'remove');
|
|
const getNextSibling = lookupGetter(ElementPrototype, 'nextSibling');
|
|
const getChildNodes = lookupGetter(ElementPrototype, 'childNodes');
|
|
const getParentNode = lookupGetter(ElementPrototype, 'parentNode');
|
|
const getShadowRoot = lookupGetter(ElementPrototype, 'shadowRoot');
|
|
const getAttributes = lookupGetter(ElementPrototype, 'attributes');
|
|
const getNodeType =
|
|
Node && Node.prototype ? lookupGetter(Node.prototype, 'nodeType') : null;
|
|
const getNodeName =
|
|
Node && Node.prototype ? lookupGetter(Node.prototype, 'nodeName') : null;
|
|
|
|
// As per issue #47, the web-components registry is inherited by a
|
|
// new document created via createHTMLDocument. As per the spec
|
|
// (http://w3c.github.io/webcomponents/spec/custom/#creating-and-passing-registries)
|
|
// a new empty registry is used when creating a template contents owner
|
|
// document, so we use that as our parent document to ensure nothing
|
|
// is inherited.
|
|
if (typeof HTMLTemplateElement === 'function') {
|
|
const template = document.createElement('template');
|
|
if (template.content && template.content.ownerDocument) {
|
|
document = template.content.ownerDocument;
|
|
}
|
|
}
|
|
|
|
let trustedTypesPolicy;
|
|
let emptyHTML = '';
|
|
|
|
// The instance's own internal Trusted Types policy. Unlike a caller-supplied
|
|
// `TRUSTED_TYPES_POLICY`, this is created at most once — Trusted Types throws
|
|
// on duplicate policy names — and is the only policy allowed to persist
|
|
// across configurations and survive `clearConfig()`.
|
|
let defaultTrustedTypesPolicy;
|
|
let defaultTrustedTypesPolicyResolved = false;
|
|
|
|
// Tracks whether we are already inside a call to the configured Trusted Types
|
|
// policy (`createHTML` or `createScriptURL`). If a supplied policy callback
|
|
// itself calls `DOMPurify.sanitize` (the cause of #1422), `sanitize` would
|
|
// re-enter the policy and recurse until the stack overflows. We detect that
|
|
// re-entry and throw a clear, actionable error instead. The guard is shared
|
|
// across both callbacks, because either one re-entering `sanitize` triggers
|
|
// the same unbounded recursion.
|
|
let IN_TRUSTED_TYPES_POLICY = 0;
|
|
const _assertNotInTrustedTypesPolicy = function (): void {
|
|
if (IN_TRUSTED_TYPES_POLICY > 0) {
|
|
throw typeErrorCreate(
|
|
'A configured TRUSTED_TYPES_POLICY callback (createHTML or ' +
|
|
'createScriptURL) must not call DOMPurify.sanitize, as that causes ' +
|
|
'infinite recursion. Do not pass a policy whose callbacks wrap ' +
|
|
'DOMPurify as TRUSTED_TYPES_POLICY; see the "DOMPurify and Trusted ' +
|
|
'Types" section of the README.'
|
|
);
|
|
}
|
|
};
|
|
|
|
const _createTrustedHTML = function (html: string): string {
|
|
_assertNotInTrustedTypesPolicy();
|
|
|
|
IN_TRUSTED_TYPES_POLICY++;
|
|
try {
|
|
return trustedTypesPolicy.createHTML(html);
|
|
} finally {
|
|
IN_TRUSTED_TYPES_POLICY--;
|
|
}
|
|
};
|
|
|
|
const _createTrustedScriptURL = function (scriptUrl: string): string {
|
|
_assertNotInTrustedTypesPolicy();
|
|
|
|
IN_TRUSTED_TYPES_POLICY++;
|
|
try {
|
|
return trustedTypesPolicy.createScriptURL(scriptUrl);
|
|
} finally {
|
|
IN_TRUSTED_TYPES_POLICY--;
|
|
}
|
|
};
|
|
|
|
// Lazily resolve (and cache) the instance's internal default policy.
|
|
// Resolution is attempted at most once: a successful `createPolicy` cannot be
|
|
// repeated (Trusted Types throws on duplicate names), and a failed or
|
|
// unsupported attempt must not be retried on every parse.
|
|
const _getDefaultTrustedTypesPolicy = function () {
|
|
if (!defaultTrustedTypesPolicyResolved) {
|
|
defaultTrustedTypesPolicy = _createTrustedTypesPolicy(
|
|
trustedTypes,
|
|
currentScript
|
|
);
|
|
defaultTrustedTypesPolicyResolved = true;
|
|
}
|
|
|
|
return defaultTrustedTypesPolicy;
|
|
};
|
|
|
|
const {
|
|
implementation,
|
|
createNodeIterator,
|
|
createDocumentFragment,
|
|
getElementsByTagName,
|
|
} = document;
|
|
const { importNode } = originalDocument;
|
|
|
|
let hooks = _createHooksMap();
|
|
|
|
/**
|
|
* Expose whether this browser supports running the full DOMPurify.
|
|
*/
|
|
DOMPurify.isSupported =
|
|
typeof entries === 'function' &&
|
|
typeof getParentNode === 'function' &&
|
|
implementation &&
|
|
implementation.createHTMLDocument !== undefined;
|
|
|
|
const {
|
|
MUSTACHE_EXPR,
|
|
ERB_EXPR,
|
|
TMPLIT_EXPR,
|
|
DATA_ATTR,
|
|
ARIA_ATTR,
|
|
IS_SCRIPT_OR_DATA,
|
|
ATTR_WHITESPACE,
|
|
CUSTOM_ELEMENT,
|
|
} = EXPRESSIONS;
|
|
|
|
let { IS_ALLOWED_URI } = EXPRESSIONS;
|
|
|
|
/**
|
|
* We consider the elements and attributes below to be safe. Ideally
|
|
* don't add any new ones but feel free to remove unwanted ones.
|
|
*/
|
|
|
|
/* allowed element names */
|
|
let ALLOWED_TAGS = null;
|
|
const DEFAULT_ALLOWED_TAGS = addToSet({}, [
|
|
...TAGS.html,
|
|
...TAGS.svg,
|
|
...TAGS.svgFilters,
|
|
...TAGS.mathMl,
|
|
...TAGS.text,
|
|
]);
|
|
|
|
/* Allowed attribute names */
|
|
let ALLOWED_ATTR = null;
|
|
const DEFAULT_ALLOWED_ATTR = addToSet({}, [
|
|
...ATTRS.html,
|
|
...ATTRS.svg,
|
|
...ATTRS.mathMl,
|
|
...ATTRS.xml,
|
|
]);
|
|
|
|
/*
|
|
* Configure how DOMPurify should handle custom elements and their attributes as well as customized built-in elements.
|
|
* @property {RegExp|Function|null} tagNameCheck one of [null, regexPattern, predicate]. Default: `null` (disallow any custom elements)
|
|
* @property {RegExp|Function|null} attributeNameCheck one of [null, regexPattern, predicate]. Default: `null` (disallow any attributes not on the allow list)
|
|
* @property {boolean} allowCustomizedBuiltInElements allow custom elements derived from built-ins if they pass CUSTOM_ELEMENT_HANDLING.tagNameCheck. Default: `false`.
|
|
*/
|
|
let CUSTOM_ELEMENT_HANDLING = Object.seal(
|
|
create(null, {
|
|
tagNameCheck: {
|
|
writable: true,
|
|
configurable: false,
|
|
enumerable: true,
|
|
value: null,
|
|
},
|
|
attributeNameCheck: {
|
|
writable: true,
|
|
configurable: false,
|
|
enumerable: true,
|
|
value: null,
|
|
},
|
|
allowCustomizedBuiltInElements: {
|
|
writable: true,
|
|
configurable: false,
|
|
enumerable: true,
|
|
value: false,
|
|
},
|
|
})
|
|
);
|
|
|
|
/* Explicitly forbidden tags (overrides ALLOWED_TAGS/ADD_TAGS) */
|
|
let FORBID_TAGS = null;
|
|
|
|
/* Explicitly forbidden attributes (overrides ALLOWED_ATTR/ADD_ATTR) */
|
|
let FORBID_ATTR = null;
|
|
|
|
/* Config object to store ADD_TAGS/ADD_ATTR functions (when used as functions) */
|
|
const EXTRA_ELEMENT_HANDLING = Object.seal(
|
|
create(null, {
|
|
tagCheck: {
|
|
writable: true,
|
|
configurable: false,
|
|
enumerable: true,
|
|
value: null,
|
|
},
|
|
attributeCheck: {
|
|
writable: true,
|
|
configurable: false,
|
|
enumerable: true,
|
|
value: null,
|
|
},
|
|
})
|
|
);
|
|
|
|
/* Decide if ARIA attributes are okay */
|
|
let ALLOW_ARIA_ATTR = true;
|
|
|
|
/* Decide if custom data attributes are okay */
|
|
let ALLOW_DATA_ATTR = true;
|
|
|
|
/* Decide if unknown protocols are okay */
|
|
let ALLOW_UNKNOWN_PROTOCOLS = false;
|
|
|
|
/* Decide if self-closing tags in attributes are allowed.
|
|
* Usually removed due to a mXSS issue in jQuery 3.0 */
|
|
let ALLOW_SELF_CLOSE_IN_ATTR = true;
|
|
|
|
/* Output should be safe for common template engines.
|
|
* This means, DOMPurify removes data attributes, mustaches and ERB
|
|
*/
|
|
let SAFE_FOR_TEMPLATES = false;
|
|
|
|
/* Output should be safe even for XML used within HTML and alike.
|
|
* This means, DOMPurify removes comments when containing risky content.
|
|
*/
|
|
let SAFE_FOR_XML = true;
|
|
|
|
/* Decide if document with <html>... should be returned */
|
|
let WHOLE_DOCUMENT = false;
|
|
|
|
/* Track whether config is already set on this instance of DOMPurify. */
|
|
let SET_CONFIG = false;
|
|
|
|
/* Decide if all elements (e.g. style, script) must be children of
|
|
* document.body. By default, browsers might move them to document.head */
|
|
let FORCE_BODY = false;
|
|
|
|
/* Decide if a DOM `HTMLBodyElement` should be returned, instead of a html
|
|
* string (or a TrustedHTML object if Trusted Types are supported).
|
|
* If `WHOLE_DOCUMENT` is enabled a `HTMLHtmlElement` will be returned instead
|
|
*/
|
|
let RETURN_DOM = false;
|
|
|
|
/* Decide if a DOM `DocumentFragment` should be returned, instead of a html
|
|
* string (or a TrustedHTML object if Trusted Types are supported) */
|
|
let RETURN_DOM_FRAGMENT = false;
|
|
|
|
/* Try to return a Trusted Type object instead of a string, return a string in
|
|
* case Trusted Types are not supported */
|
|
let RETURN_TRUSTED_TYPE = false;
|
|
|
|
/* Output should be free from DOM clobbering attacks?
|
|
* This sanitizes markups named with colliding, clobberable built-in DOM APIs.
|
|
*/
|
|
let SANITIZE_DOM = true;
|
|
|
|
/* Achieve full DOM Clobbering protection by isolating the namespace of named
|
|
* properties and JS variables, mitigating attacks that abuse the HTML/DOM spec rules.
|
|
*
|
|
* HTML/DOM spec rules that enable DOM Clobbering:
|
|
* - Named Access on Window (§7.3.3)
|
|
* - DOM Tree Accessors (§3.1.5)
|
|
* - Form Element Parent-Child Relations (§4.10.3)
|
|
* - Iframe srcdoc / Nested WindowProxies (§4.8.5)
|
|
* - HTMLCollection (§4.2.10.2)
|
|
*
|
|
* Namespace isolation is implemented by prefixing `id` and `name` attributes
|
|
* with a constant string, i.e., `user-content-`
|
|
*/
|
|
let SANITIZE_NAMED_PROPS = false;
|
|
const SANITIZE_NAMED_PROPS_PREFIX = 'user-content-';
|
|
|
|
/* Keep element content when removing element? */
|
|
let KEEP_CONTENT = true;
|
|
|
|
/* If a `Node` is passed to sanitize(), then performs sanitization in-place instead
|
|
* of importing it into a new Document and returning a sanitized copy */
|
|
let IN_PLACE = false;
|
|
|
|
/* Allow usage of profiles like html, svg and mathMl */
|
|
let USE_PROFILES: UseProfilesConfig | false = {};
|
|
|
|
/* Tags to ignore content of when KEEP_CONTENT is true */
|
|
let FORBID_CONTENTS = null;
|
|
const DEFAULT_FORBID_CONTENTS = addToSet({}, [
|
|
'annotation-xml',
|
|
'audio',
|
|
'colgroup',
|
|
'desc',
|
|
'foreignobject',
|
|
'head',
|
|
'iframe',
|
|
'math',
|
|
'mi',
|
|
'mn',
|
|
'mo',
|
|
'ms',
|
|
'mtext',
|
|
'noembed',
|
|
'noframes',
|
|
'noscript',
|
|
'plaintext',
|
|
'script',
|
|
// <selectedcontent> mirrors the selected <option>'s subtree, cloned by
|
|
// the UA (customizable <select>) — including any on* handlers — and the
|
|
// engine re-mirrors synchronously whenever a removal changes which
|
|
// option/selectedcontent is current, even inside DOMPurify's inert
|
|
// DOMParser document. Hoisting its children on removal re-inserts a fresh
|
|
// mirror target ahead of the walk, which the engine refills, looping
|
|
// forever (DoS) and amplifying output. Dropping its content on removal
|
|
// (rather than hoisting) breaks that cascade; the content is a duplicate
|
|
// of the option, which is sanitized on its own. See campaign-3 F1/F6.
|
|
'selectedcontent',
|
|
'style',
|
|
'svg',
|
|
'template',
|
|
'thead',
|
|
'title',
|
|
'video',
|
|
'xmp',
|
|
]);
|
|
|
|
/* Tags that are safe for data: URIs */
|
|
let DATA_URI_TAGS = null;
|
|
const DEFAULT_DATA_URI_TAGS = addToSet({}, [
|
|
'audio',
|
|
'video',
|
|
'img',
|
|
'source',
|
|
'image',
|
|
'track',
|
|
]);
|
|
|
|
/* Attributes safe for values like "javascript:" */
|
|
let URI_SAFE_ATTRIBUTES = null;
|
|
const DEFAULT_URI_SAFE_ATTRIBUTES = addToSet({}, [
|
|
'alt',
|
|
'class',
|
|
'for',
|
|
'id',
|
|
'label',
|
|
'name',
|
|
'pattern',
|
|
'placeholder',
|
|
'role',
|
|
'summary',
|
|
'title',
|
|
'value',
|
|
'style',
|
|
'xmlns',
|
|
]);
|
|
|
|
const MATHML_NAMESPACE = 'http://www.w3.org/1998/Math/MathML';
|
|
const SVG_NAMESPACE = 'http://www.w3.org/2000/svg';
|
|
const HTML_NAMESPACE = 'http://www.w3.org/1999/xhtml';
|
|
/* Document namespace */
|
|
let NAMESPACE = HTML_NAMESPACE;
|
|
let IS_EMPTY_INPUT = false;
|
|
|
|
/* Allowed XHTML+XML namespaces */
|
|
let ALLOWED_NAMESPACES = null;
|
|
const DEFAULT_ALLOWED_NAMESPACES = addToSet(
|
|
{},
|
|
[MATHML_NAMESPACE, SVG_NAMESPACE, HTML_NAMESPACE],
|
|
stringToString
|
|
);
|
|
|
|
const DEFAULT_MATHML_TEXT_INTEGRATION_POINTS = freeze([
|
|
'mi',
|
|
'mo',
|
|
'mn',
|
|
'ms',
|
|
'mtext',
|
|
]);
|
|
let MATHML_TEXT_INTEGRATION_POINTS = addToSet(
|
|
{},
|
|
DEFAULT_MATHML_TEXT_INTEGRATION_POINTS
|
|
);
|
|
|
|
const DEFAULT_HTML_INTEGRATION_POINTS = freeze(['annotation-xml']);
|
|
let HTML_INTEGRATION_POINTS = addToSet({}, DEFAULT_HTML_INTEGRATION_POINTS);
|
|
|
|
// Certain elements are allowed in both SVG and HTML
|
|
// namespace. We need to specify them explicitly
|
|
// so that they don't get erroneously deleted from
|
|
// HTML namespace.
|
|
const COMMON_SVG_AND_HTML_ELEMENTS = addToSet({}, [
|
|
'title',
|
|
'style',
|
|
'font',
|
|
'a',
|
|
'script',
|
|
]);
|
|
|
|
/* Parsing of strict XHTML documents */
|
|
let PARSER_MEDIA_TYPE: null | DOMParserSupportedType = null;
|
|
const SUPPORTED_PARSER_MEDIA_TYPES = ['application/xhtml+xml', 'text/html'];
|
|
const DEFAULT_PARSER_MEDIA_TYPE = 'text/html';
|
|
let transformCaseFunc: null | Parameters<typeof addToSet>[2] = null;
|
|
|
|
/* Keep a reference to config to pass to hooks */
|
|
let CONFIG: Config | null = null;
|
|
|
|
/* Ideally, do not touch anything below this line */
|
|
/* ______________________________________________ */
|
|
|
|
const formElement = document.createElement('form');
|
|
|
|
const isRegexOrFunction = function (
|
|
testValue: unknown
|
|
): testValue is Function | RegExp {
|
|
return testValue instanceof RegExp || testValue instanceof Function;
|
|
};
|
|
|
|
/**
|
|
* _parseConfig
|
|
*
|
|
* @param cfg optional config literal
|
|
*/
|
|
// eslint-disable-next-line complexity
|
|
const _parseConfig = function (cfg: Config = {}): void {
|
|
if (CONFIG && CONFIG === cfg) {
|
|
return;
|
|
}
|
|
|
|
/* Shield configuration object from tampering */
|
|
if (!cfg || typeof cfg !== 'object') {
|
|
cfg = {};
|
|
}
|
|
|
|
/* Shield configuration object from prototype pollution */
|
|
cfg = clone(cfg);
|
|
|
|
PARSER_MEDIA_TYPE =
|
|
// eslint-disable-next-line unicorn/prefer-includes
|
|
SUPPORTED_PARSER_MEDIA_TYPES.indexOf(cfg.PARSER_MEDIA_TYPE) === -1
|
|
? DEFAULT_PARSER_MEDIA_TYPE
|
|
: cfg.PARSER_MEDIA_TYPE;
|
|
|
|
// HTML tags and attributes are not case-sensitive, converting to lowercase. Keeping XHTML as is.
|
|
transformCaseFunc =
|
|
PARSER_MEDIA_TYPE === 'application/xhtml+xml'
|
|
? stringToString
|
|
: stringToLowerCase;
|
|
|
|
/* Set configuration parameters */
|
|
ALLOWED_TAGS = _resolveSetOption(
|
|
cfg,
|
|
'ALLOWED_TAGS',
|
|
DEFAULT_ALLOWED_TAGS,
|
|
{ transform: transformCaseFunc }
|
|
);
|
|
ALLOWED_ATTR = _resolveSetOption(
|
|
cfg,
|
|
'ALLOWED_ATTR',
|
|
DEFAULT_ALLOWED_ATTR,
|
|
{ transform: transformCaseFunc }
|
|
);
|
|
ALLOWED_NAMESPACES = _resolveSetOption(
|
|
cfg,
|
|
'ALLOWED_NAMESPACES',
|
|
DEFAULT_ALLOWED_NAMESPACES,
|
|
{ transform: stringToString }
|
|
);
|
|
URI_SAFE_ATTRIBUTES = _resolveSetOption(
|
|
cfg,
|
|
'ADD_URI_SAFE_ATTR',
|
|
DEFAULT_URI_SAFE_ATTRIBUTES,
|
|
{ transform: transformCaseFunc, base: DEFAULT_URI_SAFE_ATTRIBUTES }
|
|
);
|
|
DATA_URI_TAGS = _resolveSetOption(
|
|
cfg,
|
|
'ADD_DATA_URI_TAGS',
|
|
DEFAULT_DATA_URI_TAGS,
|
|
{ transform: transformCaseFunc, base: DEFAULT_DATA_URI_TAGS }
|
|
);
|
|
FORBID_CONTENTS = _resolveSetOption(
|
|
cfg,
|
|
'FORBID_CONTENTS',
|
|
DEFAULT_FORBID_CONTENTS,
|
|
{ transform: transformCaseFunc }
|
|
);
|
|
FORBID_TAGS = _resolveSetOption(cfg, 'FORBID_TAGS', clone({}), {
|
|
transform: transformCaseFunc,
|
|
});
|
|
FORBID_ATTR = _resolveSetOption(cfg, 'FORBID_ATTR', clone({}), {
|
|
transform: transformCaseFunc,
|
|
});
|
|
USE_PROFILES = objectHasOwnProperty(cfg, 'USE_PROFILES')
|
|
? cfg.USE_PROFILES && typeof cfg.USE_PROFILES === 'object'
|
|
? clone(cfg.USE_PROFILES)
|
|
: cfg.USE_PROFILES
|
|
: false;
|
|
|
|
ALLOW_ARIA_ATTR = cfg.ALLOW_ARIA_ATTR !== false; // Default true
|
|
ALLOW_DATA_ATTR = cfg.ALLOW_DATA_ATTR !== false; // Default true
|
|
ALLOW_UNKNOWN_PROTOCOLS = cfg.ALLOW_UNKNOWN_PROTOCOLS || false; // Default false
|
|
ALLOW_SELF_CLOSE_IN_ATTR = cfg.ALLOW_SELF_CLOSE_IN_ATTR !== false; // Default true
|
|
SAFE_FOR_TEMPLATES = cfg.SAFE_FOR_TEMPLATES || false; // Default false
|
|
SAFE_FOR_XML = cfg.SAFE_FOR_XML !== false; // Default true
|
|
WHOLE_DOCUMENT = cfg.WHOLE_DOCUMENT || false; // Default false
|
|
RETURN_DOM = cfg.RETURN_DOM || false; // Default false
|
|
RETURN_DOM_FRAGMENT = cfg.RETURN_DOM_FRAGMENT || false; // Default false
|
|
RETURN_TRUSTED_TYPE = cfg.RETURN_TRUSTED_TYPE || false; // Default false
|
|
FORCE_BODY = cfg.FORCE_BODY || false; // Default false
|
|
SANITIZE_DOM = cfg.SANITIZE_DOM !== false; // Default true
|
|
SANITIZE_NAMED_PROPS = cfg.SANITIZE_NAMED_PROPS || false; // Default false
|
|
KEEP_CONTENT = cfg.KEEP_CONTENT !== false; // Default true
|
|
IN_PLACE = cfg.IN_PLACE || false; // Default false
|
|
IS_ALLOWED_URI = isRegex(cfg.ALLOWED_URI_REGEXP)
|
|
? cfg.ALLOWED_URI_REGEXP
|
|
: EXPRESSIONS.IS_ALLOWED_URI; // Default regexp
|
|
|
|
NAMESPACE =
|
|
typeof cfg.NAMESPACE === 'string' ? cfg.NAMESPACE : HTML_NAMESPACE; // Default HTML namespace
|
|
|
|
MATHML_TEXT_INTEGRATION_POINTS =
|
|
objectHasOwnProperty(cfg, 'MATHML_TEXT_INTEGRATION_POINTS') &&
|
|
cfg.MATHML_TEXT_INTEGRATION_POINTS &&
|
|
typeof cfg.MATHML_TEXT_INTEGRATION_POINTS === 'object'
|
|
? clone(cfg.MATHML_TEXT_INTEGRATION_POINTS)
|
|
: addToSet({}, DEFAULT_MATHML_TEXT_INTEGRATION_POINTS); // Default built-in map
|
|
|
|
HTML_INTEGRATION_POINTS =
|
|
objectHasOwnProperty(cfg, 'HTML_INTEGRATION_POINTS') &&
|
|
cfg.HTML_INTEGRATION_POINTS &&
|
|
typeof cfg.HTML_INTEGRATION_POINTS === 'object'
|
|
? clone(cfg.HTML_INTEGRATION_POINTS)
|
|
: addToSet({}, DEFAULT_HTML_INTEGRATION_POINTS); // Default built-in map
|
|
|
|
const customElementHandling =
|
|
objectHasOwnProperty(cfg, 'CUSTOM_ELEMENT_HANDLING') &&
|
|
cfg.CUSTOM_ELEMENT_HANDLING &&
|
|
typeof cfg.CUSTOM_ELEMENT_HANDLING === 'object'
|
|
? clone(cfg.CUSTOM_ELEMENT_HANDLING)
|
|
: create(null);
|
|
|
|
CUSTOM_ELEMENT_HANDLING = create(null);
|
|
|
|
if (
|
|
objectHasOwnProperty(customElementHandling, 'tagNameCheck') &&
|
|
isRegexOrFunction(customElementHandling.tagNameCheck)
|
|
) {
|
|
CUSTOM_ELEMENT_HANDLING.tagNameCheck = customElementHandling.tagNameCheck; // Default undefined
|
|
}
|
|
|
|
if (
|
|
objectHasOwnProperty(customElementHandling, 'attributeNameCheck') &&
|
|
isRegexOrFunction(customElementHandling.attributeNameCheck)
|
|
) {
|
|
CUSTOM_ELEMENT_HANDLING.attributeNameCheck =
|
|
customElementHandling.attributeNameCheck; // Default undefined
|
|
}
|
|
|
|
if (
|
|
objectHasOwnProperty(
|
|
customElementHandling,
|
|
'allowCustomizedBuiltInElements'
|
|
) &&
|
|
typeof customElementHandling.allowCustomizedBuiltInElements === 'boolean'
|
|
) {
|
|
CUSTOM_ELEMENT_HANDLING.allowCustomizedBuiltInElements =
|
|
customElementHandling.allowCustomizedBuiltInElements; // Default undefined
|
|
}
|
|
|
|
seal(CUSTOM_ELEMENT_HANDLING);
|
|
|
|
if (SAFE_FOR_TEMPLATES) {
|
|
ALLOW_DATA_ATTR = false;
|
|
}
|
|
|
|
if (RETURN_DOM_FRAGMENT) {
|
|
RETURN_DOM = true;
|
|
}
|
|
|
|
/* Parse profile info */
|
|
if (USE_PROFILES) {
|
|
ALLOWED_TAGS = addToSet({}, TAGS.text);
|
|
ALLOWED_ATTR = create(null);
|
|
if (USE_PROFILES.html === true) {
|
|
addToSet(ALLOWED_TAGS, TAGS.html);
|
|
addToSet(ALLOWED_ATTR, ATTRS.html);
|
|
}
|
|
|
|
if (USE_PROFILES.svg === true) {
|
|
addToSet(ALLOWED_TAGS, TAGS.svg);
|
|
addToSet(ALLOWED_ATTR, ATTRS.svg);
|
|
addToSet(ALLOWED_ATTR, ATTRS.xml);
|
|
}
|
|
|
|
if (USE_PROFILES.svgFilters === true) {
|
|
addToSet(ALLOWED_TAGS, TAGS.svgFilters);
|
|
addToSet(ALLOWED_ATTR, ATTRS.svg);
|
|
addToSet(ALLOWED_ATTR, ATTRS.xml);
|
|
}
|
|
|
|
if (USE_PROFILES.mathMl === true) {
|
|
addToSet(ALLOWED_TAGS, TAGS.mathMl);
|
|
addToSet(ALLOWED_ATTR, ATTRS.mathMl);
|
|
addToSet(ALLOWED_ATTR, ATTRS.xml);
|
|
}
|
|
}
|
|
|
|
/* Always reset function-based ADD_TAGS / ADD_ATTR checks to prevent
|
|
* leaking across calls when switching from function to array config */
|
|
EXTRA_ELEMENT_HANDLING.tagCheck = null;
|
|
EXTRA_ELEMENT_HANDLING.attributeCheck = null;
|
|
|
|
/* Merge configuration parameters */
|
|
if (objectHasOwnProperty(cfg, 'ADD_TAGS')) {
|
|
if (typeof cfg.ADD_TAGS === 'function') {
|
|
EXTRA_ELEMENT_HANDLING.tagCheck = cfg.ADD_TAGS;
|
|
} else if (arrayIsArray(cfg.ADD_TAGS)) {
|
|
if (ALLOWED_TAGS === DEFAULT_ALLOWED_TAGS) {
|
|
ALLOWED_TAGS = clone(ALLOWED_TAGS);
|
|
}
|
|
|
|
addToSet(ALLOWED_TAGS, cfg.ADD_TAGS, transformCaseFunc);
|
|
}
|
|
}
|
|
|
|
if (objectHasOwnProperty(cfg, 'ADD_ATTR')) {
|
|
if (typeof cfg.ADD_ATTR === 'function') {
|
|
EXTRA_ELEMENT_HANDLING.attributeCheck = cfg.ADD_ATTR;
|
|
} else if (arrayIsArray(cfg.ADD_ATTR)) {
|
|
if (ALLOWED_ATTR === DEFAULT_ALLOWED_ATTR) {
|
|
ALLOWED_ATTR = clone(ALLOWED_ATTR);
|
|
}
|
|
|
|
addToSet(ALLOWED_ATTR, cfg.ADD_ATTR, transformCaseFunc);
|
|
}
|
|
}
|
|
|
|
if (
|
|
objectHasOwnProperty(cfg, 'ADD_URI_SAFE_ATTR') &&
|
|
arrayIsArray(cfg.ADD_URI_SAFE_ATTR)
|
|
) {
|
|
addToSet(URI_SAFE_ATTRIBUTES, cfg.ADD_URI_SAFE_ATTR, transformCaseFunc);
|
|
}
|
|
|
|
if (
|
|
objectHasOwnProperty(cfg, 'FORBID_CONTENTS') &&
|
|
arrayIsArray(cfg.FORBID_CONTENTS)
|
|
) {
|
|
if (FORBID_CONTENTS === DEFAULT_FORBID_CONTENTS) {
|
|
FORBID_CONTENTS = clone(FORBID_CONTENTS);
|
|
}
|
|
|
|
addToSet(FORBID_CONTENTS, cfg.FORBID_CONTENTS, transformCaseFunc);
|
|
}
|
|
|
|
if (
|
|
objectHasOwnProperty(cfg, 'ADD_FORBID_CONTENTS') &&
|
|
arrayIsArray(cfg.ADD_FORBID_CONTENTS)
|
|
) {
|
|
if (FORBID_CONTENTS === DEFAULT_FORBID_CONTENTS) {
|
|
FORBID_CONTENTS = clone(FORBID_CONTENTS);
|
|
}
|
|
|
|
addToSet(FORBID_CONTENTS, cfg.ADD_FORBID_CONTENTS, transformCaseFunc);
|
|
}
|
|
|
|
/* Add #text in case KEEP_CONTENT is set to true */
|
|
if (KEEP_CONTENT) {
|
|
ALLOWED_TAGS['#text'] = true;
|
|
}
|
|
|
|
/* Add html, head and body to ALLOWED_TAGS in case WHOLE_DOCUMENT is true */
|
|
if (WHOLE_DOCUMENT) {
|
|
addToSet(ALLOWED_TAGS, ['html', 'head', 'body']);
|
|
}
|
|
|
|
/* Add tbody to ALLOWED_TAGS in case tables are permitted, see #286, #365 */
|
|
if (ALLOWED_TAGS.table) {
|
|
addToSet(ALLOWED_TAGS, ['tbody']);
|
|
delete FORBID_TAGS.tbody;
|
|
}
|
|
|
|
// Re-derive the active Trusted Types policy from this configuration on
|
|
// every parse. The active policy must never be sticky closure state that
|
|
// outlives the config that set it: a caller-supplied policy left in place
|
|
// after `clearConfig()` — or after a later call that supplied none, or
|
|
// `TRUSTED_TYPES_POLICY: null` — could sign a subsequent "default"
|
|
// `RETURN_TRUSTED_TYPE` result with a foreign, possibly unsafe policy.
|
|
// See GHSA-vxr8-fq34-vvx9.
|
|
if (cfg.TRUSTED_TYPES_POLICY) {
|
|
if (typeof cfg.TRUSTED_TYPES_POLICY.createHTML !== 'function') {
|
|
throw typeErrorCreate(
|
|
'TRUSTED_TYPES_POLICY configuration option must provide a "createHTML" hook.'
|
|
);
|
|
}
|
|
|
|
if (typeof cfg.TRUSTED_TYPES_POLICY.createScriptURL !== 'function') {
|
|
throw typeErrorCreate(
|
|
'TRUSTED_TYPES_POLICY configuration option must provide a "createScriptURL" hook.'
|
|
);
|
|
}
|
|
|
|
// A caller-supplied policy applies to this configuration only.
|
|
const previousTrustedTypesPolicy = trustedTypesPolicy;
|
|
trustedTypesPolicy = cfg.TRUSTED_TYPES_POLICY;
|
|
|
|
// Sign local variables required by `sanitize`. If the supplied policy's
|
|
// `createHTML` is circular (i.e. it calls `DOMPurify.sanitize`), this
|
|
// throws via the re-entrancy guard. Restore the previous policy first so
|
|
// the instance is not left in a poisoned state. See #1422.
|
|
try {
|
|
emptyHTML = _createTrustedHTML('');
|
|
} catch (error) {
|
|
trustedTypesPolicy = previousTrustedTypesPolicy;
|
|
throw error;
|
|
}
|
|
} else if (cfg.TRUSTED_TYPES_POLICY === null) {
|
|
// Explicit opt-out for this call: perform no Trusted Types signing and
|
|
// create nothing (so a strict `trusted-types` CSP that disallows a
|
|
// `dompurify` policy can still call `sanitize` from inside its own
|
|
// policy — see #1422). Resetting to `undefined` rather than a sticky
|
|
// `null` also drops any previously retained caller policy, so it cannot
|
|
// resurface on a later call, while still allowing the next config-less
|
|
// call to restore the internal default policy. See GHSA-vxr8-fq34-vvx9.
|
|
trustedTypesPolicy = undefined;
|
|
emptyHTML = '';
|
|
} else {
|
|
// No policy supplied: keep the currently active policy if one is set — a
|
|
// previously supplied policy is intentionally sticky across config-less
|
|
// calls — otherwise fall back to the instance's own internal policy,
|
|
// created at most once. (A policy supplied for a *single* call still
|
|
// lingers by design; what must not linger is a policy whose configuration
|
|
// has been torn down via `clearConfig()`, which restores the default.)
|
|
if (trustedTypesPolicy === undefined) {
|
|
trustedTypesPolicy = _getDefaultTrustedTypesPolicy();
|
|
}
|
|
|
|
// Sign internal variables only when a policy is active. A falsy policy
|
|
// (Trusted Types unsupported, creation failed, or an explicit opt-out)
|
|
// leaves `emptyHTML` as a plain string, so we never call `.createHTML` on
|
|
// a non-policy and throw. See #1422.
|
|
if (trustedTypesPolicy && typeof emptyHTML === 'string') {
|
|
emptyHTML = _createTrustedHTML('');
|
|
}
|
|
}
|
|
|
|
/*
|
|
* Mirror the clone-before-mutate pattern already applied above for
|
|
* cfg.ADD_TAGS / cfg.ADD_ATTR: if any uponSanitize* hook is
|
|
* registered AND the set still points at the default constant,
|
|
* clone it. The hook then mutates the clone (in-call widening
|
|
* still works exactly as documented) and the next default-cfg
|
|
* call rebinds to the untouched original via the reassignment at
|
|
* the top of this function.
|
|
*/
|
|
if (
|
|
(hooks.uponSanitizeElement.length > 0 ||
|
|
hooks.uponSanitizeAttribute.length > 0) &&
|
|
ALLOWED_TAGS === DEFAULT_ALLOWED_TAGS
|
|
) {
|
|
ALLOWED_TAGS = clone(ALLOWED_TAGS);
|
|
}
|
|
|
|
if (
|
|
hooks.uponSanitizeAttribute.length > 0 &&
|
|
ALLOWED_ATTR === DEFAULT_ALLOWED_ATTR
|
|
) {
|
|
ALLOWED_ATTR = clone(ALLOWED_ATTR);
|
|
}
|
|
|
|
// Prevent further manipulation of configuration.
|
|
// Not available in IE8, Safari 5, etc.
|
|
if (freeze) {
|
|
freeze(cfg);
|
|
}
|
|
|
|
CONFIG = cfg;
|
|
};
|
|
|
|
/* Keep track of all possible SVG and MathML tags
|
|
* so that we can perform the namespace checks
|
|
* correctly. */
|
|
const ALL_SVG_TAGS = addToSet({}, [
|
|
...TAGS.svg,
|
|
...TAGS.svgFilters,
|
|
...TAGS.svgDisallowed,
|
|
]);
|
|
const ALL_MATHML_TAGS = addToSet({}, [
|
|
...TAGS.mathMl,
|
|
...TAGS.mathMlDisallowed,
|
|
]);
|
|
|
|
/**
|
|
* Namespace rules for an element in the SVG namespace.
|
|
*
|
|
* @param tagName the element's lowercase tag name
|
|
* @param parent the (possibly simulated) parent node
|
|
* @param parentTagName the parent's lowercase tag name
|
|
* @returns true if a spec-compliant parser could produce this element
|
|
*/
|
|
const _checkSvgNamespace = function (
|
|
tagName: string,
|
|
parent: { namespaceURI?: string },
|
|
parentTagName: string
|
|
): boolean {
|
|
// The only way to switch from HTML namespace to SVG
|
|
// is via <svg>. If it happens via any other tag, then
|
|
// it should be killed.
|
|
if (parent.namespaceURI === HTML_NAMESPACE) {
|
|
return tagName === 'svg';
|
|
}
|
|
|
|
// The only way to switch from MathML to SVG is via <svg>
|
|
// if the parent is either <annotation-xml> or a MathML
|
|
// text integration point.
|
|
if (parent.namespaceURI === MATHML_NAMESPACE) {
|
|
return (
|
|
tagName === 'svg' &&
|
|
(parentTagName === 'annotation-xml' ||
|
|
MATHML_TEXT_INTEGRATION_POINTS[parentTagName])
|
|
);
|
|
}
|
|
|
|
// We only allow elements that are defined in SVG
|
|
// spec. All others are disallowed in SVG namespace.
|
|
return Boolean(ALL_SVG_TAGS[tagName]);
|
|
};
|
|
|
|
/**
|
|
* Namespace rules for an element in the MathML namespace.
|
|
*
|
|
* @param tagName the element's lowercase tag name
|
|
* @param parent the (possibly simulated) parent node
|
|
* @param parentTagName the parent's lowercase tag name
|
|
* @returns true if a spec-compliant parser could produce this element
|
|
*/
|
|
const _checkMathMlNamespace = function (
|
|
tagName: string,
|
|
parent: { namespaceURI?: string },
|
|
parentTagName: string
|
|
): boolean {
|
|
// The only way to switch from HTML namespace to MathML
|
|
// is via <math>. If it happens via any other tag, then
|
|
// it should be killed.
|
|
if (parent.namespaceURI === HTML_NAMESPACE) {
|
|
return tagName === 'math';
|
|
}
|
|
|
|
// The only way to switch from SVG to MathML is via
|
|
// <math> and HTML integration points
|
|
if (parent.namespaceURI === SVG_NAMESPACE) {
|
|
return tagName === 'math' && HTML_INTEGRATION_POINTS[parentTagName];
|
|
}
|
|
|
|
// We only allow elements that are defined in MathML
|
|
// spec. All others are disallowed in MathML namespace.
|
|
return Boolean(ALL_MATHML_TAGS[tagName]);
|
|
};
|
|
|
|
/**
|
|
* Namespace rules for an element in the HTML namespace.
|
|
*
|
|
* @param tagName the element's lowercase tag name
|
|
* @param parent the (possibly simulated) parent node
|
|
* @param parentTagName the parent's lowercase tag name
|
|
* @returns true if a spec-compliant parser could produce this element
|
|
*/
|
|
const _checkHtmlNamespace = function (
|
|
tagName: string,
|
|
parent: { namespaceURI?: string },
|
|
parentTagName: string
|
|
): boolean {
|
|
// The only way to switch from SVG to HTML is via
|
|
// HTML integration points, and from MathML to HTML
|
|
// is via MathML text integration points
|
|
if (
|
|
parent.namespaceURI === SVG_NAMESPACE &&
|
|
!HTML_INTEGRATION_POINTS[parentTagName]
|
|
) {
|
|
return false;
|
|
}
|
|
|
|
if (
|
|
parent.namespaceURI === MATHML_NAMESPACE &&
|
|
!MATHML_TEXT_INTEGRATION_POINTS[parentTagName]
|
|
) {
|
|
return false;
|
|
}
|
|
|
|
// We disallow tags that are specific for MathML
|
|
// or SVG and should never appear in HTML namespace
|
|
return (
|
|
!ALL_MATHML_TAGS[tagName] &&
|
|
(COMMON_SVG_AND_HTML_ELEMENTS[tagName] || !ALL_SVG_TAGS[tagName])
|
|
);
|
|
};
|
|
|
|
/**
|
|
* @param element a DOM element whose namespace is being checked
|
|
* @returns Return false if the element has a
|
|
* namespace that a spec-compliant parser would never
|
|
* return. Return true otherwise.
|
|
*/
|
|
const _checkValidNamespace = function (element: Element): boolean {
|
|
let parent = getParentNode(element);
|
|
|
|
// In JSDOM, if we're inside shadow DOM, then parentNode
|
|
// can be null. We just simulate parent in this case.
|
|
if (!parent || !parent.tagName) {
|
|
parent = {
|
|
namespaceURI: NAMESPACE,
|
|
tagName: 'template',
|
|
};
|
|
}
|
|
|
|
const tagName = stringToLowerCase(element.tagName);
|
|
const parentTagName = stringToLowerCase(parent.tagName);
|
|
|
|
if (!ALLOWED_NAMESPACES[element.namespaceURI]) {
|
|
return false;
|
|
}
|
|
|
|
if (element.namespaceURI === SVG_NAMESPACE) {
|
|
return _checkSvgNamespace(tagName, parent, parentTagName);
|
|
}
|
|
|
|
if (element.namespaceURI === MATHML_NAMESPACE) {
|
|
return _checkMathMlNamespace(tagName, parent, parentTagName);
|
|
}
|
|
|
|
if (element.namespaceURI === HTML_NAMESPACE) {
|
|
return _checkHtmlNamespace(tagName, parent, parentTagName);
|
|
}
|
|
|
|
// For XHTML and XML documents that support custom namespaces
|
|
if (
|
|
PARSER_MEDIA_TYPE === 'application/xhtml+xml' &&
|
|
ALLOWED_NAMESPACES[element.namespaceURI]
|
|
) {
|
|
return true;
|
|
}
|
|
|
|
// The code should never reach this place (this means
|
|
// that the element somehow got namespace that is not
|
|
// HTML, SVG, MathML or allowed via ALLOWED_NAMESPACES).
|
|
// Return false just in case.
|
|
return false;
|
|
};
|
|
|
|
/**
|
|
* _forceRemove
|
|
*
|
|
* @param node a DOM node
|
|
*/
|
|
const _forceRemove = function (node: Node): void {
|
|
arrayPush(DOMPurify.removed, { element: node });
|
|
|
|
try {
|
|
// eslint-disable-next-line unicorn/prefer-dom-node-remove
|
|
getParentNode(node).removeChild(node);
|
|
} catch (_) {
|
|
/* The normal detach failed — this is reached for a parentless node
|
|
(getParentNode() is null, so .removeChild throws). Element.prototype
|
|
.remove() is itself a spec no-op on a parentless node, so a recorded
|
|
"removal" would otherwise hand the caller back an intact,
|
|
payload-bearing node (e.g. a detached IN_PLACE root the mXSS canary or
|
|
the style-with-element-child rule decided to kill). Fail closed by
|
|
throwing — exactly as a clobbered root does at the IN_PLACE entry —
|
|
rather than trying to "neutralize" the node via its own methods.
|
|
Neutralizing would mean calling getAttributeNames()/removeAttribute()
|
|
on the node, both of which a <form> root can clobber via a named child
|
|
(and _isClobbered does not even probe getAttributeNames), so the
|
|
neutralize step could itself be silently defeated, leaving the payload
|
|
intact. A throw touches only the cached, clobber-safe remove() and
|
|
getParentNode(). Generalizes GHSA-r47g-fvhr-h676 (clobbered-form root)
|
|
to every root-kill reason. REPORT-3.
|
|
|
|
This lives inside the catch, so it never fires for a normally-removed
|
|
in-tree node: those have a parent, removeChild() succeeds, and the
|
|
catch is not entered. Only a kept (parentless) root reaches here. */
|
|
remove(node);
|
|
|
|
if (!getParentNode(node)) {
|
|
throw typeErrorCreate(
|
|
'a node selected for removal could not be detached from its tree ' +
|
|
'and cannot be safely returned; refusing to sanitize in place'
|
|
);
|
|
}
|
|
}
|
|
};
|
|
|
|
/**
|
|
* _neutralizeRoot
|
|
*
|
|
* Fail-closed teardown of an in-place root after the sanitize walk aborts
|
|
* (campaign-3 F2). An internal throw mid-walk — e.g. a page-registered
|
|
* custom element's reaction detaches a node so `_forceRemove`'s deliberate
|
|
* parentless guard throws, or any other re-entrant engine mutation — would
|
|
* otherwise leave the caller's *live* tree half-sanitized, with everything
|
|
* after the abort point still carrying its handlers. There is no safe way
|
|
* to resume the walk (the tree mutated under us), so we strip the root bare:
|
|
* remove every child and every attribute, then let the caller's catch see
|
|
* the original error. Clobber-safe (cached `remove`/`childNodes`/`attributes`
|
|
* getters; the root was already clobber-pre-flighted at the IN_PLACE entry).
|
|
*
|
|
* @param root the in-place root to empty
|
|
*/
|
|
const _neutralizeRoot = function (root: Node): void {
|
|
const childNodes = getChildNodes(root);
|
|
if (childNodes) {
|
|
const snapshot: Node[] = [];
|
|
arrayForEach(childNodes, (child) => {
|
|
arrayPush(snapshot, child);
|
|
});
|
|
arrayForEach(snapshot, (child) => {
|
|
try {
|
|
remove(child);
|
|
} catch (_) {
|
|
/* Best-effort teardown; a still-attached child is handled below */
|
|
}
|
|
});
|
|
}
|
|
|
|
const attributes = getAttributes(root);
|
|
if (attributes) {
|
|
for (let i = attributes.length - 1; i >= 0; --i) {
|
|
const attribute = attributes[i];
|
|
const name = attribute && attribute.name;
|
|
if (typeof name === 'string') {
|
|
try {
|
|
(root as Element).removeAttribute(name);
|
|
} catch (_) {
|
|
/* Clobbered removeAttribute — ignore (fail-closed best effort) */
|
|
}
|
|
}
|
|
}
|
|
}
|
|
};
|
|
|
|
/**
|
|
* _removeAttribute
|
|
*
|
|
* @param name an Attribute name
|
|
* @param element a DOM node
|
|
*/
|
|
const _removeAttribute = function (name: string, element: Element): void {
|
|
try {
|
|
arrayPush(DOMPurify.removed, {
|
|
attribute: element.getAttributeNode(name),
|
|
from: element,
|
|
});
|
|
} catch (_) {
|
|
arrayPush(DOMPurify.removed, {
|
|
attribute: null,
|
|
from: element,
|
|
});
|
|
}
|
|
|
|
element.removeAttribute(name);
|
|
|
|
// We void attribute values for unremovable "is" attributes
|
|
if (name === 'is') {
|
|
if (RETURN_DOM || RETURN_DOM_FRAGMENT) {
|
|
try {
|
|
_forceRemove(element);
|
|
} catch (_) {}
|
|
} else {
|
|
try {
|
|
element.setAttribute(name, '');
|
|
} catch (_) {}
|
|
}
|
|
}
|
|
};
|
|
|
|
/**
|
|
* _stripDisallowedAttributes
|
|
*
|
|
* Removes every attribute the active configuration does not allow from a
|
|
* single element, using the same allowlist as the main attribute pass (so
|
|
* `on*` handlers go, but no `/^on/` blocklist is introduced). Used only to
|
|
* neutralise nodes that are being discarded from an in-place tree.
|
|
*
|
|
* @param element the element to strip
|
|
*/
|
|
const _stripDisallowedAttributes = function (element: Element): void {
|
|
const attributes = getAttributes(element);
|
|
if (!attributes) {
|
|
return;
|
|
}
|
|
|
|
for (let i = attributes.length - 1; i >= 0; --i) {
|
|
const attribute = attributes[i];
|
|
const name = attribute && attribute.name;
|
|
if (typeof name !== 'string' || ALLOWED_ATTR[transformCaseFunc(name)]) {
|
|
continue;
|
|
}
|
|
|
|
try {
|
|
element.removeAttribute(name);
|
|
} catch (_) {
|
|
/* Clobbered removeAttribute on a doomed node — ignore */
|
|
}
|
|
}
|
|
};
|
|
|
|
/**
|
|
* _neutralizeSubtree
|
|
*
|
|
* Completes the audit-5 F1 fix across every removal path. The KEEP_CONTENT
|
|
* move-hoist neutralises only disallowed-tag removals; clobber, mXSS-canary,
|
|
* namespace, comment, processing-instruction and KEEP_CONTENT:false removals
|
|
* all drop their subtree wholesale via `_forceRemove`. On the IN_PLACE path
|
|
* those dropped nodes are detached from the caller's LIVE tree but a
|
|
* handler-bearing original among them (an `<img onerror>`/`<video>` that was
|
|
* loading) keeps its queued resource event, which fires in page scope after
|
|
* sanitize returns. This walks a removed subtree and strips every attribute
|
|
* the active configuration does not allow — so `on*` handlers are cancelled
|
|
* through the SAME allowlist that governs kept nodes, not a separate `/^on/`
|
|
* blocklist. Run synchronously before sanitize returns, i.e. before any
|
|
* queued event can fire. Hook-free by design: these nodes leave the output,
|
|
* so firing attribute hooks for them would be surprising. Clobber-safe reads;
|
|
* a doomed clobbered node may shadow `removeAttribute` (its own attributes are
|
|
* irrelevant — it is discarded — while its non-clobbered descendants, e.g.
|
|
* the `<img>`, are reached and scrubbed).
|
|
*
|
|
* @param root the root of a removed subtree to neutralise
|
|
*/
|
|
const _neutralizeSubtree = function (root: Node): void {
|
|
const stack: Node[] = [root];
|
|
|
|
while (stack.length > 0) {
|
|
const node = stack.pop();
|
|
const nodeType = getNodeType ? getNodeType(node) : (node as any).nodeType;
|
|
|
|
if (nodeType === NODE_TYPE.element) {
|
|
_stripDisallowedAttributes(node as Element);
|
|
}
|
|
|
|
const childNodes = getChildNodes(node);
|
|
if (childNodes) {
|
|
for (let i = childNodes.length - 1; i >= 0; --i) {
|
|
stack.push(childNodes[i]);
|
|
}
|
|
}
|
|
}
|
|
};
|
|
|
|
/**
|
|
* _initDocument
|
|
*
|
|
* @param dirty - a string of dirty markup
|
|
* @return a DOM, filled with the dirty markup
|
|
*/
|
|
const _initDocument = function (dirty: string): Document {
|
|
/* Create a HTML document */
|
|
let doc = null;
|
|
let leadingWhitespace = null;
|
|
|
|
if (FORCE_BODY) {
|
|
dirty = '<remove></remove>' + dirty;
|
|
} else {
|
|
/* If FORCE_BODY isn't used, leading whitespace needs to be preserved manually */
|
|
const matches = stringMatch(dirty, /^[\r\n\t ]+/);
|
|
leadingWhitespace = matches && matches[0];
|
|
}
|
|
|
|
if (
|
|
PARSER_MEDIA_TYPE === 'application/xhtml+xml' &&
|
|
NAMESPACE === HTML_NAMESPACE
|
|
) {
|
|
// Root of XHTML doc must contain xmlns declaration (see https://www.w3.org/TR/xhtml1/normative.html#strict)
|
|
dirty =
|
|
'<html xmlns="http://www.w3.org/1999/xhtml"><head></head><body>' +
|
|
dirty +
|
|
'</body></html>';
|
|
}
|
|
|
|
const dirtyPayload = trustedTypesPolicy ? _createTrustedHTML(dirty) : dirty;
|
|
/*
|
|
* Use the DOMParser API by default, fallback later if needs be
|
|
* DOMParser not work for svg when has multiple root element.
|
|
*/
|
|
if (NAMESPACE === HTML_NAMESPACE) {
|
|
try {
|
|
doc = new DOMParser().parseFromString(dirtyPayload, PARSER_MEDIA_TYPE);
|
|
} catch (_) {}
|
|
}
|
|
|
|
/* Use createHTMLDocument in case DOMParser is not available */
|
|
if (!doc || !doc.documentElement) {
|
|
doc = implementation.createDocument(NAMESPACE, 'template', null);
|
|
try {
|
|
doc.documentElement.innerHTML = IS_EMPTY_INPUT
|
|
? emptyHTML
|
|
: dirtyPayload;
|
|
} catch (_) {
|
|
// Syntax error if dirtyPayload is invalid xml
|
|
}
|
|
}
|
|
|
|
const body = doc.body || doc.documentElement;
|
|
|
|
if (dirty && leadingWhitespace) {
|
|
body.insertBefore(
|
|
document.createTextNode(leadingWhitespace),
|
|
body.childNodes[0] || null
|
|
);
|
|
}
|
|
|
|
/* Work on whole document or just its body */
|
|
if (NAMESPACE === HTML_NAMESPACE) {
|
|
return getElementsByTagName.call(
|
|
doc,
|
|
WHOLE_DOCUMENT ? 'html' : 'body'
|
|
)[0];
|
|
}
|
|
|
|
return WHOLE_DOCUMENT ? doc.documentElement : body;
|
|
};
|
|
|
|
/**
|
|
* Creates a NodeIterator object that you can use to traverse filtered lists of nodes or elements in a document.
|
|
*
|
|
* @param root The root element or node to start traversing on.
|
|
* @return The created NodeIterator
|
|
*/
|
|
const _createNodeIterator = function (root: Node): NodeIterator {
|
|
return createNodeIterator.call(
|
|
root.ownerDocument || root,
|
|
root,
|
|
// eslint-disable-next-line no-bitwise
|
|
NodeFilter.SHOW_ELEMENT |
|
|
NodeFilter.SHOW_COMMENT |
|
|
NodeFilter.SHOW_TEXT |
|
|
NodeFilter.SHOW_PROCESSING_INSTRUCTION |
|
|
NodeFilter.SHOW_CDATA_SECTION,
|
|
null
|
|
);
|
|
};
|
|
|
|
/**
|
|
* Replace template expression syntax (mustache, ERB, template
|
|
* literal) with a space; shared by all SAFE_FOR_TEMPLATES scrub
|
|
* sites. Order matters: mustache, then ERB, then template literal.
|
|
*
|
|
* @param value the string to scrub
|
|
* @returns the scrubbed string
|
|
*/
|
|
const _stripTemplateExpressions = function (value: string): string {
|
|
value = stringReplace(value, MUSTACHE_EXPR, ' ');
|
|
value = stringReplace(value, ERB_EXPR, ' ');
|
|
value = stringReplace(value, TMPLIT_EXPR, ' ');
|
|
return value;
|
|
};
|
|
|
|
/**
|
|
* Strip template-engine expressions ({{...}}, ${...}, <%...%>) from the
|
|
* character data of an element subtree. Used as the final safety net for
|
|
* SAFE_FOR_TEMPLATES on every DOM-returning code path so that expressions
|
|
* which only form after text-node normalization (e.g. fragments split across
|
|
* stripped elements) cannot survive into a template-evaluating framework.
|
|
*
|
|
* Walks text/comment/CDATA/processing-instruction nodes and mutates `.data`
|
|
* in place rather than round-tripping through innerHTML. This preserves
|
|
* descendant node references (important for IN_PLACE callers), avoids a
|
|
* serialize/reparse cycle, and reads literal character data — which means
|
|
* `<%...%>` in text content matches the ERB regex against its real bytes
|
|
* instead of the HTML-entity-escaped form innerHTML would produce.
|
|
*
|
|
* Attribute values are not visited here; SAFE_FOR_TEMPLATES handling for
|
|
* attributes is performed during the per-node `_sanitizeAttributes` pass.
|
|
*
|
|
* @param node The root element whose character data should be scrubbed.
|
|
*/
|
|
const _scrubTemplateExpressions = function (node: Element): void {
|
|
node.normalize();
|
|
const walker = createNodeIterator.call(
|
|
node.ownerDocument || node,
|
|
node,
|
|
// eslint-disable-next-line no-bitwise
|
|
NodeFilter.SHOW_TEXT |
|
|
NodeFilter.SHOW_COMMENT |
|
|
NodeFilter.SHOW_CDATA_SECTION |
|
|
NodeFilter.SHOW_PROCESSING_INSTRUCTION,
|
|
null
|
|
);
|
|
|
|
let currentNode = walker.nextNode() as CharacterData | null;
|
|
while (currentNode) {
|
|
currentNode.data = _stripTemplateExpressions(currentNode.data);
|
|
currentNode = walker.nextNode() as CharacterData | null;
|
|
}
|
|
|
|
// NodeIterator does not descend into <template>.content per the DOM spec,
|
|
// so we must explicitly recurse into each template's content fragment,
|
|
// mirroring the approach used by _sanitizeShadowDOM.
|
|
const templates = node.querySelectorAll?.('template');
|
|
if (templates) {
|
|
arrayForEach(templates, (tmpl: HTMLTemplateElement) => {
|
|
if (_isDocumentFragment(tmpl.content)) {
|
|
_scrubTemplateExpressions(tmpl.content as unknown as Element);
|
|
}
|
|
});
|
|
}
|
|
};
|
|
|
|
/**
|
|
* _isClobbered
|
|
*
|
|
* Detect DOM-clobbering on HTMLFormElement nodes. Form is the only HTML
|
|
* interface with [LegacyOverrideBuiltIns]; a descendant element with a
|
|
* `name` attribute matching a prototype property shadows that property
|
|
* on direct reads. We use this check at the IN_PLACE entry-point and
|
|
* during attribute sanitization to refuse clobbered forms.
|
|
*
|
|
* @param element element to check for clobbering attacks
|
|
* @return true if clobbered, false if safe
|
|
*/
|
|
const _isClobbered = function (element: Element): boolean {
|
|
// Realm-independent tag-name probe. If we can't determine the tag
|
|
// name at all, we can't reason about clobbering — return false
|
|
// (the caller's other defences still apply).
|
|
const realTagName = getNodeName ? getNodeName(element) : null;
|
|
if (typeof realTagName !== 'string') {
|
|
return false;
|
|
}
|
|
|
|
if (transformCaseFunc(realTagName) !== 'form') {
|
|
return false;
|
|
}
|
|
|
|
return (
|
|
typeof element.nodeName !== 'string' ||
|
|
typeof element.textContent !== 'string' ||
|
|
typeof element.removeChild !== 'function' ||
|
|
// Realm-safe NamedNodeMap detection: equality against the cached
|
|
// prototype getter. Clobbered .attributes (e.g. <input name="attributes">)
|
|
// makes the direct read diverge from the cached read; a clean form
|
|
// (same-realm OR foreign-realm) has both reads pointing at the same
|
|
// canonical NamedNodeMap.
|
|
element.attributes !== getAttributes(element) ||
|
|
typeof element.removeAttribute !== 'function' ||
|
|
typeof element.setAttribute !== 'function' ||
|
|
typeof element.namespaceURI !== 'string' ||
|
|
typeof element.insertBefore !== 'function' ||
|
|
typeof element.hasChildNodes !== 'function' ||
|
|
// NodeType clobbering probe. Cached Node.prototype.nodeType getter
|
|
// returns the integer 1 for any Element regardless of realm; direct
|
|
// read on a clobbered form (e.g. <input name="nodeType">) returns
|
|
// the named child element. Cheap addition — nodeType is read from
|
|
// an internal slot, no serialization cost — and removes a residual
|
|
// clobbering surface used by several mXSS / PI / comment branches
|
|
// in _sanitizeElements that compare currentNode.nodeType directly.
|
|
element.nodeType !== getNodeType(element) ||
|
|
// HTMLFormElement has [LegacyOverrideBuiltIns]: a descendant named
|
|
// "childNodes" shadows the prototype getter. Direct reads of
|
|
// form.childNodes from a clobbered form return the named child
|
|
// instead of the real NodeList, so any walk that reads it directly
|
|
// skips the form's real children. Compare the direct read to the
|
|
// cached Node.prototype getter — when the form's named-property
|
|
// getter intercepts the read, the two values differ and we flag
|
|
// the form. This catches every clobbering child type (input,
|
|
// select, etc.) regardless of whether the named child happens to
|
|
// carry a numeric .length, which a typeof-based probe would miss
|
|
// (e.g. HTMLSelectElement.length is a defined unsigned-long).
|
|
element.childNodes !== getChildNodes(element)
|
|
);
|
|
};
|
|
|
|
/**
|
|
* Checks whether the given value is a DocumentFragment from any realm.
|
|
*
|
|
* The realm-independent replacement reads `nodeType` through the cached
|
|
* Node.prototype getter and compares to the DOCUMENT_FRAGMENT_NODE
|
|
* constant (11). nodeType is a numeric value resolved from the node's
|
|
* internal slot, identical across realms for the same kind of node.
|
|
*
|
|
* @param value object to check
|
|
* @return true if value is a DocumentFragment-shaped node from any realm
|
|
*/
|
|
const _isDocumentFragment = function (value: unknown): boolean {
|
|
if (!getNodeType || typeof value !== 'object' || value === null) {
|
|
return false;
|
|
}
|
|
|
|
try {
|
|
return getNodeType(value as Node) === NODE_TYPE.documentFragment;
|
|
} catch (_) {
|
|
return false;
|
|
}
|
|
};
|
|
|
|
/**
|
|
* Checks whether the given object is a DOM node, including nodes that
|
|
* originate from a different window/realm (e.g. an iframe's
|
|
* contentDocument). The previous `value instanceof Node` check was
|
|
* realm-bound: nodes from a different window failed it, causing
|
|
* sanitize() to silently stringify them and reset IN_PLACE to false,
|
|
* returning the original node unsanitized. See GHSA-4w3q-35jp-p934.
|
|
*
|
|
* @param value object to check whether it's a DOM node
|
|
* @return true if value is a DOM node from any realm
|
|
*/
|
|
const _isNode = function (value: unknown): value is Node {
|
|
if (!getNodeType || typeof value !== 'object' || value === null) {
|
|
return false;
|
|
}
|
|
|
|
try {
|
|
return typeof getNodeType(value) === 'number';
|
|
} catch (_) {
|
|
return false;
|
|
}
|
|
};
|
|
|
|
function _executeHooks<T extends HookFunction>(
|
|
hooks: HookFunction[],
|
|
currentNode: Parameters<T>[0],
|
|
data: Parameters<T>[1]
|
|
): void {
|
|
if (hooks.length === 0) {
|
|
return;
|
|
}
|
|
|
|
arrayForEach(hooks, (hook: T) => {
|
|
hook.call(DOMPurify, currentNode, data, CONFIG);
|
|
});
|
|
}
|
|
|
|
/**
|
|
* Structural-threat checks that condemn a node regardless of the
|
|
* allowlists: mXSS via namespace confusion, risky CSS construction,
|
|
* processing instructions, markup-bearing comments. Pure predicate;
|
|
* the caller removes. Check order is load-bearing.
|
|
*
|
|
* @param currentNode the node to inspect
|
|
* @param tagName the node's transformCaseFunc'd tag name
|
|
* @return true if the node must be removed
|
|
*/
|
|
const _isUnsafeNode = function (currentNode: any, tagName: string): boolean {
|
|
/* Detect mXSS attempts abusing namespace confusion */
|
|
if (
|
|
SAFE_FOR_XML &&
|
|
currentNode.hasChildNodes() &&
|
|
!_isNode(currentNode.firstElementChild) &&
|
|
regExpTest(EXPRESSIONS.ELEMENT_MARKUP_PROBE, currentNode.textContent) &&
|
|
regExpTest(EXPRESSIONS.ELEMENT_MARKUP_PROBE, currentNode.innerHTML)
|
|
) {
|
|
return true;
|
|
}
|
|
|
|
/* Remove risky CSS construction leading to mXSS */
|
|
if (
|
|
SAFE_FOR_XML &&
|
|
currentNode.namespaceURI === HTML_NAMESPACE &&
|
|
tagName === 'style' &&
|
|
_isNode(currentNode.firstElementChild)
|
|
) {
|
|
return true;
|
|
}
|
|
|
|
/* Remove any occurrence of processing instructions */
|
|
if (currentNode.nodeType === NODE_TYPE.processingInstruction) {
|
|
return true;
|
|
}
|
|
|
|
/* Remove any kind of possibly harmful comments */
|
|
if (
|
|
SAFE_FOR_XML &&
|
|
currentNode.nodeType === NODE_TYPE.comment &&
|
|
regExpTest(EXPRESSIONS.COMMENT_MARKUP_PROBE, currentNode.data)
|
|
) {
|
|
return true;
|
|
}
|
|
|
|
return false;
|
|
};
|
|
|
|
/**
|
|
* Handle a node whose tag is forbidden or not allowlisted: keep
|
|
* allowed custom elements (false return exits _sanitizeElements
|
|
* early - namespace/fallback checks and the afterSanitizeElements
|
|
* hook are intentionally skipped for kept custom elements), else
|
|
* hoist content per KEEP_CONTENT and remove.
|
|
*
|
|
* @param currentNode the disallowed node
|
|
* @param tagName the node's transformCaseFunc'd tag name
|
|
* @return true if the node was removed, false if kept
|
|
*/
|
|
const _sanitizeDisallowedNode = function (
|
|
currentNode: any,
|
|
tagName: string
|
|
): boolean {
|
|
/* Check if we have a custom element to handle */
|
|
if (!FORBID_TAGS[tagName] && _isBasicCustomElement(tagName)) {
|
|
if (
|
|
CUSTOM_ELEMENT_HANDLING.tagNameCheck instanceof RegExp &&
|
|
regExpTest(CUSTOM_ELEMENT_HANDLING.tagNameCheck, tagName)
|
|
) {
|
|
return false;
|
|
}
|
|
|
|
if (
|
|
CUSTOM_ELEMENT_HANDLING.tagNameCheck instanceof Function &&
|
|
CUSTOM_ELEMENT_HANDLING.tagNameCheck(tagName)
|
|
) {
|
|
return false;
|
|
}
|
|
}
|
|
|
|
/* Keep content except for bad-listed elements.
|
|
Use the cached prototype getters exclusively — the previous code
|
|
had `|| currentNode.parentNode` / `|| currentNode.childNodes`
|
|
fallbacks, but the cached getters always return the canonical
|
|
value (or null for a real parent-less node), so the fallback
|
|
path was dead in safe cases and a clobbering surface in unsafe
|
|
ones. Falsy cached results stay falsy; the `if (childNodes &&
|
|
parentNode)` check already gates correctly. */
|
|
if (KEEP_CONTENT && !FORBID_CONTENTS[tagName]) {
|
|
const parentNode = getParentNode(currentNode);
|
|
const childNodes = getChildNodes(currentNode);
|
|
|
|
if (childNodes && parentNode) {
|
|
const childCount = childNodes.length;
|
|
|
|
/* In-place: hoist the *original* children so the iterator visits
|
|
and sanitises them through the same allowlist pass as every other
|
|
node. The caller built the tree in the live document, so the
|
|
originals carry already-queued resource events (`<img onerror>`,
|
|
`<video>`/`<audio>` error, lazy/`onload`, …); cloning would leave
|
|
those originals detached but still armed, firing in page scope
|
|
while the returned tree looked clean. Moving is safe in-place: the
|
|
root is pre-validated as an allowed tag and so is never the node
|
|
being removed, which keeps `parentNode` inside the iterator root
|
|
and the relocated child inside the serialised tree.
|
|
|
|
Otherwise (string / DOM-copy paths): clone. The iterator is rooted
|
|
at — and the result serialised from — `body`, so a restrictive
|
|
ALLOWED_TAGS that removes `body` itself must leave its content in
|
|
place, which only cloning does; and those paths parse into an
|
|
inert document, so their discarded originals never had a queued
|
|
event to neutralise.
|
|
|
|
`childNodes` is live; a tail-to-head walk keeps `childNodes[i]`
|
|
valid whether we move (drops the trailing entry) or clone (leaves
|
|
the list intact). */
|
|
for (let i = childCount - 1; i >= 0; --i) {
|
|
const hoisted = IN_PLACE
|
|
? childNodes[i]
|
|
: cloneNode(childNodes[i], true);
|
|
parentNode.insertBefore(hoisted, getNextSibling(currentNode));
|
|
}
|
|
}
|
|
}
|
|
|
|
_forceRemove(currentNode);
|
|
return true;
|
|
};
|
|
|
|
/**
|
|
* _sanitizeElements
|
|
*
|
|
* @protect nodeName
|
|
* @protect textContent
|
|
* @protect removeChild
|
|
* @param currentNode to check for permission to exist
|
|
* @return true if node was killed, false if left alive
|
|
*/
|
|
const _sanitizeElements = function (currentNode: any): boolean {
|
|
/* Execute a hook if present */
|
|
_executeHooks(hooks.beforeSanitizeElements, currentNode, null);
|
|
|
|
/* Check if element is clobbered or can clobber */
|
|
if (_isClobbered(currentNode)) {
|
|
_forceRemove(currentNode);
|
|
return true;
|
|
}
|
|
|
|
/* Now let's check the element's type and name */
|
|
const tagName = transformCaseFunc(
|
|
getNodeName ? getNodeName(currentNode) : currentNode.nodeName
|
|
);
|
|
|
|
/* Execute a hook if present */
|
|
_executeHooks(hooks.uponSanitizeElement, currentNode, {
|
|
tagName,
|
|
allowedTags: ALLOWED_TAGS,
|
|
});
|
|
|
|
/* Remove mXSS vectors, processing instructions and risky comments */
|
|
if (_isUnsafeNode(currentNode, tagName)) {
|
|
_forceRemove(currentNode);
|
|
return true;
|
|
}
|
|
|
|
/* Remove element if anything forbids its presence */
|
|
if (
|
|
FORBID_TAGS[tagName] ||
|
|
(!(
|
|
EXTRA_ELEMENT_HANDLING.tagCheck instanceof Function &&
|
|
EXTRA_ELEMENT_HANDLING.tagCheck(tagName)
|
|
) &&
|
|
!ALLOWED_TAGS[tagName])
|
|
) {
|
|
return _sanitizeDisallowedNode(currentNode, tagName);
|
|
}
|
|
|
|
/* Check whether element has a valid namespace.
|
|
Realm-safe check (GHSA-hpcv-96wg-7vj8): use the cached Node.prototype
|
|
nodeType getter rather than `instanceof Element`, which is realm-
|
|
bound and short-circuits to false for any node minted in a different
|
|
realm — letting a foreign-realm element with a forbidden namespace
|
|
slip past the namespace check entirely. */
|
|
const nt = getNodeType ? getNodeType(currentNode) : currentNode.nodeType;
|
|
if (nt === NODE_TYPE.element && !_checkValidNamespace(currentNode)) {
|
|
_forceRemove(currentNode);
|
|
return true;
|
|
}
|
|
|
|
/* Make sure that older browsers don't get fallback-tag mXSS */
|
|
if (
|
|
(tagName === 'noscript' ||
|
|
tagName === 'noembed' ||
|
|
tagName === 'noframes') &&
|
|
regExpTest(EXPRESSIONS.FALLBACK_TAG_CLOSE, currentNode.innerHTML)
|
|
) {
|
|
_forceRemove(currentNode);
|
|
return true;
|
|
}
|
|
|
|
/* Sanitize element content to be template-safe */
|
|
if (SAFE_FOR_TEMPLATES && currentNode.nodeType === NODE_TYPE.text) {
|
|
/* Get the element's text content */
|
|
const content = _stripTemplateExpressions(currentNode.textContent);
|
|
|
|
if (currentNode.textContent !== content) {
|
|
arrayPush(DOMPurify.removed, { element: currentNode.cloneNode() });
|
|
currentNode.textContent = content;
|
|
}
|
|
}
|
|
|
|
/* Execute a hook if present */
|
|
_executeHooks(hooks.afterSanitizeElements, currentNode, null);
|
|
|
|
return false;
|
|
};
|
|
|
|
/**
|
|
* _isValidAttribute
|
|
*
|
|
* @param lcTag Lowercase tag name of containing element.
|
|
* @param lcName Lowercase attribute name.
|
|
* @param value Attribute value.
|
|
* @return Returns true if `value` is valid, otherwise false.
|
|
*/
|
|
// eslint-disable-next-line complexity
|
|
const _isValidAttribute = function (
|
|
lcTag: string,
|
|
lcName: string,
|
|
value: string
|
|
): boolean {
|
|
/* FORBID_ATTR must always win, even if ADD_ATTR predicate would allow it */
|
|
if (FORBID_ATTR[lcName]) {
|
|
return false;
|
|
}
|
|
|
|
/* Make sure attribute cannot clobber */
|
|
if (
|
|
SANITIZE_DOM &&
|
|
(lcName === 'id' || lcName === 'name') &&
|
|
(value in document || value in formElement)
|
|
) {
|
|
return false;
|
|
}
|
|
|
|
const nameIsPermitted =
|
|
ALLOWED_ATTR[lcName] ||
|
|
(EXTRA_ELEMENT_HANDLING.attributeCheck instanceof Function &&
|
|
EXTRA_ELEMENT_HANDLING.attributeCheck(lcName, lcTag));
|
|
|
|
/* Allow valid data-* attributes: At least one character after "-"
|
|
(https://html.spec.whatwg.org/multipage/dom.html#embedding-custom-non-visible-data-with-the-data-*-attributes)
|
|
XML-compatible (https://html.spec.whatwg.org/multipage/infrastructure.html#xml-compatible and http://www.w3.org/TR/xml/#d0e804)
|
|
We don't need to check the value; it's always URI safe. */
|
|
if (ALLOW_DATA_ATTR && regExpTest(DATA_ATTR, lcName)) {
|
|
// This attribute is safe
|
|
} else if (ALLOW_ARIA_ATTR && regExpTest(ARIA_ATTR, lcName)) {
|
|
// This attribute is safe
|
|
/* Otherwise, check the name is permitted */
|
|
} else if (!nameIsPermitted) {
|
|
if (
|
|
// First condition does a very basic check if a) it's basically a valid custom element tagname AND
|
|
// b) if the tagName passes whatever the user has configured for CUSTOM_ELEMENT_HANDLING.tagNameCheck
|
|
// and c) if the attribute name passes whatever the user has configured for CUSTOM_ELEMENT_HANDLING.attributeNameCheck
|
|
(_isBasicCustomElement(lcTag) &&
|
|
((CUSTOM_ELEMENT_HANDLING.tagNameCheck instanceof RegExp &&
|
|
regExpTest(CUSTOM_ELEMENT_HANDLING.tagNameCheck, lcTag)) ||
|
|
(CUSTOM_ELEMENT_HANDLING.tagNameCheck instanceof Function &&
|
|
CUSTOM_ELEMENT_HANDLING.tagNameCheck(lcTag))) &&
|
|
((CUSTOM_ELEMENT_HANDLING.attributeNameCheck instanceof RegExp &&
|
|
regExpTest(CUSTOM_ELEMENT_HANDLING.attributeNameCheck, lcName)) ||
|
|
(CUSTOM_ELEMENT_HANDLING.attributeNameCheck instanceof Function &&
|
|
CUSTOM_ELEMENT_HANDLING.attributeNameCheck(lcName, lcTag)))) ||
|
|
// Alternative, second condition checks if it's an `is`-attribute, AND
|
|
// the value passes whatever the user has configured for CUSTOM_ELEMENT_HANDLING.tagNameCheck
|
|
(lcName === 'is' &&
|
|
CUSTOM_ELEMENT_HANDLING.allowCustomizedBuiltInElements &&
|
|
((CUSTOM_ELEMENT_HANDLING.tagNameCheck instanceof RegExp &&
|
|
regExpTest(CUSTOM_ELEMENT_HANDLING.tagNameCheck, value)) ||
|
|
(CUSTOM_ELEMENT_HANDLING.tagNameCheck instanceof Function &&
|
|
CUSTOM_ELEMENT_HANDLING.tagNameCheck(value))))
|
|
) {
|
|
// If user has supplied a regexp or function in CUSTOM_ELEMENT_HANDLING.tagNameCheck, we need to also allow derived custom elements using the same tagName test.
|
|
// Additionally, we need to allow attributes passing the CUSTOM_ELEMENT_HANDLING.attributeNameCheck user has configured, as custom elements can define these at their own discretion.
|
|
} else {
|
|
return false;
|
|
}
|
|
/* Check value is safe. First, is attr inert? If so, is safe */
|
|
} else if (URI_SAFE_ATTRIBUTES[lcName]) {
|
|
// This attribute is safe
|
|
/* Check no script, data or unknown possibly unsafe URI
|
|
unless we know URI values are safe for that attribute */
|
|
} else if (
|
|
regExpTest(IS_ALLOWED_URI, stringReplace(value, ATTR_WHITESPACE, ''))
|
|
) {
|
|
// This attribute is safe
|
|
/* Keep image data URIs alive if src/xlink:href is allowed */
|
|
/* Further prevent gadget XSS for dynamically built script tags */
|
|
} else if (
|
|
(lcName === 'src' || lcName === 'xlink:href' || lcName === 'href') &&
|
|
lcTag !== 'script' &&
|
|
stringIndexOf(value, 'data:') === 0 &&
|
|
DATA_URI_TAGS[lcTag]
|
|
) {
|
|
// This attribute is safe
|
|
/* Allow unknown protocols: This provides support for links that
|
|
are handled by protocol handlers which may be unknown ahead of
|
|
time, e.g. fb:, spotify: */
|
|
} else if (
|
|
ALLOW_UNKNOWN_PROTOCOLS &&
|
|
!regExpTest(IS_SCRIPT_OR_DATA, stringReplace(value, ATTR_WHITESPACE, ''))
|
|
) {
|
|
// This attribute is safe
|
|
/* Check for binary attributes */
|
|
} else if (value) {
|
|
return false;
|
|
} else {
|
|
// Binary attributes are safe at this point
|
|
/* Anything else, presume unsafe, do not add it back */
|
|
}
|
|
|
|
return true;
|
|
};
|
|
|
|
/* Names the HTML spec reserves from valid-custom-element-name; these must
|
|
* never be treated as basic custom elements even when a permissive
|
|
* CUSTOM_ELEMENT_HANDLING.tagNameCheck is configured. */
|
|
const RESERVED_CUSTOM_ELEMENT_NAMES = addToSet({}, [
|
|
'annotation-xml',
|
|
'color-profile',
|
|
'font-face',
|
|
'font-face-format',
|
|
'font-face-name',
|
|
'font-face-src',
|
|
'font-face-uri',
|
|
'missing-glyph',
|
|
]);
|
|
|
|
/**
|
|
* _isBasicCustomElement
|
|
* checks if at least one dash is included in tagName, and it's not the first char
|
|
* for more sophisticated checking see https://github.com/sindresorhus/validate-element-name
|
|
*
|
|
* @param tagName name of the tag of the node to sanitize
|
|
* @returns Returns true if the tag name meets the basic criteria for a custom element, otherwise false.
|
|
*/
|
|
const _isBasicCustomElement = function (tagName: string): boolean {
|
|
return (
|
|
!RESERVED_CUSTOM_ELEMENT_NAMES[stringToLowerCase(tagName)] &&
|
|
regExpTest(CUSTOM_ELEMENT, tagName)
|
|
);
|
|
};
|
|
|
|
/**
|
|
* Wrap an attribute value in the matching Trusted Types object when
|
|
* the active policy requires it. Namespaced attributes pass through
|
|
* unchanged (no TT support yet, see
|
|
* https://bugs.chromium.org/p/chromium/issues/detail?id=1305293).
|
|
*
|
|
* @param lcTag lowercase tag name of the containing element
|
|
* @param lcName lowercase attribute name
|
|
* @param namespaceURI the attribute's namespace, if any
|
|
* @param value the attribute value to wrap
|
|
* @return the value, wrapped when Trusted Types demand it
|
|
*/
|
|
const _applyTrustedTypesToAttribute = function (
|
|
lcTag: string,
|
|
lcName: string,
|
|
namespaceURI: string | null,
|
|
value: string
|
|
): string {
|
|
if (
|
|
trustedTypesPolicy &&
|
|
typeof trustedTypes === 'object' &&
|
|
typeof trustedTypes.getAttributeType === 'function' &&
|
|
!namespaceURI
|
|
) {
|
|
switch (trustedTypes.getAttributeType(lcTag, lcName)) {
|
|
case 'TrustedHTML': {
|
|
return _createTrustedHTML(value);
|
|
}
|
|
|
|
case 'TrustedScriptURL': {
|
|
return _createTrustedScriptURL(value);
|
|
}
|
|
|
|
default: {
|
|
break;
|
|
}
|
|
}
|
|
}
|
|
|
|
return value;
|
|
};
|
|
|
|
/**
|
|
* Write a modified attribute value back onto the element. On
|
|
* success, re-probe for clobbering introduced by the new value and
|
|
* remove the element when found; otherwise pop the removal entry
|
|
* recorded by the earlier _removeAttribute (long-standing pairing
|
|
* with the SANITIZE_NAMED_PROPS path - do not "fix" casually). On
|
|
* failure, remove the attribute instead.
|
|
*
|
|
* @param currentNode the element carrying the attribute
|
|
* @param name the attribute name as present on the element
|
|
* @param namespaceURI the attribute's namespace, if any
|
|
* @param value the new attribute value
|
|
*/
|
|
const _setAttributeValue = function (
|
|
currentNode: Element,
|
|
name: string,
|
|
namespaceURI: string | null,
|
|
value: string
|
|
): void {
|
|
try {
|
|
if (namespaceURI) {
|
|
currentNode.setAttributeNS(namespaceURI, name, value);
|
|
} else {
|
|
/* Fallback to setAttribute() for browser-unrecognized namespaces e.g. "x-schema". */
|
|
currentNode.setAttribute(name, value);
|
|
}
|
|
|
|
if (_isClobbered(currentNode)) {
|
|
_forceRemove(currentNode);
|
|
} else {
|
|
arrayPop(DOMPurify.removed);
|
|
}
|
|
} catch (_) {
|
|
_removeAttribute(name, currentNode);
|
|
}
|
|
};
|
|
|
|
/**
|
|
* _sanitizeAttributes
|
|
*
|
|
* @protect attributes
|
|
* @protect nodeName
|
|
* @protect removeAttribute
|
|
* @protect setAttribute
|
|
*
|
|
* @param currentNode to sanitize
|
|
*/
|
|
const _sanitizeAttributes = function (currentNode: Element): void {
|
|
/* Execute a hook if present */
|
|
_executeHooks(hooks.beforeSanitizeAttributes, currentNode, null);
|
|
|
|
const { attributes } = currentNode;
|
|
|
|
/* Check if we have attributes; if not we might have a text node */
|
|
if (!attributes || _isClobbered(currentNode)) {
|
|
return;
|
|
}
|
|
|
|
const hookEvent = {
|
|
attrName: '',
|
|
attrValue: '',
|
|
keepAttr: true,
|
|
allowedAttributes: ALLOWED_ATTR,
|
|
forceKeepAttr: undefined,
|
|
};
|
|
let l = attributes.length;
|
|
const lcTag = transformCaseFunc(currentNode.nodeName);
|
|
|
|
/* Go backwards over all attributes; safely remove bad ones */
|
|
while (l--) {
|
|
const attr = attributes[l];
|
|
const { name, namespaceURI, value: attrValue } = attr;
|
|
const lcName = transformCaseFunc(name);
|
|
|
|
const initValue = attrValue;
|
|
let value = name === 'value' ? initValue : stringTrim(initValue);
|
|
|
|
/* Execute a hook if present */
|
|
hookEvent.attrName = lcName;
|
|
hookEvent.attrValue = value;
|
|
hookEvent.keepAttr = true;
|
|
hookEvent.forceKeepAttr = undefined; // Allows developers to see this is a property they can set
|
|
_executeHooks(hooks.uponSanitizeAttribute, currentNode, hookEvent);
|
|
value = hookEvent.attrValue;
|
|
|
|
/* Full DOM Clobbering protection via namespace isolation,
|
|
* Prefix id and name attributes with `user-content-`
|
|
*/
|
|
if (
|
|
SANITIZE_NAMED_PROPS &&
|
|
(lcName === 'id' || lcName === 'name') &&
|
|
stringIndexOf(value, SANITIZE_NAMED_PROPS_PREFIX) !== 0
|
|
) {
|
|
// Remove the attribute with this value
|
|
_removeAttribute(name, currentNode);
|
|
// Prefix the value and later re-create the attribute with the sanitized value
|
|
value = SANITIZE_NAMED_PROPS_PREFIX + value;
|
|
}
|
|
// Else: already prefixed, leave the attribute alone — the prefix is
|
|
// itself the clobbering protection, and re-applying it is incorrect.
|
|
|
|
/* Work around a security issue with comments inside attributes */
|
|
if (
|
|
SAFE_FOR_XML &&
|
|
regExpTest(
|
|
/((--!?|])>)|<\/(style|script|title|xmp|textarea|noscript|iframe|noembed|noframes)/i,
|
|
value
|
|
)
|
|
) {
|
|
_removeAttribute(name, currentNode);
|
|
continue;
|
|
}
|
|
|
|
/* Make sure we cannot easily use animated hrefs, even if animations are allowed */
|
|
if (lcName === 'attributename' && stringMatch(value, 'href')) {
|
|
_removeAttribute(name, currentNode);
|
|
continue;
|
|
}
|
|
|
|
/* Did the hooks force-keep the attribute? */
|
|
if (hookEvent.forceKeepAttr) {
|
|
continue;
|
|
}
|
|
|
|
/* Did the hooks approve of the attribute? */
|
|
if (!hookEvent.keepAttr) {
|
|
_removeAttribute(name, currentNode);
|
|
continue;
|
|
}
|
|
|
|
/* Work around a security issue in jQuery 3.0 */
|
|
if (
|
|
!ALLOW_SELF_CLOSE_IN_ATTR &&
|
|
regExpTest(EXPRESSIONS.SELF_CLOSING_TAG, value)
|
|
) {
|
|
_removeAttribute(name, currentNode);
|
|
continue;
|
|
}
|
|
|
|
/* Sanitize attribute content to be template-safe */
|
|
if (SAFE_FOR_TEMPLATES) {
|
|
value = _stripTemplateExpressions(value);
|
|
}
|
|
|
|
/* Is `value` valid for this attribute? */
|
|
if (!_isValidAttribute(lcTag, lcName, value)) {
|
|
_removeAttribute(name, currentNode);
|
|
continue;
|
|
}
|
|
|
|
/* Handle attributes that require Trusted Types */
|
|
value = _applyTrustedTypesToAttribute(lcTag, lcName, namespaceURI, value);
|
|
|
|
/* Handle invalid data-* attribute set by try-catching it */
|
|
if (value !== initValue) {
|
|
_setAttributeValue(currentNode, name, namespaceURI, value);
|
|
}
|
|
}
|
|
|
|
/* Execute a hook if present */
|
|
_executeHooks(hooks.afterSanitizeAttributes, currentNode, null);
|
|
};
|
|
|
|
/**
|
|
* _sanitizeShadowDOM
|
|
*
|
|
* @param fragment to iterate over recursively
|
|
*/
|
|
const _sanitizeShadowDOM = function (fragment: DocumentFragment): void {
|
|
let shadowNode = null;
|
|
const shadowIterator = _createNodeIterator(fragment);
|
|
|
|
/* Execute a hook if present */
|
|
_executeHooks(hooks.beforeSanitizeShadowDOM, fragment, null);
|
|
|
|
while ((shadowNode = shadowIterator.nextNode())) {
|
|
/* Execute a hook if present */
|
|
_executeHooks(hooks.uponSanitizeShadowNode, shadowNode, null);
|
|
|
|
/* Sanitize tags and elements */
|
|
_sanitizeElements(shadowNode);
|
|
|
|
/* Check attributes next */
|
|
_sanitizeAttributes(shadowNode);
|
|
|
|
/* Deep shadow DOM detected.
|
|
Realm-safe check (GHSA-hpcv-96wg-7vj8): use nodeType against the
|
|
DOCUMENT_FRAGMENT_NODE constant rather than instanceof, so we
|
|
recurse into <template>.content from foreign realms too. */
|
|
if (_isDocumentFragment(shadowNode.content)) {
|
|
_sanitizeShadowDOM(shadowNode.content);
|
|
}
|
|
|
|
/* An element iterated here may itself host an attached
|
|
shadow root. The default NodeIterator does not enter shadow
|
|
trees, so a shadow root nested inside template.content was
|
|
previously reached by no walk at all (the pre-pass at
|
|
_sanitizeAttachedShadowRoots descends via childNodes, which
|
|
doesn't enter template.content; the template-content recursion
|
|
above iterates the content but never inspected shadowRoot).
|
|
Walk it explicitly. The nodeType guard avoids reading
|
|
shadowRoot off text / comment / CDATA / PI nodes that the
|
|
iterator also surfaces. */
|
|
const shadowNodeType = getNodeType
|
|
? getNodeType(shadowNode)
|
|
: shadowNode.nodeType;
|
|
if (shadowNodeType === NODE_TYPE.element) {
|
|
const innerSr = getShadowRoot(shadowNode);
|
|
if (_isDocumentFragment(innerSr)) {
|
|
_sanitizeAttachedShadowRoots(innerSr);
|
|
_sanitizeShadowDOM(innerSr);
|
|
}
|
|
}
|
|
}
|
|
|
|
/* Execute a hook if present */
|
|
_executeHooks(hooks.afterSanitizeShadowDOM, fragment, null);
|
|
};
|
|
|
|
/**
|
|
* _sanitizeAttachedShadowRoots
|
|
*
|
|
* Walks `root` and feeds every attached shadow root we encounter into
|
|
* the existing _sanitizeShadowDOM pipeline. The default node iterator
|
|
* does not descend into shadow trees, so nodes inside an attached
|
|
* shadow root would otherwise be skipped entirely.
|
|
*
|
|
* Two real input paths put attached shadow roots in front of us:
|
|
* 1. IN_PLACE on a DOM node that already has shadow roots attached.
|
|
* 2. DOM-node input where importNode(dirty, true) deep-clones the
|
|
* shadow root because it was created with `clonable: true`.
|
|
*
|
|
* This pass runs once, up front, so the main iteration loop (and the
|
|
* existing _sanitizeShadowDOM template-content recursion) stay
|
|
* untouched — string-input paths are not affected.
|
|
*
|
|
* @param root the subtree root to walk for attached shadow roots
|
|
*/
|
|
const _sanitizeAttachedShadowRoots = function (root: Node): void {
|
|
/* Iterative (explicit stack) rather than per-child recursion. DOM APIs
|
|
impose no depth cap, so an attacker-shaped tree (JSON/CRDT/editor data
|
|
built straight into the DOM — the IN_PLACE surface) deeper than the JS
|
|
call-stack budget would otherwise overflow native recursion here and
|
|
throw at the IN_PLACE entry pre-pass, before a single node is
|
|
sanitized, leaving the caller's live tree untouched (fail-open). See
|
|
campaign-3 F4. A heap stack keeps depth off the call stack.
|
|
|
|
Each work item is either a node to descend into, or a deferred
|
|
`_sanitizeShadowDOM` for an already-walked shadow root. The deferred
|
|
form preserves the original post-order discipline: a shadow root's
|
|
nested shadow roots are discovered before the outer shadow is
|
|
sanitized (which may remove hosts). Pushes are in reverse of the
|
|
desired processing order (LIFO): template content, then children, then
|
|
the shadow-sanitize, then the shadow walk — so the order matches the
|
|
previous recursion exactly. */
|
|
const stack: Array<{ node: Node | null; shadow: DocumentFragment | null }> =
|
|
[{ node: root, shadow: null }];
|
|
|
|
while (stack.length > 0) {
|
|
const item = stack.pop();
|
|
|
|
/* Deferred shadow-DOM sanitisation: runs after its subtree was walked. */
|
|
if (item.shadow) {
|
|
_sanitizeShadowDOM(item.shadow);
|
|
continue;
|
|
}
|
|
|
|
const node = item.node;
|
|
const nodeType = getNodeType ? getNodeType(node) : (node as any).nodeType;
|
|
const isElement = nodeType === NODE_TYPE.element;
|
|
|
|
/* (pushed last → processed first) Children, snapshotted in reverse so
|
|
the first child is processed first. Snapshotting matters because a
|
|
hook may detach siblings mid-walk. */
|
|
const childNodes = getChildNodes(node);
|
|
if (childNodes) {
|
|
for (let i = childNodes.length - 1; i >= 0; --i) {
|
|
stack.push({ node: childNodes[i], shadow: null });
|
|
}
|
|
}
|
|
|
|
/* (pushed before children → processed after them, matching the old
|
|
"template content last" order) When the node is a <template>,
|
|
descend into its content. */
|
|
if (isElement) {
|
|
const rootName = getNodeName ? getNodeName(node) : null;
|
|
if (
|
|
typeof rootName === 'string' &&
|
|
transformCaseFunc(rootName) === 'template'
|
|
) {
|
|
const content = (node as HTMLTemplateElement).content;
|
|
if (_isDocumentFragment(content)) {
|
|
stack.push({ node: content, shadow: null });
|
|
}
|
|
}
|
|
}
|
|
|
|
/* Shadow root (processed first): walk its subtree, then sanitise it.
|
|
Realm-safe check (GHSA-hpcv-96wg-7vj8): nodeType-based detection
|
|
rather than `instanceof DocumentFragment`, which is realm-bound and
|
|
silently skipped foreign-realm shadow roots (e.g.
|
|
iframe.contentDocument attachShadow). */
|
|
if (isElement) {
|
|
const sr = getShadowRoot(node);
|
|
if (_isDocumentFragment(sr)) {
|
|
/* Push the deferred sanitise first so it pops after the shadow
|
|
walk we push next, i.e. nested shadow roots are discovered
|
|
before this one is sanitised. */
|
|
stack.push({ node: null, shadow: sr }, { node: sr, shadow: null });
|
|
}
|
|
}
|
|
}
|
|
};
|
|
|
|
// eslint-disable-next-line complexity
|
|
DOMPurify.sanitize = function (dirty, cfg = {}) {
|
|
let body = null;
|
|
let importedNode = null;
|
|
let currentNode = null;
|
|
let returnNode = null;
|
|
/* Make sure we have a string to sanitize.
|
|
DO NOT return early, as this will return the wrong type if
|
|
the user has requested a DOM object rather than a string */
|
|
IS_EMPTY_INPUT = !dirty;
|
|
if (IS_EMPTY_INPUT) {
|
|
dirty = '<!-->';
|
|
}
|
|
|
|
/* Stringify, in case dirty is an object */
|
|
if (typeof dirty !== 'string' && !_isNode(dirty)) {
|
|
dirty = stringifyValue(dirty);
|
|
|
|
if (typeof dirty !== 'string') {
|
|
throw typeErrorCreate('dirty is not a string, aborting');
|
|
}
|
|
}
|
|
|
|
/* Return dirty HTML if DOMPurify cannot run */
|
|
if (!DOMPurify.isSupported) {
|
|
return dirty;
|
|
}
|
|
|
|
/* Assign config vars */
|
|
if (!SET_CONFIG) {
|
|
_parseConfig(cfg);
|
|
}
|
|
|
|
/* Clean up removed elements */
|
|
DOMPurify.removed = [];
|
|
|
|
/* Resolve IN_PLACE for this call without mutating persistent config.
|
|
Writing the IN_PLACE closure variable here leaks under setConfig(),
|
|
where _parseConfig is skipped on later calls: a single string call would
|
|
disable in-place mode for every subsequent node call, returning a
|
|
sanitized copy while leaving the caller's node — which in-place callers
|
|
keep using and whose return value they ignore — unsanitized. REPORT-2. */
|
|
const inPlace = IN_PLACE && typeof dirty !== 'string' && _isNode(dirty);
|
|
|
|
if (inPlace) {
|
|
/* Do some early pre-sanitization to avoid unsafe root nodes.
|
|
Read nodeName through the cached prototype getter — a clobbering
|
|
child named "nodeName" on the form root would otherwise shadow
|
|
the property and let this check skip the root-allowlist
|
|
validation entirely. */
|
|
const nn = getNodeName
|
|
? getNodeName(dirty as Node)
|
|
: (dirty as Node).nodeName;
|
|
if (typeof nn === 'string') {
|
|
const tagName = transformCaseFunc(nn);
|
|
if (!ALLOWED_TAGS[tagName] || FORBID_TAGS[tagName]) {
|
|
throw typeErrorCreate(
|
|
'root node is forbidden and cannot be sanitized in-place'
|
|
);
|
|
}
|
|
}
|
|
|
|
/* Pre-flight the root through _isClobbered. The iterator-driven
|
|
removal path can not detach a parent-less root: _forceRemove
|
|
falls through to Element.prototype.remove(), which per spec
|
|
is a no-op on a node with no parent. A clobbered root would
|
|
then survive the main loop with its attributes uninspected,
|
|
because _sanitizeAttributes early-returns on _isClobbered. The
|
|
result would be an attacker-controlled form, complete with any
|
|
event-handler attributes the caller passed in, handed back to
|
|
the application unsanitized. Refuse to sanitize such a root
|
|
the same way we refuse a forbidden tag. GHSA-r47g-fvhr-h676. */
|
|
if (_isClobbered(dirty as Element)) {
|
|
throw typeErrorCreate(
|
|
'root node is clobbered and cannot be sanitized in-place'
|
|
);
|
|
}
|
|
|
|
/* Sanitize attached shadow roots before the main iterator runs.
|
|
The iterator does not descend into shadow trees. Same fail-closed
|
|
barrier as the main walk (campaign-3 F2): a custom-element reaction
|
|
inside a shadow root could abort this pre-pass before the walk runs,
|
|
which would otherwise leave the entire live tree unsanitized. */
|
|
try {
|
|
_sanitizeAttachedShadowRoots(dirty as Node);
|
|
} catch (error) {
|
|
_neutralizeRoot(dirty as Node);
|
|
|
|
throw error;
|
|
}
|
|
} else if (_isNode(dirty)) {
|
|
/* If dirty is a DOM element, append to an empty document to avoid
|
|
elements being stripped by the parser */
|
|
body = _initDocument('<!---->');
|
|
importedNode = body.ownerDocument.importNode(dirty, true);
|
|
if (
|
|
importedNode.nodeType === NODE_TYPE.element &&
|
|
importedNode.nodeName === 'BODY'
|
|
) {
|
|
/* Node is already a body, use as is */
|
|
body = importedNode;
|
|
} else if (importedNode.nodeName === 'HTML') {
|
|
body = importedNode;
|
|
} else {
|
|
// eslint-disable-next-line unicorn/prefer-dom-node-append
|
|
body.appendChild(importedNode);
|
|
}
|
|
|
|
/* Clonable shadow roots are deep-cloned by importNode(); sanitize
|
|
them before the main iterator runs, since the iterator does not
|
|
descend into shadow trees. The walk routes every read through a
|
|
cached prototype getter so clobbering descendants on a form root
|
|
cannot hide a shadow host from this pass. */
|
|
_sanitizeAttachedShadowRoots(importedNode);
|
|
} else {
|
|
/* Exit directly if we have nothing to do */
|
|
if (
|
|
!RETURN_DOM &&
|
|
!SAFE_FOR_TEMPLATES &&
|
|
!WHOLE_DOCUMENT &&
|
|
// eslint-disable-next-line unicorn/prefer-includes
|
|
dirty.indexOf('<') === -1
|
|
) {
|
|
return trustedTypesPolicy && RETURN_TRUSTED_TYPE
|
|
? _createTrustedHTML(dirty)
|
|
: dirty;
|
|
}
|
|
|
|
/* Initialize the document to work on */
|
|
body = _initDocument(dirty);
|
|
|
|
/* Check we have a DOM node from the data */
|
|
if (!body) {
|
|
return RETURN_DOM ? null : RETURN_TRUSTED_TYPE ? emptyHTML : '';
|
|
}
|
|
}
|
|
|
|
/* Remove first element node (ours) if FORCE_BODY is set */
|
|
if (body && FORCE_BODY) {
|
|
_forceRemove(body.firstChild);
|
|
}
|
|
|
|
/* Get node iterator */
|
|
const nodeIterator = _createNodeIterator(inPlace ? dirty : body);
|
|
|
|
/* Now start iterating over the created document.
|
|
The walk runs inside an exception barrier (campaign-3 F2): a re-entrant
|
|
engine/custom-element mutation can detach a node mid-walk so
|
|
`_forceRemove`'s parentless guard throws, aborting the loop. Without the
|
|
barrier the caller's in-place tree would be left half-sanitized with the
|
|
unvisited tail still armed. On any throw we fail closed — strip the
|
|
in-place root bare — then rethrow so the existing throw contract is
|
|
preserved. (String/DOM-copy paths never return the partial body, so the
|
|
propagating throw is already fail-closed there.) */
|
|
try {
|
|
while ((currentNode = nodeIterator.nextNode())) {
|
|
/* Sanitize tags and elements */
|
|
_sanitizeElements(currentNode);
|
|
|
|
/* Check attributes next */
|
|
_sanitizeAttributes(currentNode);
|
|
|
|
/* Shadow DOM detected, sanitize it.
|
|
Realm-safe check (GHSA-hpcv-96wg-7vj8): nodeType-based detection
|
|
instead of instanceof, so foreign-realm <template>.content is
|
|
walked correctly. */
|
|
if (_isDocumentFragment(currentNode.content)) {
|
|
_sanitizeShadowDOM(currentNode.content);
|
|
}
|
|
}
|
|
} catch (error) {
|
|
if (inPlace) {
|
|
_neutralizeRoot(dirty as Node);
|
|
}
|
|
|
|
throw error;
|
|
}
|
|
|
|
/* If we sanitized `dirty` in-place, return it. */
|
|
if (inPlace) {
|
|
/* Fail-closed completion of the audit-5 F1 fix: every node removed from
|
|
the caller's live tree is detached but may still hold a queued
|
|
resource-event handler that fires in page scope after we return. The
|
|
move-hoist covers only disallowed-tag KEEP_CONTENT removals; strip the
|
|
non-allow-listed attributes off every other removed subtree (clobber,
|
|
mXSS, namespace, comments, KEEP_CONTENT:false, …) so those handlers are
|
|
cancelled before any event can fire. Runs synchronously, pre-return. */
|
|
arrayForEach(DOMPurify.removed, (entry) => {
|
|
if (entry.element) {
|
|
_neutralizeSubtree(entry.element as Node);
|
|
}
|
|
});
|
|
|
|
if (SAFE_FOR_TEMPLATES) {
|
|
_scrubTemplateExpressions(dirty as Element);
|
|
}
|
|
|
|
return dirty;
|
|
}
|
|
|
|
/* Return sanitized string or DOM */
|
|
if (RETURN_DOM) {
|
|
if (SAFE_FOR_TEMPLATES) {
|
|
_scrubTemplateExpressions(body);
|
|
}
|
|
|
|
if (RETURN_DOM_FRAGMENT) {
|
|
returnNode = createDocumentFragment.call(body.ownerDocument);
|
|
|
|
while (body.firstChild) {
|
|
// eslint-disable-next-line unicorn/prefer-dom-node-append
|
|
returnNode.appendChild(body.firstChild);
|
|
}
|
|
} else {
|
|
returnNode = body;
|
|
}
|
|
|
|
if (ALLOWED_ATTR.shadowroot || ALLOWED_ATTR.shadowrootmode) {
|
|
/*
|
|
AdoptNode() is not used because internal state is not reset
|
|
(e.g. the past names map of a HTMLFormElement), this is safe
|
|
in theory but we would rather not risk another attack vector.
|
|
The state that is cloned by importNode() is explicitly defined
|
|
by the specs.
|
|
*/
|
|
returnNode = importNode.call(originalDocument, returnNode, true);
|
|
}
|
|
|
|
return returnNode;
|
|
}
|
|
|
|
let serializedHTML = WHOLE_DOCUMENT ? body.outerHTML : body.innerHTML;
|
|
|
|
/* Serialize doctype if allowed */
|
|
if (
|
|
WHOLE_DOCUMENT &&
|
|
ALLOWED_TAGS['!doctype'] &&
|
|
body.ownerDocument &&
|
|
body.ownerDocument.doctype &&
|
|
body.ownerDocument.doctype.name &&
|
|
regExpTest(EXPRESSIONS.DOCTYPE_NAME, body.ownerDocument.doctype.name)
|
|
) {
|
|
serializedHTML =
|
|
'<!DOCTYPE ' + body.ownerDocument.doctype.name + '>\n' + serializedHTML;
|
|
}
|
|
|
|
/* Sanitize final string template-safe */
|
|
if (SAFE_FOR_TEMPLATES) {
|
|
serializedHTML = _stripTemplateExpressions(serializedHTML);
|
|
}
|
|
|
|
return trustedTypesPolicy && RETURN_TRUSTED_TYPE
|
|
? _createTrustedHTML(serializedHTML)
|
|
: serializedHTML;
|
|
};
|
|
|
|
DOMPurify.setConfig = function (cfg = {}) {
|
|
_parseConfig(cfg);
|
|
SET_CONFIG = true;
|
|
};
|
|
|
|
DOMPurify.clearConfig = function () {
|
|
CONFIG = null;
|
|
SET_CONFIG = false;
|
|
|
|
// Drop any caller-supplied Trusted Types policy so it cannot poison later
|
|
// `RETURN_TRUSTED_TYPE` output. The internal default policy (cached, and
|
|
// never recreated — Trusted Types throws on duplicate names) is restored by
|
|
// the next `_parseConfig`. See GHSA-vxr8-fq34-vvx9.
|
|
trustedTypesPolicy = defaultTrustedTypesPolicy;
|
|
emptyHTML = '';
|
|
};
|
|
|
|
DOMPurify.isValidAttribute = function (tag, attr, value) {
|
|
/* Initialize shared config vars if necessary. */
|
|
if (!CONFIG) {
|
|
_parseConfig({});
|
|
}
|
|
|
|
const lcTag = transformCaseFunc(tag);
|
|
const lcName = transformCaseFunc(attr);
|
|
return _isValidAttribute(lcTag, lcName, value);
|
|
};
|
|
|
|
DOMPurify.addHook = function (
|
|
entryPoint: keyof HooksMap,
|
|
hookFunction: HookFunction
|
|
) {
|
|
if (typeof hookFunction !== 'function') {
|
|
return;
|
|
}
|
|
|
|
arrayPush(hooks[entryPoint], hookFunction);
|
|
};
|
|
|
|
DOMPurify.removeHook = function (
|
|
entryPoint: keyof HooksMap,
|
|
hookFunction: HookFunction
|
|
) {
|
|
if (hookFunction !== undefined) {
|
|
const index = arrayLastIndexOf(hooks[entryPoint], hookFunction);
|
|
|
|
return index === -1
|
|
? undefined
|
|
: arraySplice(hooks[entryPoint], index, 1)[0];
|
|
}
|
|
|
|
return arrayPop(hooks[entryPoint]);
|
|
};
|
|
|
|
DOMPurify.removeHooks = function (entryPoint: keyof HooksMap) {
|
|
hooks[entryPoint] = [];
|
|
};
|
|
|
|
DOMPurify.removeAllHooks = function () {
|
|
hooks = _createHooksMap();
|
|
};
|
|
|
|
return DOMPurify;
|
|
}
|
|
|
|
export default createDOMPurify();
|