"use strict";
Object.defineProperty(exports, "__esModule", { value: true });
var tslib_1 = require("tslib");
var regex_lib_1 = require("../regex-lib");
var utils_1 = require("../utils");
// For debugging: search for other "For debugging" lines
// import CliTable from 'cli-table';
/**
* Parses an HTML string, calling the callbacks to notify of tags and text.
*
* ## History
*
* This file previously used a regular expression to find html tags in the input
* text. Unfortunately, we ran into a bunch of catastrophic backtracking issues
* with certain input text, causing Autolinker to either hang or just take a
* really long time to parse the string.
*
* The current code is intended to be a O(n) algorithm that walks through
* the string in one pass, and tries to be as cheap as possible. We don't need
* to implement the full HTML spec, but rather simply determine where the string
* looks like an HTML tag, and where it looks like text (so that we can autolink
* that).
*
* This state machine parser is intended just to be a simple but performant
* parser of HTML for the subset of requirements we have. We simply need to:
*
* 1. Determine where HTML tags are
* 2. Determine the tag name (Autolinker specifically only cares about <a>,
* <script>, and <style> tags, so as not to link any text within them)
*
* We don't need to:
*
* 1. Create a parse tree
* 2. Auto-close tags with invalid markup
* 3. etc.
*
* The other intention behind this is that we didn't want to add external
* dependencies on the Autolinker utility which would increase its size. For
* instance, adding htmlparser2 adds 125kb to the minified output file,
* increasing its final size from 47kb to 172kb (at the time of writing). It
* also doesn't work exactly correctly, treating the string "<3 blah blah blah"
* as an HTML tag.
*
* Reference for HTML spec:
*
* https://www.w3.org/TR/html51/syntax.html#sec-tokenization
*
* @param {String} html The HTML to parse
* @param {Object} callbacks
* @param {Function} callbacks.onOpenTag Callback function to call when an open
* tag is parsed. Called with the tagName as its argument.
* @param {Function} callbacks.onCloseTag Callback function to call when a close
* tag is parsed. Called with the tagName as its argument. If a self-closing
* tag is found, `onCloseTag` is called immediately after `onOpenTag`.
* @param {Function} callbacks.onText Callback function to call when text (i.e
* not an HTML tag) is parsed. Called with the text (string) as its first
* argument, and offset (number) into the string as its second.
*/
function parseHtml(html, _a) {
var onOpenTag = _a.onOpenTag, onCloseTag = _a.onCloseTag, onText = _a.onText, onComment = _a.onComment, onDoctype = _a.onDoctype;
var noCurrentTag = new CurrentTag();
var charIdx = 0, len = html.length, state = 0 /* Data */, currentDataIdx = 0, // where the current data start index is
currentTag = noCurrentTag; // describes the current tag that is being read
// For debugging: search for other "For debugging" lines
// const table = new CliTable( {
// head: [ 'charIdx', 'char', 'state', 'currentDataIdx', 'currentOpenTagIdx', 'tag.type' ]
// } );
while (charIdx < len) {
var char = html.charAt(charIdx);
// For debugging: search for other "For debugging" lines
// ALSO: Temporarily remove the 'const' keyword on the State enum
// table.push(
// [ charIdx, char, State[ state ], currentDataIdx, currentTag.idx, currentTag.idx === -1 ? '' : currentTag.type ]
// );
switch (state) {
case 0 /* Data */:
stateData(char);
break;
case 1 /* TagOpen */:
stateTagOpen(char);
break;
case 2 /* EndTagOpen */:
stateEndTagOpen(char);
break;
case 3 /* TagName */:
stateTagName(char);
break;
case 4 /* BeforeAttributeName */:
stateBeforeAttributeName(char);
break;
case 5 /* AttributeName */:
stateAttributeName(char);
break;
case 6 /* AfterAttributeName */:
stateAfterAttributeName(char);
break;
case 7 /* BeforeAttributeValue */:
stateBeforeAttributeValue(char);
break;
case 8 /* AttributeValueDoubleQuoted */:
stateAttributeValueDoubleQuoted(char);
break;
case 9 /* AttributeValueSingleQuoted */:
stateAttributeValueSingleQuoted(char);
break;
case 10 /* AttributeValueUnquoted */:
stateAttributeValueUnquoted(char);
break;
case 11 /* AfterAttributeValueQuoted */:
stateAfterAttributeValueQuoted(char);
break;
case 12 /* SelfClosingStartTag */:
stateSelfClosingStartTag(char);
break;
case 13 /* MarkupDeclarationOpenState */:
stateMarkupDeclarationOpen(char);
break;
case 14 /* CommentStart */:
stateCommentStart(char);
break;
case 15 /* CommentStartDash */:
stateCommentStartDash(char);
break;
case 16 /* Comment */:
stateComment(char);
break;
case 17 /* CommentEndDash */:
stateCommentEndDash(char);
break;
case 18 /* CommentEnd */:
stateCommentEnd(char);
break;
case 19 /* CommentEndBang */:
stateCommentEndBang(char);
break;
case 20 /* Doctype */:
stateDoctype(char);
break;
default:
utils_1.throwUnhandledCaseError(state);
}
// For debugging: search for other "For debugging" lines
// ALSO: Temporarily remove the 'const' keyword on the State enum
// table.push(
// [ charIdx, char, State[ state ], currentDataIdx, currentTag.idx, currentTag.idx === -1 ? '' : currentTag.type ]
// );
charIdx++;
}
if (currentDataIdx < charIdx) {
emitText();
}
// For debugging: search for other "For debugging" lines
// console.log( '\n' + table.toString() );
// Called when non-tags are being read (i.e. the text around HTML †ags)
// https://www.w3.org/TR/html51/syntax.html#data-state
function stateData(char) {
if (char === '<') {
startNewTag();
}
}
// Called after a '<' is read from the Data state
// https://www.w3.org/TR/html51/syntax.html#tag-open-state
function stateTagOpen(char) {
if (char === '!') {
state = 13 /* MarkupDeclarationOpenState */;
}
else if (char === '/') {
state = 2 /* EndTagOpen */;
currentTag = new CurrentTag(tslib_1.__assign({}, currentTag, { isClosing: true }));
}
else if (char === '<') {
// start of another tag (ignore the previous, incomplete one)
startNewTag();
}
else if (regex_lib_1.letterRe.test(char)) {
// tag name start (and no '/' read)
state = 3 /* TagName */;
currentTag = new CurrentTag(tslib_1.__assign({}, currentTag, { isOpening: true }));
}
else {
// Any other
state = 0 /* Data */;
currentTag = noCurrentTag;
}
}
// After a '<x', '</x' sequence is read (where 'x' is a letter character),
// this is to continue reading the tag name
// https://www.w3.org/TR/html51/syntax.html#tag-name-state
function stateTagName(char) {
if (regex_lib_1.whitespaceRe.test(char)) {
currentTag = new CurrentTag(tslib_1.__assign({}, currentTag, { name: captureTagName() }));
state = 4 /* BeforeAttributeName */;
}
else if (char === '<') {
// start of another tag (ignore the previous, incomplete one)
startNewTag();
}
else if (char === '/') {
currentTag = new CurrentTag(tslib_1.__assign({}, currentTag, { name: captureTagName() }));
state = 12 /* SelfClosingStartTag */;
}
else if (char === '>') {
currentTag = new CurrentTag(tslib_1.__assign({}, currentTag, { name: captureTagName() }));
emitTagAndPreviousTextNode(); // resets to Data state as well
}
else if (!regex_lib_1.letterRe.test(char) && !regex_lib_1.digitRe.test(char) && char !== ':') {
// Anything else that does not form an html tag. Note: the colon
// character is accepted for XML namespaced tags
resetToDataState();
}
else {
// continue reading tag name
}
}
// Called after the '/' is read from a '</' sequence
// https://www.w3.org/TR/html51/syntax.html#end-tag-open-state
function stateEndTagOpen(char) {
if (char === '>') { // parse error. Encountered "</>". Skip it without treating as a tag
resetToDataState();
}
else if (regex_lib_1.letterRe.test(char)) {
state = 3 /* TagName */;
}
else {
// some other non-tag-like character, don't treat this as a tag
resetToDataState();
}
}
// https://www.w3.org/TR/html51/syntax.html#before-attribute-name-state
function stateBeforeAttributeName(char) {
if (regex_lib_1.whitespaceRe.test(char)) {
// stay in BeforeAttributeName state - continue reading chars
}
else if (char === '/') {
state = 12 /* SelfClosingStartTag */;
}
else if (char === '>') {
emitTagAndPreviousTextNode(); // resets to Data state as well
}
else if (char === '<') {
// start of another tag (ignore the previous, incomplete one)
startNewTag();
}
else if (char === "=" || regex_lib_1.quoteRe.test(char) || regex_lib_1.controlCharsRe.test(char)) {
// "Parse error" characters that, according to the spec, should be
// appended to the attribute name, but we'll treat these characters
// as not forming a real HTML tag
resetToDataState();
}
else {
// Any other char, start of a new attribute name
state = 5 /* AttributeName */;
}
}
// https://www.w3.org/TR/html51/syntax.html#attribute-name-state
function stateAttributeName(char) {
if (regex_lib_1.whitespaceRe.test(char)) {
state = 6 /* AfterAttributeName */;
}
else if (char === '/') {
state = 12 /* SelfClosingStartTag */;
}
else if (char === '=') {
state = 7 /* BeforeAttributeValue */;
}
else if (char === '>') {
emitTagAndPreviousTextNode(); // resets to Data state as well
}
else if (char === '<') {
// start of another tag (ignore the previous, incomplete one)
startNewTag();
}
else if (regex_lib_1.quoteRe.test(char)) {
// "Parse error" characters that, according to the spec, should be
// appended to the attribute name, but we'll treat these characters
// as not forming a real HTML tag
resetToDataState();
}
else {
// anything else: continue reading attribute name
}
}
// https://www.w3.org/TR/html51/syntax.html#after-attribute-name-state
function stateAfterAttributeName(char) {
if (regex_lib_1.whitespaceRe.test(char)) {
// ignore the character - continue reading
}
else if (char === '/') {
state = 12 /* SelfClosingStartTag */;
}
else if (char === '=') {
state = 7 /* BeforeAttributeValue */;
}
else if (char === '>') {
emitTagAndPreviousTextNode();
}
else if (char === '<') {
// start of another tag (ignore the previous, incomplete one)
startNewTag();
}
else if (regex_lib_1.quoteRe.test(char)) {
// "Parse error" characters that, according to the spec, should be
// appended to the attribute name, but we'll treat these characters
// as not forming a real HTML tag
resetToDataState();
}
else {
// Any other character, start a new attribute in the current tag
state = 5 /* AttributeName */;
}
}
// https://www.w3.org/TR/html51/syntax.html#before-attribute-value-state
function stateBeforeAttributeValue(char) {
if (regex_lib_1.whitespaceRe.test(char)) {
// ignore the character - continue reading
}
else if (char === "\"") {
state = 8 /* AttributeValueDoubleQuoted */;
}
else if (char === "'") {
state = 9 /* AttributeValueSingleQuoted */;
}
else if (/[>=`]/.test(char)) {
// Invalid chars after an '=' for an attribute value, don't count
// the current tag as an HTML tag
resetToDataState();
}
else if (char === '<') {
// start of another tag (ignore the previous, incomplete one)
startNewTag();
}
else {
// Any other character, consider it an unquoted attribute value
state = 10 /* AttributeValueUnquoted */;
}
}
// https://www.w3.org/TR/html51/syntax.html#attribute-value-double-quoted-state
function stateAttributeValueDoubleQuoted(char) {
if (char === "\"") { // end the current double-quoted attribute
state = 11 /* AfterAttributeValueQuoted */;
}
else {
// consume the character as part of the double-quoted attribute value
}
}
// https://www.w3.org/TR/html51/syntax.html#attribute-value-single-quoted-state
function stateAttributeValueSingleQuoted(char) {
if (char === "'") { // end the current single-quoted attribute
state = 11 /* AfterAttributeValueQuoted */;
}
else {
// consume the character as part of the double-quoted attribute value
}
}
// https://www.w3.org/TR/html51/syntax.html#attribute-value-unquoted-state
function stateAttributeValueUnquoted(char) {
if (regex_lib_1.whitespaceRe.test(char)) {
state = 4 /* BeforeAttributeName */;
}
else if (char === '>') {
emitTagAndPreviousTextNode();
}
else if (char === '<') {
// start of another tag (ignore the previous, incomplete one)
startNewTag();
}
else {
// Any other character, treat it as part of the attribute value
}
}
// https://www.w3.org/TR/html51/syntax.html#after-attribute-value-quoted-state
function stateAfterAttributeValueQuoted(char) {
if (regex_lib_1.whitespaceRe.test(char)) {
state = 4 /* BeforeAttributeName */;
}
else if (char === '/') {
state = 12 /* SelfClosingStartTag */;
}
else if (char === '>') {
emitTagAndPreviousTextNode();
}
else if (char === '<') {
// start of another tag (ignore the previous, incomplete one)
startNewTag();
}
else {
// Any other character, "parse error". Spec says to switch to the
// BeforeAttributeState and re-consume the character, as it may be
// the start of a new attribute name
state = 4 /* BeforeAttributeName */;
reconsumeCurrentCharacter();
}
}
// A '/' has just been read in the current tag (presumably for '/>'), and
// this handles the next character
// https://www.w3.org/TR/html51/syntax.html#self-closing-start-tag-state
function stateSelfClosingStartTag(char) {
if (char === '>') {
currentTag = new CurrentTag(tslib_1.__assign({}, currentTag, { isClosing: true }));
emitTagAndPreviousTextNode(); // resets to Data state as well
}
else {
state = 4 /* BeforeAttributeName */;
}
}
// https://www.w3.org/TR/html51/syntax.html#markup-declaration-open-state
// (HTML Comments or !DOCTYPE)
function stateMarkupDeclarationOpen(char) {
if (html.substr(charIdx, 2) === '--') { // html comment
charIdx += 2; // "consume" characters
currentTag = new CurrentTag(tslib_1.__assign({}, currentTag, { type: 'comment' }));
state = 14 /* CommentStart */;
}
else if (html.substr(charIdx, 7).toUpperCase() === 'DOCTYPE') {
charIdx += 7; // "consume" characters
currentTag = new CurrentTag(tslib_1.__assign({}, currentTag, { type: 'doctype' }));
state = 20 /* Doctype */;
}
else {
// At this point, the spec specifies that the state machine should
// enter the "bogus comment" state, in which case any character(s)
// after the '<!' that were read should become an HTML comment up
// until the first '>' that is read (or EOF). Instead, we'll assume
// that a user just typed '<!' as part of text data
resetToDataState();
}
}
// Handles after the sequence '<!--' has been read
// https://www.w3.org/TR/html51/syntax.html#comment-start-state
function stateCommentStart(char) {
if (char === '-') {
// We've read the sequence '<!---' at this point (3 dashes)
state = 15 /* CommentStartDash */;
}
else if (char === '>') {
// At this point, we'll assume the comment wasn't a real comment
// so we'll just emit it as data. We basically read the sequence
// '<!-->'
resetToDataState();
}
else {
// Any other char, take it as part of the comment
state = 16 /* Comment */;
}
}
// We've read the sequence '<!---' at this point (3 dashes)
// https://www.w3.org/TR/html51/syntax.html#comment-start-dash-state
function stateCommentStartDash(char) {
if (char === '-') {
// We've read '<!----' (4 dashes) at this point
state = 18 /* CommentEnd */;
}
else if (char === '>') {
// At this point, we'll assume the comment wasn't a real comment
// so we'll just emit it as data. We basically read the sequence
// '<!--->'
resetToDataState();
}
else {
// Anything else, take it as a valid comment
state = 16 /* Comment */;
}
}
// Currently reading the comment's text (data)
// https://www.w3.org/TR/html51/syntax.html#comment-state
function stateComment(char) {
if (char === '-') {
state = 17 /* CommentEndDash */;
}
else {
// Any other character, stay in the Comment state
}
}
// When we we've read the first dash inside a comment, it may signal the
// end of the comment if we read another dash
// https://www.w3.org/TR/html51/syntax.html#comment-end-dash-state
function stateCommentEndDash(char) {
if (char === '-') {
state = 18 /* CommentEnd */;
}
else {
// Wasn't a dash, must still be part of the comment
state = 16 /* Comment */;
}
}
// After we've read two dashes inside a comment, it may signal the end of
// the comment if we then read a '>' char
// https://www.w3.org/TR/html51/syntax.html#comment-end-state
function stateCommentEnd(char) {
if (char === '>') {
emitTagAndPreviousTextNode();
}
else if (char === '!') {
state = 19 /* CommentEndBang */;
}
else if (char === '-') {
// A 3rd '-' has been read: stay in the CommentEnd state
}
else {
// Anything else, switch back to the comment state since we didn't
// read the full "end comment" sequence (i.e. '-->')
state = 16 /* Comment */;
}
}
// We've read the sequence '--!' inside of a comment
// https://www.w3.org/TR/html51/syntax.html#comment-end-bang-state
function stateCommentEndBang(char) {
if (char === '-') {
// We read the sequence '--!-' inside of a comment. The last dash
// could signify that the comment is going to close
state = 17 /* CommentEndDash */;
}
else if (char === '>') {
// End of comment with the sequence '--!>'
emitTagAndPreviousTextNode();
}
else {
// The '--!' was not followed by a '>', continue reading the
// comment's text
state = 16 /* Comment */;
}
}
/**
* For DOCTYPES in particular, we don't care about the attributes. Just
* advance to the '>' character and emit the tag, unless we find a '<'
* character in which case we'll start a new tag.
*
* Example doctype tag:
* <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN" "http://www.w3.org/TR/html4/strict.dtd">
*
* Actual spec: https://www.w3.org/TR/html51/syntax.html#doctype-state
*/
function stateDoctype(char) {
if (char === '>') {
emitTagAndPreviousTextNode();
}
else if (char === '<') {
startNewTag();
}
else {
// stay in the Doctype state
}
}
/**
* Resets the state back to the Data state, and removes the current tag.
*
* We'll generally run this function whenever a "parse error" is
* encountered, where the current tag that is being read no longer looks
* like a real HTML tag.
*/
function resetToDataState() {
state = 0 /* Data */;
currentTag = noCurrentTag;
}
/**
* Starts a new HTML tag at the current index, ignoring any previous HTML
* tag that was being read.
*
* We'll generally run this function whenever we read a new '<' character,
* including when we read a '<' character inside of an HTML tag that we were
* previously reading.
*/
function startNewTag() {
state = 1 /* TagOpen */;
currentTag = new CurrentTag({ idx: charIdx });
}
/**
* Once we've decided to emit an open tag, that means we can also emit the
* text node before it.
*/
function emitTagAndPreviousTextNode() {
var textBeforeTag = html.slice(currentDataIdx, currentTag.idx);
if (textBeforeTag) {
// the html tag was the first element in the html string, or two
// tags next to each other, in which case we should not emit a text
// node
onText(textBeforeTag, currentDataIdx);
}
if (currentTag.type === 'comment') {
onComment(currentTag.idx);
}
else if (currentTag.type === 'doctype') {
onDoctype(currentTag.idx);
}
else {
if (currentTag.isOpening) {
onOpenTag(currentTag.name, currentTag.idx);
}
if (currentTag.isClosing) { // note: self-closing tags will emit both opening and closing
onCloseTag(currentTag.name, currentTag.idx);
}
}
// Since we just emitted a tag, reset to the data state for the next char
resetToDataState();
currentDataIdx = charIdx + 1;
}
function emitText() {
var text = html.slice(currentDataIdx, charIdx);
onText(text, currentDataIdx);
currentDataIdx = charIdx + 1;
}
/**
* Captures the tag name from the start of the tag to the current character
* index, and converts it to lower case
*/
function captureTagName() {
var startIdx = currentTag.idx + (currentTag.isClosing ? 2 : 1);
return html.slice(startIdx, charIdx).toLowerCase();
}
/**
* Causes the main loop to re-consume the current character, such as after
* encountering a "parse error" that changed state and needs to reconsume
* the same character in that new state.
*/
function reconsumeCurrentCharacter() {
charIdx--;
}
}
exports.parseHtml = parseHtml;
var CurrentTag = (function () {
function CurrentTag(cfg) {
if (cfg === void 0) { cfg = {}; }
this.idx = cfg.idx !== undefined ? cfg.idx : -1;
this.type = cfg.type || 'tag';
this.name = cfg.name || '';
this.isOpening = !!cfg.isOpening;
this.isClosing = !!cfg.isClosing;
}
return CurrentTag;
}());
//# sourceMappingURL=parse-html.js.map