meteor/packages/html-tools/parse.js

// Parse a "fragment" of HTML, up to the end of the input or a particular
// template tag (using the "shouldStop" option).
HTMLTools.parseFragment = function (input, options) {
  var scanner;
  if (typeof input === 'string')
    scanner = new Scanner(input);
  else
    // input can be a scanner.  We'd better not have a different
    // value for the "getTemplateTag" option as when the scanner
    // was created, because we don't do anything special to reset
    // the value (which is attached to the scanner).
    scanner = input;

  // ```
  // { getTemplateTag: function (scanner, templateTagPosition) {
  //     if (templateTagPosition === HTMLTools.TEMPLATE_TAG_POSITION.ELEMENT) {
  //       ...
  // ```
  if (options && options.getTemplateTag)
    scanner.getTemplateTag = options.getTemplateTag;

  // function (scanner) -> boolean
  var shouldStop = options && options.shouldStop;

  var result;
  if (options && options.textMode) {
    if (options.textMode === HTML.TEXTMODE.STRING) {
      result = getRawText(scanner, null, shouldStop);
    } else if (options.textMode === HTML.TEXTMODE.RCDATA) {
      result = getRCData(scanner, null, shouldStop);
    } else {
      throw new Error("Unsupported textMode: " + options.textMode);
    }
  } else {
    result = getContent(scanner, shouldStop);
  }
  if (! scanner.isEOF()) {
    // If we aren't at the end of the input, we either stopped at an unmatched
    // HTML end tag or at a template tag (like `{{else}}` or `{{/if}}`).
    // Detect the former case (stopped at an HTML end tag) and throw a good
    // error.

    var posBefore = scanner.pos;

    try {
      var endTag = getHTMLToken(scanner);
    } catch (e) {
      // ignore errors from getTemplateTag
    }

    // XXX we make some assumptions about shouldStop here, like that it
    // won't tell us to stop at an HTML end tag.  Should refactor
    // `shouldStop` into something more suitable.
    if (endTag && endTag.t === 'Tag' && endTag.isEnd) {
      var closeTag = endTag.n;
      var isVoidElement = HTML.isVoidElement(closeTag);
      scanner.fatal("Unexpected HTML close tag" +
                    (isVoidElement ?
                     '.  <' + endTag.n + '> should have no close tag.' : ''));
    }

    scanner.pos = posBefore; // rewind, we'll continue parsing as usual

    // If no "shouldStop" option was provided, we should have consumed the whole
    // input.
    if (! shouldStop)
      scanner.fatal("Expected EOF");
  }

  return result;
};

// Take a numeric Unicode code point, which may be larger than 16 bits,
// and encode it as a JavaScript UTF-16 string.
//
// Adapted from
// http://stackoverflow.com/questions/7126384/expressing-utf-16-unicode-characters-in-javascript/7126661.
codePointToString = HTMLTools.codePointToString = function(cp) {
  if (cp >= 0 && cp <= 0xD7FF || cp >= 0xE000 && cp <= 0xFFFF) {
    return String.fromCharCode(cp);
  } else if (cp >= 0x10000 && cp <= 0x10FFFF) {

    // we substract 0x10000 from cp to get a 20-bit number
    // in the range 0..0xFFFF
    cp -= 0x10000;

    // we add 0xD800 to the number formed by the first 10 bits
    // to give the first byte
    var first = ((0xffc00 & cp) >> 10) + 0xD800;

    // we add 0xDC00 to the number formed by the low 10 bits
    // to give the second byte
    var second = (0x3ff & cp) + 0xDC00;

    return String.fromCharCode(first) + String.fromCharCode(second);
  } else {
    return '';
  }
};

getContent = HTMLTools.Parse.getContent = function (scanner, shouldStopFunc) {
  var items = [];

  while (! scanner.isEOF()) {
    if (shouldStopFunc && shouldStopFunc(scanner))
      break;

    var posBefore = scanner.pos;
    var token = getHTMLToken(scanner);
    if (! token)
      // tokenizer reached EOF on its own, e.g. while scanning
      // template comments like `{{! foo}}`.
      continue;

    if (token.t === 'Doctype') {
      scanner.fatal("Unexpected Doctype");
    } else if (token.t === 'Chars') {
      pushOrAppendString(items, token.v);
    } else if (token.t === 'CharRef') {
      items.push(convertCharRef(token));
    } else if (token.t === 'Comment') {
      items.push(HTML.Comment(token.v));
    } else if (token.t === 'TemplateTag') {
      items.push(token.v);
    } else if (token.t === 'Tag') {
      if (token.isEnd) {
        // Stop when we encounter an end tag at the top level.
        // Rewind; we'll re-parse the end tag later.
        scanner.pos = posBefore;
        break;
      }

      var tagName = token.n;
      // is this an element with no close tag (a BR, HR, IMG, etc.) based
      // on its name?
      var isVoid = HTML.isVoidElement(tagName);
      if (token.isSelfClosing) {
        if (! (isVoid || HTML.isKnownSVGElement(tagName) || tagName.indexOf(':') >= 0))
          scanner.fatal('Only certain elements like BR, HR, IMG, etc. (and foreign elements like SVG) are allowed to self-close');
      }

      // may be null
      var attrs = parseAttrs(token.attrs);

      var tagFunc = HTML.getTag(tagName);
      if (isVoid || token.isSelfClosing) {
        items.push(attrs ? tagFunc(attrs) : tagFunc());
      } else {
        // parse HTML tag contents.

        // HTML treats a final `/` in a tag as part of an attribute, as in `<a href=/foo/>`, but the template author who writes `<circle r={{r}}/>`, say, may not be thinking about that, so generate a good error message in the "looks like self-close" case.
        var looksLikeSelfClose = (scanner.input.substr(scanner.pos - 2, 2) === '/>');

        var content;
        if (token.n === 'textarea') {
          if (scanner.peek() === '\n')
            scanner.pos++;
          attrs = (attrs || {});
          var textareaValue = getRCData(scanner, token.n, shouldStopFunc);
          if (HTML.isArray(attrs)) {
            attrs.push({value: textareaValue});
          } else {
            attrs.value = textareaValue;
          }
        } else {
          content = getContent(scanner, shouldStopFunc);
        }

        var endTag = getHTMLToken(scanner);

        if (! (endTag && endTag.t === 'Tag' && endTag.isEnd && endTag.n === tagName))
          scanner.fatal('Expected "' + tagName + '" end tag' + (looksLikeSelfClose ? ' -- if the "<' + token.n + ' />" tag was supposed to self-close, try adding a space before the "/"' : ''));

        // XXX support implied end tags in cases allowed by the spec

        // make `content` into an array suitable for applying tag constructor
        // as in `FOO.apply(null, content)`.
        if (content == null)
          content = [];
        else if (! (content instanceof Array))
          content = [content];

        items.push(HTML.getTag(tagName).apply(
          null, (attrs ? [attrs] : []).concat(content)));
      }
    } else {
      scanner.fatal("Unknown token type: " + token.t);
    }
  }

  if (items.length === 0)
    return null;
  else if (items.length === 1)
    return items[0];
  else
    return items;
};

var pushOrAppendString = function (items, string) {
  if (items.length &&
      typeof items[items.length - 1] === 'string')
    items[items.length - 1] += string;
  else
    items.push(string);
};

// get RCDATA to go in the lowercase (or camel case) tagName (e.g. "textarea")
getRCData = HTMLTools.Parse.getRCData = function (scanner, tagName, shouldStopFunc) {
  var items = [];

  while (! scanner.isEOF()) {
    // break at appropriate end tag
    if (tagName && isLookingAtEndTag(scanner, tagName))
      break;

    if (shouldStopFunc && shouldStopFunc(scanner))
      break;

    var token = getHTMLToken(scanner, 'rcdata');
    if (! token)
      // tokenizer reached EOF on its own, e.g. while scanning
      // template comments like `{{! foo}}`.
      continue;

    if (token.t === 'Chars') {
      pushOrAppendString(items, token.v);
    } else if (token.t === 'CharRef') {
      items.push(convertCharRef(token));
    } else if (token.t === 'TemplateTag') {
      items.push(token.v);
    } else {
      // (can't happen)
      scanner.fatal("Unknown or unexpected token type: " + token.t);
    }
  }

  if (items.length === 0)
    return null;
  else if (items.length === 1)
    return items[0];
  else
    return items;
};

var getRawText = function (scanner, tagName, shouldStopFunc) {
  var items = [];

  while (! scanner.isEOF()) {
    // break at appropriate end tag
    if (tagName && isLookingAtEndTag(scanner, tagName))
      break;

    if (shouldStopFunc && shouldStopFunc(scanner))
      break;

    var token = getHTMLToken(scanner, 'rawtext');
    if (! token)
      // tokenizer reached EOF on its own, e.g. while scanning
      // template comments like `{{! foo}}`.
      continue;

    if (token.t === 'Chars') {
      pushOrAppendString(items, token.v);
    } else if (token.t === 'TemplateTag') {
      items.push(token.v);
    } else {
      // (can't happen)
      scanner.fatal("Unknown or unexpected token type: " + token.t);
    }
  }

  if (items.length === 0)
    return null;
  else if (items.length === 1)
    return items[0];
  else
    return items;
};

// Input: A token like `{ t: 'CharRef', v: '&amp;', cp: [38] }`.
//
// Output: A tag like `HTML.CharRef({ html: '&amp;', str: '&' })`.
var convertCharRef = function (token) {
  var codePoints = token.cp;
  var str = '';
  for (var i = 0; i < codePoints.length; i++)
    str += codePointToString(codePoints[i]);
  return HTML.CharRef({ html: token.v, str: str });
};

// Input is always a dictionary (even if zero attributes) and each
// value in the dictionary is an array of `Chars`, `CharRef`,
// and maybe `TemplateTag` tokens.
//
// Output is null if there are zero attributes, and otherwise a
// dictionary, or an array of dictionaries and template tags.
// Each value in the dictionary is HTMLjs (e.g. a
// string or an array of `Chars`, `CharRef`, and `TemplateTag`
// nodes).
//
// An attribute value with no input tokens is represented as "",
// not an empty array, in order to prop open empty attributes
// with no template tags.
var parseAttrs = function (attrs) {
  var result = null;

  if (HTML.isArray(attrs)) {
    // first element is nondynamic attrs, rest are template tags
    var nondynamicAttrs = parseAttrs(attrs[0]);
    if (nondynamicAttrs) {
      result = (result || []);
      result.push(nondynamicAttrs);
    }
    for (var i = 1; i < attrs.length; i++) {
      var token = attrs[i];
      if (token.t !== 'TemplateTag')
        throw new Error("Expected TemplateTag token");
      result = (result || []);
      result.push(token.v);
    }
    return HTML.Attrs.apply(null, result);
  }

  for (var k in attrs) {
    if (! result)
      result = {};

    var inValue = attrs[k];
    var outParts = [];
    for (var i = 0; i < inValue.length; i++) {
      var token = inValue[i];
      if (token.t === 'CharRef') {
        outParts.push(convertCharRef(token));
      } else if (token.t === 'TemplateTag') {
        outParts.push(token.v);
      } else if (token.t === 'Chars') {
        pushOrAppendString(outParts, token.v);
      }
    }

    var outValue = (inValue.length === 0 ? '' :
                    (outParts.length === 1 ? outParts[0] : outParts));
    var properKey = HTMLTools.properCaseAttributeName(k);
    result[properKey] = outValue;
  }

  return result;
};