preprocessor.js 4.01 KB
'use strict';

var UNICODE = require('../common/unicode');

//Aliases
var $ = UNICODE.CODE_POINTS;

//Utils

//OPTIMIZATION: these utility functions should not be moved out of this module. V8 Crankshaft will not inline
//this functions if they will be situated in another module due to context switch.
//Always perform inlining check before modifying this functions ('node --trace-inlining').
function isSurrogatePair(cp1, cp2) {
    return cp1 >= 0xD800 && cp1 <= 0xDBFF && cp2 >= 0xDC00 && cp2 <= 0xDFFF;
}

function getSurrogatePairCodePoint(cp1, cp2) {
    return (cp1 - 0xD800) * 0x400 + 0x2400 + cp2;
}


//Const
var DEFAULT_BUFFER_WATERLINE = 1 << 16;


//Preprocessor
//NOTE: HTML input preprocessing
//(see: http://www.whatwg.org/specs/web-apps/current-work/multipage/parsing.html#preprocessing-the-input-stream)
var Preprocessor = module.exports = function () {
    this.html = null;

    this.pos = -1;
    this.lastGapPos = -1;
    this.lastCharPos = -1;

    this.gapStack = [];

    this.skipNextNewLine = false;

    this.lastChunkWritten = false;
    this.endOfChunkHit = false;
    this.bufferWaterline = DEFAULT_BUFFER_WATERLINE;
};

Preprocessor.prototype.dropParsedChunk = function () {
    if (this.pos > this.bufferWaterline) {
        this.lastCharPos -= this.pos;
        this.html = this.html.substring(this.pos);
        this.pos = 0;
        this.lastGapPos = -1;
        this.gapStack = [];
    }
};

Preprocessor.prototype._addGap = function () {
    this.gapStack.push(this.lastGapPos);
    this.lastGapPos = this.pos;
};

Preprocessor.prototype._processHighRangeCodePoint = function (cp) {
    //NOTE: try to peek a surrogate pair
    if (this.pos !== this.lastCharPos) {
        var nextCp = this.html.charCodeAt(this.pos + 1);

        if (isSurrogatePair(cp, nextCp)) {
            //NOTE: we have a surrogate pair. Peek pair character and recalculate code point.
            this.pos++;
            cp = getSurrogatePairCodePoint(cp, nextCp);

            //NOTE: add gap that should be avoided during retreat
            this._addGap();
        }
    }

    // NOTE: we've hit the end of chunk, stop processing at this point
    else if (!this.lastChunkWritten) {
        this.endOfChunkHit = true;
        return $.EOF;
    }

    return cp;
};

Preprocessor.prototype.write = function (chunk, isLastChunk) {
    if (this.html)
        this.html += chunk;

    else
        this.html = chunk;

    this.lastCharPos = this.html.length - 1;
    this.endOfChunkHit = false;
    this.lastChunkWritten = isLastChunk;
};

Preprocessor.prototype.insertHtmlAtCurrentPos = function (chunk) {
    this.html = this.html.substring(0, this.pos + 1) +
                chunk +
                this.html.substring(this.pos + 1, this.html.length);

    this.lastCharPos = this.html.length - 1;
    this.endOfChunkHit = false;
};


Preprocessor.prototype.advance = function () {
    this.pos++;

    if (this.pos > this.lastCharPos) {
        if (!this.lastChunkWritten)
            this.endOfChunkHit = true;

        return $.EOF;
    }

    var cp = this.html.charCodeAt(this.pos);

    //NOTE: any U+000A LINE FEED (LF) characters that immediately follow a U+000D CARRIAGE RETURN (CR) character
    //must be ignored.
    if (this.skipNextNewLine && cp === $.LINE_FEED) {
        this.skipNextNewLine = false;
        this._addGap();
        return this.advance();
    }

    //NOTE: all U+000D CARRIAGE RETURN (CR) characters must be converted to U+000A LINE FEED (LF) characters
    if (cp === $.CARRIAGE_RETURN) {
        this.skipNextNewLine = true;
        return $.LINE_FEED;
    }

    this.skipNextNewLine = false;

    //OPTIMIZATION: first perform check if the code point in the allowed range that covers most common
    //HTML input (e.g. ASCII codes) to avoid performance-cost operations for high-range code points.
    return cp >= 0xD800 ? this._processHighRangeCodePoint(cp) : cp;
};

Preprocessor.prototype.retreat = function () {
    if (this.pos === this.lastGapPos) {
        this.lastGapPos = this.gapStack.pop();
        this.pos--;
    }

    this.pos--;
};