Custom component example

One of the best ways to understand how to create a component is to see an example of one and how it's put together.

Here we've created a simple QR code generator that contains a JSON Manifest file and two JavaScript files (one for the custom element, and one to generate the code). The entire QR code example component can be downloaded directly. It is a zip file that contains a directory of four files:

  • myqrcode.js - The custom element for the QR code generator
  • manifest.json - The JSON manifest for the component
  • demo.html - An HTML file that demonstrates the component
  • qr.js - The JavaScript that generates the QR code

You can also examine the custom element, the manifest file, and the QR code generator below.

myqrcode.js

The QR code uses the following JavaScript to create a custom element. You can also directly download the JavaScript for the QR code component.

 
/**
 * @fileoverview Implementation of my-qrcode component.
 * @author ksubramanian@google.com (Kishore Subramanian)
 */
if (document.registerElement) {

  var proto = Object.create(HTMLElement.prototype, {

    createdCallback: {
      value:
      function() {
        this.img = null;
      },
      enumerable: true
    },

    attachedCallback: {
      value:
      /**
       * Lifecycle callback that is invoked when this element is added to the
       * DOM.
       */
      function() {
        this.generateImage();
      },
      enumerable: true
    },

    attributeChangedCallback: {
      value:
      /**
       * Lifecycle callback that is invoked when an attribute is changed.
       * @param {string} attributeName Name of the attribute being changed.
       */
      function(attributeName) {
        if (!this.img) {
          // It is possible that the attribute is set before before the
          // component is added to the DOM.
          return;
        }
        switch (attributeName) {
          case 'data':
            this.generateImage();
            break;
        }
      },
      enumerable: true
    },

    generateImage: {
      value:
      function() {
        var data = this.getAttribute('data');
        if (data) {
          if (!this.img) {
            this.img = document.createElement('img');
            this.img.style.height = '100%';
            this.appendChild(this.img);
          }
          this.img.setAttribute('src', QRCode.generatePNG(data));
        }	
      }
    }
  });

  document.registerElement('my-qrcode', {prototype: proto});
}
manifest.json

This is the JSON Manifest file for the QR code generator. It follows the Component JSON manifest format. You can also download the QR code JSON manifest for customization if you like.


{
  "name": "QR Code",
  "type": "my-qrcode",
  "tagName": "my-qrcode",
  "version": 0.1,
  "description": "Generates a QR code image for the specified data.",
  "files": {
    "js": [
      "qr.js",
      "myqrcode.js"
    ]
  },
  "attributes": [
   {
    "name": "data",
    "label": "Data",
    "type": "string",
    "required": true,
    "description": "The data to encode in the QR code image"
   }
  ],
  "events": [
  ],
  "methods": [
  ],
  "nestable": false
}
qr.js

This is the JavaScript that generates the QR code. It's referenced in the Component JSON manifest. You can also download the QR generator code for customization if you like.

 
/* qr.js -- QR code generator in Javascript (revision 2011-01-19)
 * Written by Kang Seonghoon <public+qrjs@mearie.org>.
 *
 * This source code is in the public domain; if your jurisdiction does not
 * recognize the public domain the terms of Creative Commons CC0 license
 * apply. In the other words, you can always do what you want.
 */

var QRCode = (function(){

/* Quick overview: QR code composed of 2D array of modules (a rectangular
 * area that conveys one bit of information); some modules are fixed to help
 * the recognition of the code, and remaining data modules are further divided
 * into 8-bit code words which are augumented by Reed-Solomon error correcting
 * codes (ECC). There could be multiple ECCs, in the case the code is so large
 * that it is helpful to split the raw data into several chunks.
 *
 * The number of modules is determined by the code's "version", ranging from 1
 * (21x21) to 40 (177x177). How many ECC bits are used is determined by the
 * ECC level (L/M/Q/H). The number and size (and thus the order of generator
 * polynomial) of ECCs depend to the version and ECC level.
 */

// per-version information (cf. JIS X 0510:2004 pp. 30--36, 71)
//
// [0]: the degree of generator polynomial by ECC levels
// [1]: # of code blocks by ECC levels
// [2]: left-top positions of alignment patterns
//
// the number in this table (in particular, [0]) does not exactly match with
// the numbers in the specficiation. see augumenteccs below for the reason.
var VERSIONS = [
	null,
	[[10, 7,17,13], [ 1, 1, 1, 1], []],
	[[16,10,28,22], [ 1, 1, 1, 1], [4,16]],
	[[26,15,22,18], [ 1, 1, 2, 2], [4,20]],
	[[18,20,16,26], [ 2, 1, 4, 2], [4,24]],
	[[24,26,22,18], [ 2, 1, 4, 4], [4,28]],
	[[16,18,28,24], [ 4, 2, 4, 4], [4,32]],
	[[18,20,26,18], [ 4, 2, 5, 6], [4,20,36]],
	[[22,24,26,22], [ 4, 2, 6, 6], [4,22,40]],
	[[22,30,24,20], [ 5, 2, 8, 8], [4,24,44]],
	[[26,18,28,24], [ 5, 4, 8, 8], [4,26,48]],
	[[30,20,24,28], [ 5, 4,11, 8], [4,28,52]],
	[[22,24,28,26], [ 8, 4,11,10], [4,30,56]],
	[[22,26,22,24], [ 9, 4,16,12], [4,32,60]],
	[[24,30,24,20], [ 9, 4,16,16], [4,24,44,64]],
	[[24,22,24,30], [10, 6,18,12], [4,24,46,68]],
	[[28,24,30,24], [10, 6,16,17], [4,24,48,72]],
	[[28,28,28,28], [11, 6,19,16], [4,28,52,76]],
	[[26,30,28,28], [13, 6,21,18], [4,28,54,80]],
	[[26,28,26,26], [14, 7,25,21], [4,28,56,84]],
	[[26,28,28,30], [16, 8,25,20], [4,32,60,88]],
	[[26,28,30,28], [17, 8,25,23], [4,26,48,70,92]],
	[[28,28,24,30], [17, 9,34,23], [4,24,48,72,96]],
	[[28,30,30,30], [18, 9,30,25], [4,28,52,76,100]],
	[[28,30,30,30], [20,10,32,27], [4,26,52,78,104]],
	[[28,26,30,30], [21,12,35,29], [4,30,56,82,108]],
	[[28,28,30,28], [23,12,37,34], [4,28,56,84,112]],
	[[28,30,30,30], [25,12,40,34], [4,32,60,88,116]],
	[[28,30,30,30], [26,13,42,35], [4,24,48,72,96,120]],
	[[28,30,30,30], [28,14,45,38], [4,28,52,76,100,124]],
	[[28,30,30,30], [29,15,48,40], [4,24,50,76,102,128]],
	[[28,30,30,30], [31,16,51,43], [4,28,54,80,106,132]],
	[[28,30,30,30], [33,17,54,45], [4,32,58,84,110,136]],
	[[28,30,30,30], [35,18,57,48], [4,28,56,84,112,140]],
	[[28,30,30,30], [37,19,60,51], [4,32,60,88,116,144]],
	[[28,30,30,30], [38,19,63,53], [4,28,52,76,100,124,148]],
	[[28,30,30,30], [40,20,66,56], [4,22,48,74,100,126,152]],
	[[28,30,30,30], [43,21,70,59], [4,26,52,78,104,130,156]],
	[[28,30,30,30], [45,22,74,62], [4,30,56,82,108,134,160]],
	[[28,30,30,30], [47,24,77,65], [4,24,52,80,108,136,164]],
	[[28,30,30,30], [49,25,81,68], [4,28,56,84,112,140,168]]];

// mode constants (cf. Table 2 in JIS X 0510:2004 p. 16)
var MODE_TERMINATOR = 0;
var MODE_NUMERIC = 1, MODE_ALPHANUMERIC = 2, MODE_OCTET = 4, MODE_KANJI = 8;

// validation regexps
var NUMERIC_REGEXP = /^\d*$/;
var ALPHANUMERIC_REGEXP = /^[A-Za-z0-9 $%*+\-./:]*$/;
var ALPHANUMERIC_OUT_REGEXP = /^[A-Z0-9 $%*+\-./:]*$/;

// ECC levels (cf. Table 22 in JIS X 0510:2004 p. 45)
var ECCLEVEL_L = 1, ECCLEVEL_M = 0, ECCLEVEL_Q = 3, ECCLEVEL_H = 2;

// GF(2^8)-to-integer mapping with a reducing polynomial x^8+x^4+x^3+x^2+1
// invariant: GF256_MAP[GF256_INVMAP[i]] == i for all i in [1,256)
var GF256_MAP = [], GF256_INVMAP = [-1];
for (var i = 0, v = 1; i < 255; ++i) {
	GF256_MAP.push(v);
	GF256_INVMAP[v] = i;
	v = (v * 2) ^ (v >= 128 ? 0x11d : 0);
}

// generator polynomials up to degree 30
// (should match with polynomials in JIS X 0510:2004 Appendix A)
//
// generator polynomial of degree K is product of (x-\alpha^0), (x-\alpha^1),
// ..., (x-\alpha^(K-1)). by convention, we omit the K-th coefficient (always 1)
// from the result; also other coefficients are written in terms of the exponent
// to \alpha to avoid the redundant calculation. (see also calculateecc below.)
var GF256_GENPOLY = [[]];
for (var i = 0; i < 30; ++i) {
	var prevpoly = GF256_GENPOLY[i], poly = [];
	for (var j = 0; j <= i; ++j) {
		var a = (j < i ? GF256_MAP[prevpoly[j]] : 0);
		var b = GF256_MAP[(i + (prevpoly[j-1] || 0)) % 255];
		poly.push(GF256_INVMAP[a ^ b]);
	}
	GF256_GENPOLY.push(poly);
}

// alphanumeric character mapping (cf. Table 5 in JIS X 0510:2004 p. 19)
var ALPHANUMERIC_MAP = {};
for (var i = 0; i < 45; ++i) {
	ALPHANUMERIC_MAP['0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ $%*+-./:'.charAt(i)] = i;
}

// mask functions in terms of row # and column #
// (cf. Table 20 in JIS X 0510:2004 p. 42)
var MASKFUNCS = [
	function(i,j) { return (i+j) % 2 == 0; },
	function(i,j) { return i % 2 == 0; },
	function(i,j) { return j % 3 == 0; },
	function(i,j) { return (i+j) % 3 == 0; },
	function(i,j) { return (((i/2)|0) + ((j/3)|0)) % 2 == 0; },
	function(i,j) { return (i*j) % 2 + (i*j) % 3 == 0; },
	function(i,j) { return ((i*j) % 2 + (i*j) % 3) % 2 == 0; },
	function(i,j) { return ((i+j) % 2 + (i*j) % 3) % 2 == 0; }];

// returns true when the version information has to be embeded.
var needsverinfo = function(ver) { return ver > 6; };

// returns the size of entire QR code for given version.
var getsizebyver = function(ver) { return 4 * ver + 17; };

// returns the number of bits available for code words in this version.
var nfullbits = function(ver) {
	/*
	 * |<--------------- n --------------->|
	 * |        |<----- n-17 ---->|        |
	 * +-------+                ///+-------+ ----
	 * |       |                ///|       |    ^
	 * |  9x9  |       @@@@@    ///|  9x8  |    |
	 * |       | # # # @5x5@ # # # |       |    |
	 * +-------+       @@@@@       +-------+    |
	 *       #                               ---|
	 *                                        ^ |
	 *       #                                |
	 *     @@@@@       @@@@@       @@@@@      | n
	 *     @5x5@       @5x5@       @5x5@   n-17
	 *     @@@@@       @@@@@       @@@@@      | |
	 *       #                                | |
	 * //////                                 v |
	 * //////#                               ---|
	 * +-------+       @@@@@       @@@@@        |
	 * |       |       @5x5@       @5x5@        |
	 * |  8x9  |       @@@@@       @@@@@        |
	 * |       |                                v
	 * +-------+                             ----
	 *
	 * when the entire code has n^2 modules and there are m^2-3 alignment
	 * patterns, we have:
	 * - 225 (= 9x9 + 9x8 + 8x9) modules for finder patterns and
	 *   format information;
	 * - 2n-34 (= 2(n-17)) modules for timing patterns;
	 * - 36 (= 3x6 + 6x3) modules for version information, if any;
	 * - 25m^2-75 (= (m^2-3)(5x5)) modules for alignment patterns
	 *   if any, but 10m-20 (= 2(m-2)x5) of them overlaps with
	 *   timing patterns.
	 */
	var v = VERSIONS[ver];
	var nbits = 16*ver*ver + 128*ver + 64; // finder, timing and format info.
	if (needsverinfo(ver)) nbits -= 36; // version information
	if (v[2].length) { // alignment patterns
		nbits -= 25 * v[2].length * v[2].length - 10 * v[2].length - 55;
	}
	return nbits;
};

// returns the number of bits available for data portions (i.e. excludes ECC
// bits but includes mode and length bits) in this version and ECC level.
var ndatabits = function(ver, ecclevel) {
	var nbits = nfullbits(ver) & ~7; // no sub-octet code words
	var v = VERSIONS[ver];
	nbits -= 8 * v[0][ecclevel] * v[1][ecclevel]; // ecc bits
	return nbits;
}

// returns the number of bits required for the length of data.
// (cf. Table 3 in JIS X 0510:2004 p. 16)
var ndatalenbits = function(ver, mode) {
	switch (mode) {
	case MODE_NUMERIC: return (ver < 10 ? 10 : ver < 27 ? 12 : 14);
	case MODE_ALPHANUMERIC: return (ver < 10 ? 9 : ver < 27 ? 11 : 13);
	case MODE_OCTET: return (ver < 10 ? 8 : 16);
	case MODE_KANJI: return (ver < 10 ? 8 : ver < 27 ? 10 : 12);
	}
};

// returns the maximum length of data possible in given configuration.
var getmaxdatalen = function(ver, mode, ecclevel) {
	var nbits = ndatabits(ver, ecclevel) - 4 - ndatalenbits(ver, mode); // 4 for mode bits
	switch (mode) {
	case MODE_NUMERIC:
		return ((nbits/10) | 0) * 3 + (nbits%10 < 4 ? 0 : nbits%10 < 7 ? 1 : 2);
	case MODE_ALPHANUMERIC:
		return ((nbits/11) | 0) * 2 + (nbits%11 < 6 ? 0 : 1);
	case MODE_OCTET:
		return (nbits/8) | 0;
	case MODE_KANJI:
		return (nbits/13) | 0;
	}
};

// checks if the given data can be encoded in given mode, and returns
// the converted data for the further processing if possible. otherwise
// returns null.
//
// this function does not check the length of data; it is a duty of
// encode function below (as it depends on the version and ECC level too).
var validatedata = function(mode, data) {
	switch (mode) {
	case MODE_NUMERIC:
		if (!data.match(NUMERIC_REGEXP)) return null;
		return data;

	case MODE_ALPHANUMERIC:
		if (!data.match(ALPHANUMERIC_REGEXP)) return null;
		return data.toUpperCase();

	case MODE_OCTET:
		if (typeof data === 'string') { // encode as utf-8 string
			var newdata = [];
			for (var i = 0; i < data.length; ++i) {
				var ch = data.charCodeAt(i);
				if (ch < 0x80) {
					newdata.push(ch);
				} else if (ch < 0x800) {
					newdata.push(0xc0 | (ch >> 6),
						0x80 | (ch & 0x3f));
				} else if (ch < 0x10000) {
					newdata.push(0xe0 | (ch >> 12),
						0x80 | ((ch >> 6) & 0x3f),
						0x80 | (ch & 0x3f));
				} else {
					newdata.push(0xf0 | (ch >> 18),
						0x80 | ((ch >> 12) & 0x3f),
						0x80 | ((ch >> 6) & 0x3f),
						0x80 | (ch & 0x3f));
				}
			}
			return newdata;
		} else {
			return data;
		}
	}
};

// returns the code words (sans ECC bits) for given data and configurations.
// requires data to be preprocessed by validatedata. no length check is
// performed, and everything has to be checked before calling this function.
var encode = function(ver, mode, data, maxbuflen) {
	var buf = [];
	var bits = 0, remaining = 8;
	var datalen = data.length;

	// this function is intentionally no-op when n=0.
	var pack = function(x, n) {
		if (n >= remaining) {
			buf.push(bits | (x >> (n -= remaining)));
			while (n >= 8) buf.push((x >> (n -= 8)) & 255);
			bits = 0;
			remaining = 8;
		}
		if (n > 0) bits |= (x & ((1 << n) - 1)) << (remaining -= n);
	};

	var nlenbits = ndatalenbits(ver, mode);
	pack(mode, 4);
	pack(datalen, nlenbits);

	switch (mode) {
	case MODE_NUMERIC:
		for (var i = 2; i < datalen; i += 3) {
			pack(parseInt(data.substring(i-2,i+1), 10), 10);
		}
		pack(parseInt(data.substring(i-2), 10), [0,4,7][datalen%3]);
		break;

	case MODE_ALPHANUMERIC:
		for (var i = 1; i < datalen; i += 2) {
			pack(ALPHANUMERIC_MAP[data.charAt(i-1)] * 45 +
				ALPHANUMERIC_MAP[data.charAt(i)], 11);
		}
		if (datalen % 2 == 1) {
			pack(ALPHANUMERIC_MAP[data.charAt(i-1)], 6);
		}
		break;

	case MODE_OCTET:
		for (var i = 0; i < datalen; ++i) {
			pack(data[i], 8);
		}
		break;
	};

	// final bits. it is possible that adding terminator causes the buffer
	// to overflow, but then the buffer truncated to the maximum size will
	// be valid as the truncated terminator mode bits and padding is
	// identical in appearance (cf. JIS X 0510:2004 sec 8.4.8).
	pack(MODE_TERMINATOR, 4);
	if (remaining < 8) buf.push(bits);

	// the padding to fill up the remaining space. we should not add any
	// words when the overflow already occurred.
	while (buf.length + 1 < maxbuflen) buf.push(0xec, 0x11);
	if (buf.length < maxbuflen) buf.push(0xec);
	return buf;
};

// calculates ECC code words for given code words and generator polynomial.
//
// this is quite similar to CRC calculation as both Reed-Solomon and CRC use
// the certain kind of cyclic codes, which is effectively the division of
// zero-augumented polynomial by the generator polynomial. the only difference
// is that Reed-Solomon uses GF(2^8), instead of CRC's GF(2), and Reed-Solomon
// uses the different generator polynomial than CRC's.
var calculateecc = function(poly, genpoly) {
	var modulus = poly.slice(0);
	var polylen = poly.length, genpolylen = genpoly.length;
	for (var i = 0; i < genpolylen; ++i) modulus.push(0);
	for (var i = 0; i < polylen; ) {
		var quotient = GF256_INVMAP[modulus[i++]];
		if (quotient >= 0) {
			for (var j = 0; j < genpolylen; ++j) {
				modulus[i+j] ^= GF256_MAP[(quotient + genpoly[j]) % 255];
			}
		}
	}
	return modulus.slice(polylen);
};

// auguments ECC code words to given code words. the resulting words are
// ready to be encoded in the matrix.
//
// the much of actual augumenting procedure follows JIS X 0510:2004 sec 8.7.
// the code is simplified using the fact that the size of each code & ECC
// blocks is almost same; for example, when we have 4 blocks and 46 data words
// the number of code words in those blocks are 11, 11, 12, 12 respectively.
var augumenteccs = function(poly, nblocks, genpoly) {
	var subsizes = [];
	var subsize = (poly.length / nblocks) | 0, subsize0 = 0;
	var pivot = nblocks - poly.length % nblocks;
	for (var i = 0; i < pivot; ++i) {
		subsizes.push(subsize0);
		subsize0 += subsize;
	}
	for (var i = pivot; i < nblocks; ++i) {
		subsizes.push(subsize0);
		subsize0 += subsize+1;
	}
	subsizes.push(subsize0);

	var eccs = [];
	for (var i = 0; i < nblocks; ++i) {
		eccs.push(calculateecc(poly.slice(subsizes[i], subsizes[i+1]), genpoly));
	}

	var result = [];
	var nitemsperblock = (poly.length / nblocks) | 0;
	for (var i = 0; i < nitemsperblock; ++i) {
		for (var j = 0; j < nblocks; ++j) {
			result.push(poly[subsizes[j] + i]);
		}
	}
	for (var j = pivot; j < nblocks; ++j) {
		result.push(poly[subsizes[j+1] - 1]);
	}
	for (var i = 0; i < genpoly.length; ++i) {
		for (var j = 0; j < nblocks; ++j) {
			result.push(eccs[j][i]);
		}
	}
	return result;
};

// auguments BCH(p+q,q) code to the polynomial over GF(2), given the proper
// genpoly. the both input and output are in binary numbers, and unlike
// calculateecc genpoly should include the 1 bit for the highest degree.
//
// actual polynomials used for this procedure are as follows:
// - p=10, q=5, genpoly=x^10+x^8+x^5+x^4+x^2+x+1 (JIS X 0510:2004 Appendix C)
// - p=18, q=6, genpoly=x^12+x^11+x^10+x^9+x^8+x^5+x^2+1 (ibid. Appendix D)
var augumentbch = function(poly, p, genpoly, q) {
	var modulus = poly << q;
	for (var i = p - 1; i >= 0; --i) {
		if ((modulus >> (q+i)) & 1) modulus ^= genpoly << i;
	}
	return (poly << q) | modulus;
};

// creates the base matrix for given version. it returns two matrices, one of
// them is the actual one and the another represents the "reserved" portion
// (e.g. finder and timing patterns) of the matrix.
//
// some entries in the matrix may be undefined, rather than 0 or 1. this is
// intentional (no initialization needed!), and putdata below will fill
// the remaining ones.
var makebasematrix = function(ver) {
	var v = VERSIONS[ver], n = getsizebyver(ver);
	var matrix = [], reserved = [];
	for (var i = 0; i < n; ++i) {
		matrix.push([]);
		reserved.push([]);
	}

	var blit = function(y, x, h, w, bits) {
		for (var i = 0; i < h; ++i) {
			for (var j = 0; j < w; ++j) {
				matrix[y+i][x+j] = (bits[i] >> j) & 1;
				reserved[y+i][x+j] = 1;
			}
		}
	};

	// finder patterns and a part of timing patterns
	// will also mark the format information area (not yet written) as reserved.
	blit(0, 0, 9, 9, [0x7f, 0x41, 0x5d, 0x5d, 0x5d, 0x41, 0x17f, 0x00, 0x40]);
	blit(n-8, 0, 8, 9, [0x100, 0x7f, 0x41, 0x5d, 0x5d, 0x5d, 0x41, 0x7f]);
	blit(0, n-8, 9, 8, [0xfe, 0x82, 0xba, 0xba, 0xba, 0x82, 0xfe, 0x00, 0x00]);

	// the rest of timing patterns
	for (var i = 9; i < n-8; ++i) {
		matrix[6][i] = matrix[i][6] = ~i & 1;
		reserved[6][i] = reserved[i][6] = 1;
	}

	// alignment patterns
	var aligns = v[2], m = aligns.length;
	for (var i = 0; i < m; ++i) {
		var minj = (i==0 || i==m-1 ? 1 : 0), maxj = (i==0 ? m-1 : m);
		for (var j = minj; j < maxj; ++j) {
			blit(aligns[i], aligns[j], 5, 5, [0x1f, 0x11, 0x15, 0x11, 0x1f]);
		}
	}

	// version information
	if (needsverinfo(ver)) {
		var code = augumentbch(ver, 6, 0x1f25, 12);
		var k = 0;
		for (var i = 0; i < 6; ++i) {
			for (var j = 0; j < 3; ++j) {
				matrix[i][(n-11)+j] = matrix[(n-11)+j][i] = (code >> k++) & 1;
				reserved[i][(n-11)+j] = reserved[(n-11)+j][i] = 1;
			}
		}
	}

	return {matrix: matrix, reserved: reserved};
};

// fills the data portion (i.e. unmarked in reserved) of the matrix with given
// code words. the size of code words should be no more than available bits,
// and remaining bits are padded to 0 (cf. JIS X 0510:2004 sec 8.7.3).
var putdata = function(matrix, reserved, buf) {
	var n = matrix.length;
	var k = 0, dir = -1;
	for (var i = n-1; i >= 0; i -= 2) {
		if (i == 6) --i; // skip the entire timing pattern column
		var jj = (dir < 0 ? n-1 : 0);
		for (var j = 0; j < n; ++j) {
			for (var ii = i; ii > i-2; --ii) {
				if (!reserved[jj][ii]) {
					// may overflow, but (undefined >> x)
					// is 0 so it will auto-pad to zero.
					matrix[jj][ii] = (buf[k >> 3] >> (~k&7)) & 1;
					++k;
				}
			}
			jj += dir;
		}
		dir = -dir;
	}
	return matrix;
};

// XOR-masks the data portion of the matrix. repeating the call with the same
// arguments will revert the prior call (convenient in the matrix evaluation).
var maskdata = function(matrix, reserved, mask) {
	var maskf = MASKFUNCS[mask];
	var n = matrix.length;
	for (var i = 0; i < n; ++i) {
		for (var j = 0; j < n; ++j) {
			if (!reserved[i][j]) matrix[i][j] ^= maskf(i,j);
		}
	}
	return matrix;
}

// puts the format information.
var putformatinfo = function(matrix, reserved, ecclevel, mask) {
	var n = matrix.length;
	var code = augumentbch((ecclevel << 3) | mask, 5, 0x537, 10) ^ 0x5412;
	for (var i = 0; i < 15; ++i) {
		var r = [0,1,2,3,4,5,7,8,n-7,n-6,n-5,n-4,n-3,n-2,n-1][i];
		var c = [n-1,n-2,n-3,n-4,n-5,n-6,n-7,n-8,7,5,4,3,2,1,0][i];
		matrix[r][8] = matrix[8][c] = (code >> i) & 1;
		// we don't have to mark those bits reserved; always done
		// in makebasematrix above.
	}
	return matrix;
};

// evaluates the resulting matrix and returns the score (lower is better).
// (cf. JIS X 0510:2004 sec 8.8.2)
//
// the evaluation procedure tries to avoid the problematic patterns naturally
// occuring from the original matrix. for example, it penaltizes the patterns
// which just look like the finder pattern which will confuse the decoder.
// we choose the mask which results in the lowest score among 8 possible ones.
//
// note: zxing seems to use the same procedure and in many cases its choice
// agrees to ours, but sometimes it does not. practically it doesn't matter.
var evaluatematrix = function(matrix) {
	// N1+(k-5) points for each consecutive row of k same-colored modules,
	// where k >= 5. no overlapping row counts.
	var PENALTY_CONSECUTIVE = 3;
	// N2 points for each 2x2 block of same-colored modules.
	// overlapping block does count.
	var PENALTY_TWOBYTWO = 3;
	// N3 points for each pattern with >4W:1B:1W:3B:1W:1B or
	// 1B:1W:3B:1W:1B:>4W, or their multiples (e.g. highly unlikely,
	// but 13W:3B:3W:9B:3W:3B counts).
	var PENALTY_FINDERLIKE = 40;
	// N4*k points for every (5*k)% deviation from 50% black density.
	// i.e. k=1 for 55~60% and 40~45%, k=2 for 60~65% and 35~40%, etc.
	var PENALTY_DENSITY = 10;

	var evaluategroup = function(groups) { // assumes [W,B,W,B,W,...,B,W]
		var score = 0;
		for (var i = 0; i < groups.length; ++i) {
			if (groups[i] >= 5) score += PENALTY_CONSECUTIVE + (groups[i]-5);
		}
		for (var i = 5; i < groups.length; i += 2) {
			var p = groups[i];
			if (groups[i-1] == p && groups[i-2] == 3*p && groups[i-3] == p &&
					groups[i-4] == p && (groups[i-5] >= 4*p || groups[i+1] >= 4*p)) {
				// this part differs from zxing...
				score += PENALTY_FINDERLIKE;
			}
		}
		return score;
	};

	var n = matrix.length;
	var score = 0, nblacks = 0;
	for (var i = 0; i < n; ++i) {
		var row = matrix[i];
		var groups;

		// evaluate the current row
		groups = [0]; // the first empty group of white
		for (var j = 0; j < n; ) {
			var k;
			for (k = 0; j < n && row[j]; ++k) ++j;
			groups.push(k);
			for (k = 0; j < n && !row[j]; ++k) ++j;
			groups.push(k);
		}
		score += evaluategroup(groups);

		// evaluate the current column
		groups = [0];
		for (var j = 0; j < n; ) {
			var k;
			for (k = 0; j < n && matrix[j][i]; ++k) ++j;
			groups.push(k);
			for (k = 0; j < n && !matrix[j][i]; ++k) ++j;
			groups.push(k);
		}
		score += evaluategroup(groups);

		// check the 2x2 box and calculate the density
		var nextrow = matrix[i+1] || [];
		nblacks += row[0];
		for (var j = 1; j < n; ++j) {
			var p = row[j];
			nblacks += p;
			// at least comparison with next row should be strict...
			if (row[j-1] == p && nextrow[j] === p && nextrow[j-1] === p) {
				score += PENALTY_TWOBYTWO;
			}
		}
	}

	score += PENALTY_DENSITY * ((Math.abs(nblacks / n / n - 0.5) / 0.05) | 0);
	return score;
};

// returns the fully encoded QR code matrix which contains given data.
// it also chooses the best mask automatically when mask is -1.
var generate = function(data, ver, mode, ecclevel, mask) {
	var v = VERSIONS[ver];
	var buf = encode(ver, mode, data, ndatabits(ver, ecclevel) >> 3);
	buf = augumenteccs(buf, v[1][ecclevel], GF256_GENPOLY[v[0][ecclevel]]);

	var result = makebasematrix(ver);
	var matrix = result.matrix, reserved = result.reserved;
	putdata(matrix, reserved, buf);

	if (mask < 0) {
		// find the best mask
		maskdata(matrix, reserved, 0);
		putformatinfo(matrix, reserved, ecclevel, 0);
		var bestmask = 0, bestscore = evaluatematrix(matrix);
		maskdata(matrix, reserved, 0);
		for (mask = 1; mask < 8; ++mask) {
			maskdata(matrix, reserved, mask);
			putformatinfo(matrix, reserved, ecclevel, mask);
			var score = evaluatematrix(matrix);
			if (bestscore > score) {
				bestscore = score;
				bestmask = mask;
			}
			maskdata(matrix, reserved, mask);
		}
		mask = bestmask;
	}

	maskdata(matrix, reserved, mask);
	putformatinfo(matrix, reserved, ecclevel, mask);
	return matrix;
};

// the public interface is trivial; the options available are as follows:
//
// - version: an integer in [1,40]. when omitted (or -1) the smallest possible
//   version is chosen.
// - mode: one of 'numeric', 'alphanumeric', 'octet'. when omitted the smallest
//   possible mode is chosen.
// - ecclevel: one of 'L', 'M', 'Q', 'H'. defaults to 'L'.
// - mask: an integer in [0,7]. when omitted (or -1) the best mask is chosen.
//
// for generate{HTML,PNG}:
//
// - modulesize: a number. this is a size of each modules in pixels, and
//   defaults to 5px.
// - margin: a number. this is a size of margin in *modules*, and defaults to
//   4 (white modules). the specficiation mandates the margin no less than 4
//   modules, so it is better not to alter this value unless you know what
//   you're doing.
var QRCode = {
	'generate': function(data, options) {
		var MODES = {'numeric': MODE_NUMERIC, 'alphanumeric': MODE_ALPHANUMERIC,
			'octet': MODE_OCTET};
		var ECCLEVELS = {'L': ECCLEVEL_L, 'M': ECCLEVEL_M, 'Q': ECCLEVEL_Q,
			'H': ECCLEVEL_H};

		options = options || {};
		var ver = options.version || -1;
		var ecclevel = ECCLEVELS[(options.ecclevel || 'L').toUpperCase()];
		var mode = options.mode ? MODES[options.mode.toLowerCase()] : -1;
		var mask = 'mask' in options ? options.mask : -1;

		if (mode < 0) {
			if (typeof data === 'string') {
				if (data.match(NUMERIC_REGEXP)) {
					mode = MODE_NUMERIC;
				} else if (data.match(ALPHANUMERIC_OUT_REGEXP)) {
					// while encode supports case-insensitive
					// encoding, we restrict the data to be
					// uppercased when auto-selecting the mode.
					mode = MODE_ALPHANUMERIC;
				} else {
					mode = MODE_OCTET;
				}
			} else {
				mode = MODE_OCTET;
			}
		} else if (!(mode == MODE_NUMERIC || mode == MODE_ALPHANUMERIC ||
				mode == MODE_OCTET)) {
			throw 'invalid or unsupported mode';
		}

		data = validatedata(mode, data);
		if (data === null) throw 'invalid data format';

		if (ecclevel < 0 || ecclevel > 3) throw 'invalid ECC level';

		if (ver < 0) {
			for (ver = 1; ver <= 40; ++ver) {
				if (data.length <= getmaxdatalen(ver, mode, ecclevel)) break;
			}
			if (ver > 40) throw 'too large data';
		} else if (ver < 1 || ver > 40) {
			throw 'invalid version';
		}

		if (mask != -1 && (mask < 0 || mask > 8)) throw 'invalid mask';

		return generate(data, ver, mode, ecclevel, mask);
	},

	'generateHTML': function(data, options) {
		options = options || {};
		var matrix = QRCode['generate'](data, options);
		var modsize = Math.max(options.modulesize || 5, 0.5);
		var margin = Math.max(options.margin || 4, 0.0);

		var e = document.createElement('div');
		var n = matrix.length;
		var html = ['<table border="0" cellspacing="0" cellpadding="0" style="border:' +
			modsize*margin + 'px solid #fff;background:#fff">'];
		for (var i = 0; i < n; ++i) {
			html.push('<tr>');
			for (var j = 0; j < n; ++j) {
				html.push('<td style="width:' + modsize + 'px;height:' + modsize + 'px' +
					(matrix[i][j] ? ';background:#000' : '') + '"></td>');
			}
        	html.push('</tr>');
    	}
		e.className = 'qrcode';
		e.innerHTML = html.join('') + '</table>';
		return e;
	},

	'generatePNG': function(data, options) {
		options = options || {};
		var matrix = QRCode['generate'](data, options);
		var modsize = Math.max(options.modulesize || 5, 0.5);
		var margin = Math.max(options.margin || 4, 0.0);
		var n = matrix.length;
		var size = modsize * (n + 2 * margin);

		var canvas = document.createElement('canvas'), context;
		canvas.width = canvas.height = size;
		context = canvas.getContext('2d');
		if (!context) throw 'canvas support is needed for PNG output';

		context.fillStyle = '#fff';
		context.fillRect(0, 0, size, size);
		context.fillStyle = '#000';
		for (var i = 0; i < n; ++i) {
			for (var j = 0; j < n; ++j) {
				if (matrix[i][j]) {
					context.fillRect(modsize * (margin + j),
						modsize * (margin + i),
						modsize, modsize);
				}
			}
		}
		//context.fillText('evaluation: ' + evaluatematrix(matrix), 10, 10);
		return canvas.toDataURL();
	}
};

return QRCode;
})();
Was this article helpful?
How can we improve it?