mirror of
https://github.com/zint/zint
synced 2024-11-16 20:57:25 +13:00
1877 lines
59 KiB
C
1877 lines
59 KiB
C
/* composite.c - Handles GS1 Composite Symbols */
|
|
|
|
/*
|
|
libzint - the open source barcode library
|
|
Copyright (C) 2008-2017 Robin Stuart <rstuart114@gmail.com>
|
|
|
|
Redistribution and use in source and binary forms, with or without
|
|
modification, are permitted provided that the following conditions
|
|
are met:
|
|
|
|
1. Redistributions of source code must retain the above copyright
|
|
notice, this list of conditions and the following disclaimer.
|
|
2. Redistributions in binary form must reproduce the above copyright
|
|
notice, this list of conditions and the following disclaimer in the
|
|
documentation and/or other materials provided with the distribution.
|
|
3. Neither the name of the project nor the names of its contributors
|
|
may be used to endorse or promote products derived from this software
|
|
without specific prior written permission.
|
|
|
|
THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND
|
|
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
|
IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
|
|
ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE
|
|
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
|
|
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
|
|
OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
|
|
HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
|
|
LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
|
|
OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
|
|
SUCH DAMAGE.
|
|
*/
|
|
|
|
/* The functions "getBit", "init928" and "encode928" are copyright BSI and are
|
|
released with permission under the following terms:
|
|
|
|
"Copyright subsists in all BSI publications. BSI also holds the copyright, in the
|
|
UK, of the international standardisation bodies. Except as
|
|
permitted under the Copyright, Designs and Patents Act 1988 no extract may be
|
|
reproduced, stored in a retrieval system or transmitted in any form or by any
|
|
means - electronic, photocopying, recording or otherwise - without prior written
|
|
permission from BSI.
|
|
|
|
"This does not preclude the free use, in the course of implementing the standard,
|
|
of necessary details such as symbols, and size, type or grade designations. If these
|
|
details are to be used for any other purpose than implementation then the prior
|
|
written permission of BSI must be obtained."
|
|
|
|
The date of publication for these functions is 31 May 2006
|
|
*/
|
|
|
|
#include <stdio.h>
|
|
#include <string.h>
|
|
#include <stdlib.h>
|
|
#include <assert.h>
|
|
#include <math.h>
|
|
#include <assert.h>
|
|
#ifdef _MSC_VER
|
|
#include <malloc.h>
|
|
#endif
|
|
#include "common.h"
|
|
#include "composite.h"
|
|
#include "pdf417.h"
|
|
#include "gs1.h"
|
|
|
|
#define UINT unsigned short
|
|
|
|
extern int general_rules(char type[]);
|
|
extern int eanx(struct zint_symbol *symbol, unsigned char source[], int length);
|
|
extern int ean_128(struct zint_symbol *symbol, unsigned char source[], const size_t length);
|
|
extern int rss14(struct zint_symbol *symbol, unsigned char source[], int length);
|
|
extern int rsslimited(struct zint_symbol *symbol, unsigned char source[], int length);
|
|
extern int rssexpanded(struct zint_symbol *symbol, unsigned char source[], int length);
|
|
|
|
static UINT pwr928[69][7];
|
|
|
|
static int _min(int first, int second) {
|
|
|
|
if (first <= second)
|
|
return first;
|
|
else
|
|
return second;
|
|
}
|
|
|
|
/* gets bit in bitString at bitPos */
|
|
static int getBit(UINT *bitStr, int bitPos) {
|
|
return !!(bitStr[bitPos >> 4] & (0x8000 >> (bitPos & 15)));
|
|
}
|
|
|
|
/* initialize pwr928 encoding table */
|
|
static void init928(void) {
|
|
int i, j, v;
|
|
int cw[7];
|
|
cw[6] = 1L;
|
|
for (i = 5; i >= 0; i--)
|
|
cw[i] = 0;
|
|
|
|
for (i = 0; i < 7; i++)
|
|
pwr928[0][i] = cw[i];
|
|
for (j = 1; j < 69; j++) {
|
|
for (v = 0, i = 6; i >= 1; i--) {
|
|
v = (2 * cw[i]) + (v / 928);
|
|
pwr928[j][i] = cw[i] = v % 928;
|
|
}
|
|
pwr928[j][0] = cw[0] = (2 * cw[0]) + (v / 928);
|
|
}
|
|
return;
|
|
}
|
|
|
|
/* converts bit string to base 928 values, codeWords[0] is highest order */
|
|
static int encode928(UINT bitString[], UINT codeWords[], int bitLng) {
|
|
int i, j, b, cwNdx, cwLng;
|
|
for (cwNdx = cwLng = b = 0; b < bitLng; b += 69, cwNdx += 7) {
|
|
int bitCnt = _min(bitLng - b, 69);
|
|
int cwCnt;
|
|
cwLng += cwCnt = bitCnt / 10 + 1;
|
|
for (i = 0; i < cwCnt; i++)
|
|
codeWords[cwNdx + i] = 0; /* init 0 */
|
|
for (i = 0; i < bitCnt; i++) {
|
|
if (getBit(bitString, b + bitCnt - i - 1)) {
|
|
for (j = 0; j < cwCnt; j++)
|
|
codeWords[cwNdx + j] += pwr928[i][j + 7 - cwCnt];
|
|
}
|
|
}
|
|
for (i = cwCnt - 1; i > 0; i--) {
|
|
/* add "carries" */
|
|
codeWords[cwNdx + i - 1] += codeWords[cwNdx + i] / 928L;
|
|
codeWords[cwNdx + i] %= 928L;
|
|
}
|
|
}
|
|
return (cwLng);
|
|
}
|
|
|
|
/* CC-A 2D component */
|
|
static int cc_a(struct zint_symbol *symbol, char source[], int cc_width) {
|
|
int i, segment, bitlen, cwCnt, variant, rows;
|
|
int k, offset, j, total, rsCodeWords[8];
|
|
int LeftRAPStart, RightRAPStart, CentreRAPStart, StartCluster;
|
|
int LeftRAP, RightRAP, CentreRAP, Cluster, dummy[5];
|
|
int loop;
|
|
UINT codeWords[28];
|
|
UINT bitStr[13];
|
|
char pattern[580];
|
|
char local_source[210]; /* A copy of source but with padding zeroes to make 208 bits */
|
|
|
|
variant = 0;
|
|
|
|
for (i = 0; i < 13; i++) {
|
|
bitStr[i] = 0;
|
|
}
|
|
for (i = 0; i < 28; i++) {
|
|
codeWords[i] = 0;
|
|
}
|
|
|
|
bitlen = (int)strlen(source);
|
|
|
|
for (i = 0; i < 208; i++) {
|
|
local_source[i] = '0';
|
|
}
|
|
for (i = 0; i < bitlen; i++) {
|
|
local_source[i] = source[i];
|
|
}
|
|
local_source[208] = '\0';
|
|
|
|
for (segment = 0; segment < 13; segment++) {
|
|
int strpos = segment * 16;
|
|
for (i = 0; i < 16; i++) {
|
|
if (local_source[strpos + i] == '1') {
|
|
bitStr[segment] += (0x8000 >> i);
|
|
}
|
|
}
|
|
}
|
|
|
|
init928();
|
|
/* encode codeWords from bitStr */
|
|
cwCnt = encode928(bitStr, codeWords, bitlen);
|
|
|
|
switch (cc_width) {
|
|
case 2:
|
|
switch (cwCnt) {
|
|
case 6: variant = 0;
|
|
break;
|
|
case 8: variant = 1;
|
|
break;
|
|
case 9: variant = 2;
|
|
break;
|
|
case 11: variant = 3;
|
|
break;
|
|
case 12: variant = 4;
|
|
break;
|
|
case 14: variant = 5;
|
|
break;
|
|
case 17: variant = 6;
|
|
break;
|
|
}
|
|
break;
|
|
case 3:
|
|
switch (cwCnt) {
|
|
case 8: variant = 7;
|
|
break;
|
|
case 10: variant = 8;
|
|
break;
|
|
case 12: variant = 9;
|
|
break;
|
|
case 14: variant = 10;
|
|
break;
|
|
case 17: variant = 11;
|
|
break;
|
|
}
|
|
break;
|
|
case 4:
|
|
switch (cwCnt) {
|
|
case 8: variant = 12;
|
|
break;
|
|
case 11: variant = 13;
|
|
break;
|
|
case 14: variant = 14;
|
|
break;
|
|
case 17: variant = 15;
|
|
break;
|
|
case 20: variant = 16;
|
|
break;
|
|
}
|
|
break;
|
|
}
|
|
|
|
rows = ccaVariants[variant];
|
|
k = ccaVariants[17 + variant];
|
|
offset = ccaVariants[34 + variant];
|
|
|
|
/* Reed-Solomon error correction */
|
|
|
|
for (i = 0; i < 8; i++) {
|
|
rsCodeWords[i] = 0;
|
|
}
|
|
total = 0;
|
|
for (i = 0; i < cwCnt; i++) {
|
|
total = (codeWords[i] + rsCodeWords[k - 1]) % 929;
|
|
for (j = k - 1; j >= 0; j--) {
|
|
if (j == 0) {
|
|
rsCodeWords[j] = (929 - (total * ccaCoeffs[offset + j]) % 929) % 929;
|
|
} else {
|
|
rsCodeWords[j] = (rsCodeWords[j - 1] + 929 - (total * ccaCoeffs[offset + j]) % 929) % 929;
|
|
}
|
|
}
|
|
}
|
|
|
|
for (j = 0; j < k; j++) {
|
|
if (rsCodeWords[j] != 0) {
|
|
rsCodeWords[j] = 929 - rsCodeWords[j];
|
|
}
|
|
}
|
|
|
|
for (i = k - 1; i >= 0; i--) {
|
|
codeWords[cwCnt] = rsCodeWords[i];
|
|
cwCnt++;
|
|
}
|
|
|
|
/* Place data into table */
|
|
LeftRAPStart = aRAPTable[variant];
|
|
CentreRAPStart = aRAPTable[variant + 17];
|
|
RightRAPStart = aRAPTable[variant + 34];
|
|
StartCluster = aRAPTable[variant + 51] / 3;
|
|
|
|
LeftRAP = LeftRAPStart;
|
|
CentreRAP = CentreRAPStart;
|
|
RightRAP = RightRAPStart;
|
|
Cluster = StartCluster; /* Cluster can be 0, 1 or 2 for Cluster(0), Cluster(3) and Cluster(6) */
|
|
|
|
for (i = 0; i < rows; i++) {
|
|
strcpy(pattern, "");
|
|
offset = 929 * Cluster;
|
|
for (j = 0; j < 5; j++) {
|
|
dummy[j] = 0;
|
|
}
|
|
for (j = 0; j < cc_width; j++) {
|
|
dummy[j + 1] = codeWords[i * cc_width + j];
|
|
}
|
|
/* Copy the data into codebarre */
|
|
bin_append(rap_side[LeftRAP - 1], 10, pattern);
|
|
bin_append(pdf_bitpattern[offset + dummy[1]], 16, pattern);
|
|
strcat(pattern, "0");
|
|
if (cc_width == 3) {
|
|
bin_append(rap_centre[CentreRAP - 1], 10, pattern);
|
|
}
|
|
if (cc_width >= 2) {
|
|
bin_append(pdf_bitpattern[offset + dummy[2]], 16, pattern);
|
|
strcat(pattern, "0");
|
|
}
|
|
if (cc_width == 4) {
|
|
bin_append(rap_centre[CentreRAP - 1], 10, pattern);
|
|
}
|
|
if (cc_width >= 3) {
|
|
bin_append(pdf_bitpattern[offset + dummy[3]], 16, pattern);
|
|
strcat(pattern, "0");
|
|
}
|
|
if (cc_width == 4) {
|
|
bin_append(pdf_bitpattern[offset + dummy[4]], 16, pattern);
|
|
strcat(pattern, "0");
|
|
}
|
|
bin_append(rap_side[RightRAP - 1], 10, pattern);
|
|
strcat(pattern, "1"); /* stop */
|
|
|
|
/* so now pattern[] holds the string of '1's and '0's. - copy this to the symbol */
|
|
for (loop = 0; loop < (int) strlen(pattern); loop++) {
|
|
if (pattern[loop] == '1') {
|
|
set_module(symbol, i, loop);
|
|
}
|
|
}
|
|
symbol->row_height[i] = 2;
|
|
symbol->rows++;
|
|
symbol->width = strlen(pattern);
|
|
|
|
/* Set up RAPs and Cluster for next row */
|
|
LeftRAP++;
|
|
CentreRAP++;
|
|
RightRAP++;
|
|
Cluster++;
|
|
|
|
if (LeftRAP == 53) {
|
|
LeftRAP = 1;
|
|
}
|
|
if (CentreRAP == 53) {
|
|
CentreRAP = 1;
|
|
}
|
|
if (RightRAP == 53) {
|
|
RightRAP = 1;
|
|
}
|
|
if (Cluster == 3) {
|
|
Cluster = 0;
|
|
}
|
|
}
|
|
|
|
return 0;
|
|
}
|
|
|
|
/* CC-B 2D component */
|
|
static int cc_b(struct zint_symbol *symbol, char source[], int cc_width) {
|
|
int length, i;
|
|
#ifndef _MSC_VER
|
|
unsigned char data_string[(strlen(source) / 8) + 3];
|
|
#else
|
|
unsigned char* data_string = (unsigned char*) _alloca((strlen(source) / 8) + 3);
|
|
#endif
|
|
int chainemc[180], mclength;
|
|
int k, j, p, longueur, mccorrection[50], offset;
|
|
int total, dummy[5];
|
|
char pattern[580];
|
|
int variant, LeftRAPStart, CentreRAPStart, RightRAPStart, StartCluster;
|
|
int LeftRAP, CentreRAP, RightRAP, Cluster, loop;
|
|
|
|
length = strlen(source) / 8;
|
|
|
|
for (i = 0; i < length; i++) {
|
|
int binloc = i * 8;
|
|
|
|
data_string[i] = 0;
|
|
for (p = 0; p < 8; p++) {
|
|
if (source[binloc + p] == '1') {
|
|
data_string[i] += (0x80 >> p);
|
|
}
|
|
}
|
|
}
|
|
|
|
|
|
mclength = 0;
|
|
|
|
/* "the CC-B component shall have codeword 920 in the first symbol character position" (section 9a) */
|
|
chainemc[mclength] = 920;
|
|
mclength++;
|
|
|
|
byteprocess(chainemc, &mclength, data_string, 0, length);
|
|
|
|
/* Now figure out which variant of the symbol to use and load values accordingly */
|
|
|
|
variant = 0;
|
|
|
|
if (cc_width == 2) {
|
|
variant = 13;
|
|
if (mclength <= 33) {
|
|
variant = 12;
|
|
}
|
|
if (mclength <= 29) {
|
|
variant = 11;
|
|
}
|
|
if (mclength <= 24) {
|
|
variant = 10;
|
|
}
|
|
if (mclength <= 19) {
|
|
variant = 9;
|
|
}
|
|
if (mclength <= 13) {
|
|
variant = 8;
|
|
}
|
|
if (mclength <= 8) {
|
|
variant = 7;
|
|
}
|
|
}
|
|
|
|
if (cc_width == 3) {
|
|
variant = 23;
|
|
if (mclength <= 70) {
|
|
variant = 22;
|
|
}
|
|
if (mclength <= 58) {
|
|
variant = 21;
|
|
}
|
|
if (mclength <= 46) {
|
|
variant = 20;
|
|
}
|
|
if (mclength <= 34) {
|
|
variant = 19;
|
|
}
|
|
if (mclength <= 24) {
|
|
variant = 18;
|
|
}
|
|
if (mclength <= 18) {
|
|
variant = 17;
|
|
}
|
|
if (mclength <= 14) {
|
|
variant = 16;
|
|
}
|
|
if (mclength <= 10) {
|
|
variant = 15;
|
|
}
|
|
if (mclength <= 6) {
|
|
variant = 14;
|
|
}
|
|
}
|
|
|
|
if (cc_width == 4) {
|
|
variant = 34;
|
|
if (mclength <= 108) {
|
|
variant = 33;
|
|
}
|
|
if (mclength <= 90) {
|
|
variant = 32;
|
|
}
|
|
if (mclength <= 72) {
|
|
variant = 31;
|
|
}
|
|
if (mclength <= 54) {
|
|
variant = 30;
|
|
}
|
|
if (mclength <= 39) {
|
|
variant = 29;
|
|
}
|
|
if (mclength <= 30) {
|
|
variant = 28;
|
|
}
|
|
if (mclength <= 24) {
|
|
variant = 27;
|
|
}
|
|
if (mclength <= 18) {
|
|
variant = 26;
|
|
}
|
|
if (mclength <= 12) {
|
|
variant = 25;
|
|
}
|
|
if (mclength <= 8) {
|
|
variant = 24;
|
|
}
|
|
}
|
|
|
|
/* Now we have the variant we can load the data - from here on the same as MicroPDF417 code */
|
|
variant--;
|
|
assert(variant >= 0);
|
|
symbol->option_2 = MicroVariants[variant]; /* columns */
|
|
symbol->rows = MicroVariants[variant + 34]; /* rows */
|
|
k = MicroVariants[variant + 68]; /* number of EC CWs */
|
|
longueur = (symbol->option_2 * symbol->rows) - k; /* number of non-EC CWs */
|
|
i = longueur - mclength; /* amount of padding required */
|
|
offset = MicroVariants[variant + 102]; /* coefficient offset */
|
|
|
|
/* We add the padding */
|
|
while (i > 0) {
|
|
chainemc[mclength] = 900;
|
|
mclength++;
|
|
i--;
|
|
}
|
|
|
|
/* Reed-Solomon error correction */
|
|
longueur = mclength;
|
|
for (loop = 0; loop < 50; loop++) {
|
|
mccorrection[loop] = 0;
|
|
}
|
|
total = 0;
|
|
for (i = 0; i < longueur; i++) {
|
|
total = (chainemc[i] + mccorrection[k - 1]) % 929;
|
|
for (j = k - 1; j >= 0; j--) {
|
|
if (j == 0) {
|
|
mccorrection[j] = (929 - (total * Microcoeffs[offset + j]) % 929) % 929;
|
|
} else {
|
|
mccorrection[j] = (mccorrection[j - 1] + 929 - (total * Microcoeffs[offset + j]) % 929) % 929;
|
|
}
|
|
}
|
|
}
|
|
|
|
for (j = 0; j < k; j++) {
|
|
if (mccorrection[j] != 0) {
|
|
mccorrection[j] = 929 - mccorrection[j];
|
|
}
|
|
}
|
|
/* we add these codes to the string */
|
|
for (i = k - 1; i >= 0; i--) {
|
|
chainemc[mclength] = mccorrection[i];
|
|
mclength++;
|
|
}
|
|
|
|
/* Now get the RAP (Row Address Pattern) start values */
|
|
LeftRAPStart = RAPTable[variant];
|
|
CentreRAPStart = RAPTable[variant + 34];
|
|
RightRAPStart = RAPTable[variant + 68];
|
|
StartCluster = RAPTable[variant + 102] / 3;
|
|
|
|
/* That's all values loaded, get on with the encoding */
|
|
|
|
LeftRAP = LeftRAPStart;
|
|
CentreRAP = CentreRAPStart;
|
|
RightRAP = RightRAPStart;
|
|
Cluster = StartCluster;
|
|
/* Cluster can be 0, 1 or 2 for Cluster(0), Cluster(3) and Cluster(6) */
|
|
|
|
for (i = 0; i < symbol->rows; i++) {
|
|
strcpy(pattern, "");
|
|
offset = 929 * Cluster;
|
|
for (j = 0; j < 5; j++) {
|
|
dummy[j] = 0;
|
|
}
|
|
for (j = 0; j < symbol->option_2; j++) {
|
|
dummy[j + 1] = chainemc[i * symbol->option_2 + j];
|
|
}
|
|
/* Copy the data into codebarre */
|
|
bin_append(rap_side[LeftRAP - 1], 10, pattern);
|
|
bin_append(pdf_bitpattern[offset + dummy[1]], 16, pattern);
|
|
strcat(pattern, "0");
|
|
if (cc_width == 3) {
|
|
bin_append(rap_centre[CentreRAP - 1], 10, pattern);
|
|
}
|
|
if (cc_width >= 2) {
|
|
bin_append(pdf_bitpattern[offset + dummy[2]], 16, pattern);
|
|
strcat(pattern, "0");
|
|
}
|
|
if (cc_width == 4) {
|
|
bin_append(rap_centre[CentreRAP - 1], 10, pattern);
|
|
}
|
|
if (cc_width >= 3) {
|
|
bin_append(pdf_bitpattern[offset + dummy[3]], 16, pattern);
|
|
strcat(pattern, "0");
|
|
}
|
|
if (cc_width == 4) {
|
|
bin_append(pdf_bitpattern[offset + dummy[4]], 16, pattern);
|
|
strcat(pattern, "0");
|
|
}
|
|
bin_append(rap_side[RightRAP - 1], 10, pattern);
|
|
strcat(pattern, "1"); /* stop */
|
|
|
|
/* so now pattern[] holds the string of '1's and '0's. - copy this to the symbol */
|
|
for (loop = 0; loop < (int) strlen(pattern); loop++) {
|
|
if (pattern[loop] == '1') {
|
|
set_module(symbol, i, loop);
|
|
}
|
|
}
|
|
symbol->row_height[i] = 2;
|
|
symbol->width = strlen(pattern);
|
|
|
|
/* Set up RAPs and Cluster for next row */
|
|
LeftRAP++;
|
|
CentreRAP++;
|
|
RightRAP++;
|
|
Cluster++;
|
|
|
|
if (LeftRAP == 53) {
|
|
LeftRAP = 1;
|
|
}
|
|
if (CentreRAP == 53) {
|
|
CentreRAP = 1;
|
|
}
|
|
if (RightRAP == 53) {
|
|
RightRAP = 1;
|
|
}
|
|
if (Cluster == 3) {
|
|
Cluster = 0;
|
|
}
|
|
}
|
|
|
|
return 0;
|
|
}
|
|
|
|
/* CC-C 2D component - byte compressed PDF417 */
|
|
static int cc_c(struct zint_symbol *symbol, char source[], int cc_width, int ecc_level) {
|
|
int length, i, p;
|
|
#ifndef _MSC_VER
|
|
unsigned char data_string[(strlen(source) / 8) + 4];
|
|
#else
|
|
unsigned char* data_string = (unsigned char*) _alloca((strlen(source) / 8) + 4);
|
|
#endif
|
|
int chainemc[1000], mclength, k;
|
|
int offset, longueur, loop, total, j, mccorrection[520];
|
|
int c1, c2, c3, dummy[35];
|
|
char pattern[580];
|
|
|
|
length = strlen(source) / 8;
|
|
|
|
for (i = 0; i < length; i++) {
|
|
int binloc = i * 8;
|
|
|
|
data_string[i] = 0;
|
|
for (p = 0; p < 8; p++) {
|
|
if (source[binloc + p] == '1') {
|
|
data_string[i] += (0x80 >> p);
|
|
}
|
|
}
|
|
}
|
|
|
|
mclength = 0;
|
|
|
|
chainemc[mclength] = 0; /* space for length descriptor */
|
|
mclength++;
|
|
chainemc[mclength] = 920; /* CC-C identifier */
|
|
mclength++;
|
|
|
|
byteprocess(chainemc, &mclength, data_string, 0, length);
|
|
|
|
chainemc[0] = mclength;
|
|
|
|
k = 1;
|
|
for (i = 1; i <= (ecc_level + 1); i++) {
|
|
k *= 2;
|
|
}
|
|
|
|
/* 796 - we now take care of the Reed Solomon codes */
|
|
switch (ecc_level) {
|
|
case 1: offset = 2;
|
|
break;
|
|
case 2: offset = 6;
|
|
break;
|
|
case 3: offset = 14;
|
|
break;
|
|
case 4: offset = 30;
|
|
break;
|
|
case 5: offset = 62;
|
|
break;
|
|
case 6: offset = 126;
|
|
break;
|
|
case 7: offset = 254;
|
|
break;
|
|
case 8: offset = 510;
|
|
break;
|
|
default: offset = 0;
|
|
break;
|
|
}
|
|
|
|
longueur = mclength;
|
|
for (loop = 0; loop < 520; loop++) {
|
|
mccorrection[loop] = 0;
|
|
}
|
|
total = 0;
|
|
for (i = 0; i < longueur; i++) {
|
|
total = (chainemc[i] + mccorrection[k - 1]) % 929;
|
|
for (j = k - 1; j >= 0; j--) {
|
|
if (j == 0) {
|
|
mccorrection[j] = (929 - (total * coefrs[offset + j]) % 929) % 929;
|
|
} else {
|
|
mccorrection[j] = (mccorrection[j - 1] + 929 - (total * coefrs[offset + j]) % 929) % 929;
|
|
}
|
|
}
|
|
}
|
|
|
|
for (j = 0; j < k; j++) {
|
|
if (mccorrection[j] != 0) {
|
|
mccorrection[j] = 929 - mccorrection[j];
|
|
}
|
|
}
|
|
/* we add these codes to the string */
|
|
for (i = k - 1; i >= 0; i--) {
|
|
chainemc[mclength] = mccorrection[i];
|
|
mclength++;
|
|
}
|
|
|
|
/* 818 - The CW string is finished */
|
|
c1 = (mclength / cc_width - 1) / 3;
|
|
c2 = ecc_level * 3 + (mclength / cc_width - 1) % 3;
|
|
c3 = cc_width - 1;
|
|
|
|
/* we now encode each row */
|
|
for (i = 0; i <= (mclength / cc_width) - 1; i++) {
|
|
for (j = 0; j < cc_width; j++) {
|
|
dummy[j + 1] = chainemc[i * cc_width + j];
|
|
}
|
|
k = (i / 3) * 30;
|
|
switch (i % 3) {
|
|
case 0:
|
|
dummy[0] = k + c1;
|
|
dummy[cc_width + 1] = k + c3;
|
|
offset = 0; /* cluster(0) */
|
|
break;
|
|
case 1:
|
|
dummy[0] = k + c2;
|
|
dummy[cc_width + 1] = k + c1;
|
|
offset = 929; /* cluster(3) */
|
|
break;
|
|
case 2:
|
|
dummy[0] = k + c3;
|
|
dummy[cc_width + 1] = k + c2;
|
|
offset = 1858; /* cluster(6) */
|
|
break;
|
|
}
|
|
strcpy(pattern, "");
|
|
bin_append(0x1FEA8, 17, pattern); /* Row start */
|
|
|
|
for (j = 0; j <= cc_width + 1; j++) {
|
|
bin_append(pdf_bitpattern[offset + dummy[j]], 16, pattern);
|
|
strcat(pattern, "0");
|
|
}
|
|
bin_append(0x3FA29, 18, pattern); /* Row Stop */
|
|
|
|
for (loop = 0; loop < (int) strlen(pattern); loop++) {
|
|
if (pattern[loop] == '1') {
|
|
set_module(symbol, i, loop);
|
|
}
|
|
}
|
|
symbol->row_height[i] = 3;
|
|
}
|
|
symbol->rows = (mclength / cc_width);
|
|
symbol->width = (int)strlen(pattern);
|
|
|
|
return 0;
|
|
}
|
|
|
|
static int calc_padding_cca(int binary_length, int cc_width) {
|
|
int target_bitsize = 0;
|
|
|
|
switch (cc_width) {
|
|
case 2:
|
|
if (binary_length <= 167) {
|
|
target_bitsize = 167;
|
|
}
|
|
if (binary_length <= 138) {
|
|
target_bitsize = 138;
|
|
}
|
|
if (binary_length <= 118) {
|
|
target_bitsize = 118;
|
|
}
|
|
if (binary_length <= 108) {
|
|
target_bitsize = 108;
|
|
}
|
|
if (binary_length <= 88) {
|
|
target_bitsize = 88;
|
|
}
|
|
if (binary_length <= 78) {
|
|
target_bitsize = 78;
|
|
}
|
|
if (binary_length <= 59) {
|
|
target_bitsize = 59;
|
|
}
|
|
break;
|
|
case 3:
|
|
if (binary_length <= 167) {
|
|
target_bitsize = 167;
|
|
}
|
|
if (binary_length <= 138) {
|
|
target_bitsize = 138;
|
|
}
|
|
if (binary_length <= 118) {
|
|
target_bitsize = 118;
|
|
}
|
|
if (binary_length <= 98) {
|
|
target_bitsize = 98;
|
|
}
|
|
if (binary_length <= 78) {
|
|
target_bitsize = 78;
|
|
}
|
|
break;
|
|
case 4:
|
|
if (binary_length <= 197) {
|
|
target_bitsize = 197;
|
|
}
|
|
if (binary_length <= 167) {
|
|
target_bitsize = 167;
|
|
}
|
|
if (binary_length <= 138) {
|
|
target_bitsize = 138;
|
|
}
|
|
if (binary_length <= 108) {
|
|
target_bitsize = 108;
|
|
}
|
|
if (binary_length <= 78) {
|
|
target_bitsize = 78;
|
|
}
|
|
break;
|
|
}
|
|
|
|
return target_bitsize;
|
|
}
|
|
|
|
int calc_padding_ccb(int binary_length, int cc_width) {
|
|
int target_bitsize = 0;
|
|
|
|
switch (cc_width) {
|
|
case 2:
|
|
if (binary_length <= 336) {
|
|
target_bitsize = 336;
|
|
}
|
|
if (binary_length <= 296) {
|
|
target_bitsize = 296;
|
|
}
|
|
if (binary_length <= 256) {
|
|
target_bitsize = 256;
|
|
}
|
|
if (binary_length <= 208) {
|
|
target_bitsize = 208;
|
|
}
|
|
if (binary_length <= 160) {
|
|
target_bitsize = 160;
|
|
}
|
|
if (binary_length <= 104) {
|
|
target_bitsize = 104;
|
|
}
|
|
if (binary_length <= 56) {
|
|
target_bitsize = 56;
|
|
}
|
|
break;
|
|
case 3:
|
|
if (binary_length <= 768) {
|
|
target_bitsize = 768;
|
|
}
|
|
if (binary_length <= 648) {
|
|
target_bitsize = 648;
|
|
}
|
|
if (binary_length <= 536) {
|
|
target_bitsize = 536;
|
|
}
|
|
if (binary_length <= 416) {
|
|
target_bitsize = 416;
|
|
}
|
|
if (binary_length <= 304) {
|
|
target_bitsize = 304;
|
|
}
|
|
if (binary_length <= 208) {
|
|
target_bitsize = 208;
|
|
}
|
|
if (binary_length <= 152) {
|
|
target_bitsize = 152;
|
|
}
|
|
if (binary_length <= 112) {
|
|
target_bitsize = 112;
|
|
}
|
|
if (binary_length <= 72) {
|
|
target_bitsize = 72;
|
|
}
|
|
if (binary_length <= 32) {
|
|
target_bitsize = 32;
|
|
}
|
|
break;
|
|
case 4:
|
|
if (binary_length <= 1184) {
|
|
target_bitsize = 1184;
|
|
}
|
|
if (binary_length <= 1016) {
|
|
target_bitsize = 1016;
|
|
}
|
|
if (binary_length <= 840) {
|
|
target_bitsize = 840;
|
|
}
|
|
if (binary_length <= 672) {
|
|
target_bitsize = 672;
|
|
}
|
|
if (binary_length <= 496) {
|
|
target_bitsize = 496;
|
|
}
|
|
if (binary_length <= 352) {
|
|
target_bitsize = 352;
|
|
}
|
|
if (binary_length <= 264) {
|
|
target_bitsize = 264;
|
|
}
|
|
if (binary_length <= 208) {
|
|
target_bitsize = 208;
|
|
}
|
|
if (binary_length <= 152) {
|
|
target_bitsize = 152;
|
|
}
|
|
if (binary_length <= 96) {
|
|
target_bitsize = 96;
|
|
}
|
|
if (binary_length <= 56) {
|
|
target_bitsize = 56;
|
|
}
|
|
break;
|
|
}
|
|
|
|
return target_bitsize;
|
|
}
|
|
|
|
int calc_padding_ccc(int binary_length, int *cc_width, int lin_width, int *ecc) {
|
|
int target_bitsize = 0;
|
|
int byte_length, codewords_used, ecc_level, ecc_codewords, rows;
|
|
int codewords_total, target_codewords, target_bytesize;
|
|
int i;
|
|
|
|
byte_length = binary_length / 8;
|
|
if (binary_length % 8 != 0) {
|
|
byte_length++;
|
|
}
|
|
|
|
codewords_used = (byte_length / 6) * 5;
|
|
codewords_used += byte_length % 6;
|
|
|
|
ecc_level = 7;
|
|
if (codewords_used <= 1280) {
|
|
ecc_level = 6;
|
|
}
|
|
if (codewords_used <= 640) {
|
|
ecc_level = 5;
|
|
}
|
|
if (codewords_used <= 320) {
|
|
ecc_level = 4;
|
|
}
|
|
if (codewords_used <= 160) {
|
|
ecc_level = 3;
|
|
}
|
|
if (codewords_used <= 40) {
|
|
ecc_level = 2;
|
|
}
|
|
*(ecc) = ecc_level;
|
|
ecc_codewords = 1;
|
|
for (i = 1; i <= (ecc_level + 1); i++) {
|
|
ecc_codewords *= 2;
|
|
}
|
|
|
|
codewords_used += ecc_codewords;
|
|
codewords_used += 3;
|
|
|
|
*(cc_width) = (lin_width - 62) / 17;
|
|
/* stop the symbol from becoming too high */
|
|
do {
|
|
*(cc_width) = *(cc_width) + 1;
|
|
rows = codewords_used / *(cc_width);
|
|
} while (rows > 90);
|
|
|
|
if (codewords_used % *(cc_width) != 0) {
|
|
rows++;
|
|
}
|
|
|
|
codewords_total = *(cc_width) * rows;
|
|
|
|
if (codewords_total > 928) { // PDF_MAX
|
|
return 0;
|
|
}
|
|
|
|
target_codewords = codewords_total - ecc_codewords;
|
|
target_codewords -= 3;
|
|
|
|
target_bytesize = 6 * (target_codewords / 5);
|
|
target_bytesize += target_codewords % 5;
|
|
|
|
target_bitsize = 8 * target_bytesize;
|
|
|
|
return target_bitsize;
|
|
}
|
|
|
|
static int cc_binary_string(struct zint_symbol *symbol, const char source[], char binary_string[], int cc_mode, int *cc_width, int *ecc, int lin_width) { /* Handles all data encodation from section 5 of ISO/IEC 24723 */
|
|
int encoding_method, read_posn, d1, d2, alpha_pad;
|
|
int i, j, ai_crop, fnc1_latch;
|
|
int ai90_mode, latch, remainder, binary_length;
|
|
#ifndef _MSC_VER
|
|
char general_field[strlen(source) + 1], general_field_type[strlen(source) + 1];
|
|
#else
|
|
char* general_field = (char*) _alloca(strlen(source) + 1);
|
|
char* general_field_type = (char*) _alloca(strlen(source) + 1);
|
|
#endif
|
|
int target_bitsize;
|
|
|
|
encoding_method = 1;
|
|
read_posn = 0;
|
|
ai_crop = 0;
|
|
fnc1_latch = 0;
|
|
alpha_pad = 0;
|
|
ai90_mode = 0;
|
|
*ecc = 0;
|
|
target_bitsize = 0;
|
|
|
|
if ((source[0] == '1') && ((source[1] == '0') || (source[1] == '1') || (source[1] == '7')) && (strlen(source) > 8)) {
|
|
/* Source starts (10), (11) or (17) */
|
|
encoding_method = 2;
|
|
}
|
|
|
|
if ((source[0] == '9') && (source[1] == '0')) {
|
|
/* Source starts (90) */
|
|
encoding_method = 3;
|
|
}
|
|
|
|
if (encoding_method == 1) {
|
|
strcat(binary_string, "0");
|
|
}
|
|
|
|
if (encoding_method == 2) {
|
|
/* Encoding Method field "10" - date and lot number */
|
|
|
|
strcat(binary_string, "10");
|
|
|
|
if (source[1] == '0') {
|
|
/* No date data */
|
|
strcat(binary_string, "11");
|
|
read_posn = 2;
|
|
} else {
|
|
long int group_val;
|
|
/* Production Date (11) or Expiration Date (17) */
|
|
char date_str[4];
|
|
date_str[0] = source[2];
|
|
date_str[1] = source[3];
|
|
date_str[2] = '\0';
|
|
group_val = atoi(date_str) * 384;
|
|
|
|
date_str[0] = source[4];
|
|
date_str[1] = source[5];
|
|
group_val += (atoi(date_str) - 1) * 32;
|
|
|
|
date_str[0] = source[6];
|
|
date_str[1] = source[7];
|
|
group_val += atoi(date_str);
|
|
|
|
bin_append(group_val, 16, binary_string);
|
|
|
|
if (source[1] == '1') {
|
|
/* Production Date AI 11 */
|
|
strcat(binary_string, "0");
|
|
} else {
|
|
/* Expiration Date AI 17 */
|
|
strcat(binary_string, "1");
|
|
}
|
|
read_posn = 8;
|
|
}
|
|
|
|
if ((source[read_posn] == '1') && (source[read_posn + 1] == '0')) {
|
|
/* Followed by AI 10 - strip this from general field */
|
|
read_posn += 2;
|
|
} else {
|
|
/* An FNC1 character needs to be inserted in the general field */
|
|
fnc1_latch = 1;
|
|
}
|
|
}
|
|
|
|
if (encoding_method == 3) {
|
|
/* Encodation Method field of "11" - AI 90 */
|
|
#ifndef _MSC_VER
|
|
char ninety[strlen(source) + 1];
|
|
#else
|
|
char* ninety = (char*) _alloca(strlen(source) + 1);
|
|
#endif
|
|
int alpha, alphanum, numeric, test1, test2, test3;
|
|
|
|
/* "This encodation method may be used if an element string with an AI
|
|
90 occurs at the start of the data message, and if the data field
|
|
following the two-digit AI 90 starts with an alphanumeric string which
|
|
complies with a specific format." (para 5.2.2) */
|
|
|
|
i = 0;
|
|
do {
|
|
ninety[i] = source[i + 2];
|
|
i++;
|
|
} while ((strlen(source) > i + 2) && ('[' != source[i + 2]));
|
|
ninety[i] = '\0';
|
|
|
|
/* Find out if the AI 90 data is alphabetic or numeric or both */
|
|
|
|
alpha = 0;
|
|
alphanum = 0;
|
|
numeric = 0;
|
|
|
|
for (i = 0; i < (int) strlen(ninety); i++) {
|
|
|
|
if ((ninety[i] >= 'A') && (ninety[i] <= 'Z')) {
|
|
/* Character is alphabetic */
|
|
alpha += 1;
|
|
}
|
|
|
|
if ((ninety[i] >= '0') && (ninety[i] <= '9')) {
|
|
/* Character is numeric */
|
|
numeric += 1;
|
|
}
|
|
|
|
switch (ninety[i]) {
|
|
case '*':
|
|
case ',':
|
|
case '-':
|
|
case '.':
|
|
case '/': alphanum += 1;
|
|
break;
|
|
}
|
|
|
|
if (!(((ninety[i] >= '0') && (ninety[i] <= '9')) || ((ninety[i] >= 'A') && (ninety[i] <= 'Z')))) {
|
|
if ((ninety[i] != '*') && (ninety[i] != ',') && (ninety[i] != '-') && (ninety[i] != '.') && (ninety[i] != '/')) {
|
|
/* An Invalid AI 90 character */
|
|
strcpy(symbol->errtxt, "440: Invalid AI 90 data");
|
|
return ZINT_ERROR_INVALID_DATA;
|
|
}
|
|
}
|
|
}
|
|
|
|
/* must start with 0, 1, 2 or 3 digits followed by an uppercase character */
|
|
test1 = -1;
|
|
for (i = 3; i >= 0; i--) {
|
|
if ((ninety[i] >= 'A') && (ninety[i] <= 'Z')) {
|
|
test1 = i;
|
|
}
|
|
}
|
|
|
|
test2 = 0;
|
|
for (i = 0; i < test1; i++) {
|
|
if (!((ninety[i] >= '0') && (ninety[i] <= '9'))) {
|
|
test2 = 1;
|
|
}
|
|
}
|
|
|
|
/* leading zeros are not permitted */
|
|
test3 = 0;
|
|
if ((test1 >= 1) && (ninety[0] == '0')) {
|
|
test3 = 1;
|
|
}
|
|
|
|
if ((test1 != -1) && (test2 != 1) && (test3 == 0)) {
|
|
int next_ai_posn;
|
|
char numeric_part[4];
|
|
int numeric_value;
|
|
int table3_letter;
|
|
/* Encodation method "11" can be used */
|
|
strcat(binary_string, "11");
|
|
|
|
numeric -= test1;
|
|
alpha--;
|
|
|
|
/* Decide on numeric, alpha or alphanumeric mode */
|
|
/* Alpha mode is a special mode for AI 90 */
|
|
|
|
if (alphanum > 0) {
|
|
/* Alphanumeric mode */
|
|
strcat(binary_string, "0");
|
|
ai90_mode = 1;
|
|
} else {
|
|
if (alpha > numeric) {
|
|
/* Alphabetic mode */
|
|
strcat(binary_string, "11");
|
|
ai90_mode = 2;
|
|
} else {
|
|
/* Numeric mode */
|
|
strcat(binary_string, "10");
|
|
ai90_mode = 3;
|
|
}
|
|
}
|
|
|
|
next_ai_posn = 2 + (int)strlen(ninety);
|
|
|
|
if (source[next_ai_posn] == '[') {
|
|
/* There are more AIs afterwords */
|
|
if ((source[next_ai_posn + 1] == '2') && (source[next_ai_posn + 2] == '1')) {
|
|
/* AI 21 follows */
|
|
ai_crop = 1;
|
|
}
|
|
|
|
if ((source[next_ai_posn + 1] == '8') && (source[next_ai_posn + 2] == '0') && (source[next_ai_posn + 3] == '0') && (source[next_ai_posn + 4] == '4')) {
|
|
/* AI 8004 follows */
|
|
ai_crop = 2;
|
|
}
|
|
}
|
|
|
|
switch (ai_crop) {
|
|
case 0: strcat(binary_string, "0");
|
|
break;
|
|
case 1: strcat(binary_string, "10");
|
|
break;
|
|
case 2: strcat(binary_string, "11");
|
|
break;
|
|
}
|
|
|
|
if (test1 == 0) {
|
|
strcpy(numeric_part, "0");
|
|
} else {
|
|
for (i = 0; i < test1; i++) {
|
|
numeric_part[i] = ninety[i];
|
|
}
|
|
numeric_part[i] = '\0';
|
|
}
|
|
|
|
numeric_value = atoi(numeric_part);
|
|
|
|
table3_letter = -1;
|
|
if (numeric_value < 31) {
|
|
table3_letter = posn("BDHIJKLNPQRSTVWZ", ninety[test1]);
|
|
}
|
|
|
|
if (table3_letter != -1) {
|
|
/* Encoding can be done according to 5.2.2 c) 2) */
|
|
/* five bit binary string representing value before letter */
|
|
bin_append(numeric_value, 5, binary_string);
|
|
|
|
/* followed by four bit representation of letter from Table 3 */
|
|
bin_append(table3_letter, 4, binary_string);
|
|
} else {
|
|
/* Encoding is done according to 5.2.2 c) 3) */
|
|
bin_append(31, 5, binary_string);
|
|
/* ten bit representation of number */
|
|
bin_append(numeric_value, 10, binary_string);
|
|
|
|
/* five bit representation of ASCII character */
|
|
bin_append(ninety[test1] - 65, 5, binary_string);
|
|
}
|
|
|
|
read_posn = test1 + 3;
|
|
} else {
|
|
/* Use general field encodation instead */
|
|
strcat(binary_string, "0");
|
|
read_posn = 0;
|
|
}
|
|
}
|
|
|
|
/* Now encode the rest of the AI 90 data field */
|
|
if (ai90_mode == 2) {
|
|
/* Alpha encodation (section 5.2.3) */
|
|
do {
|
|
if ((source[read_posn] >= '0') && (source[read_posn] <= '9')) {
|
|
bin_append(source[read_posn] + 4, 5, binary_string);
|
|
}
|
|
|
|
if ((source[read_posn] >= 'A') && (source[read_posn] <= 'Z')) {
|
|
bin_append(source[read_posn] - 65, 6, binary_string);
|
|
}
|
|
|
|
if (source[read_posn] == '[') {
|
|
bin_append(31, 5, binary_string);
|
|
}
|
|
|
|
read_posn++;
|
|
} while ((source[read_posn - 1] != '[') && (source[read_posn - 1] != '\0'));
|
|
alpha_pad = 1; /* This is overwritten if a general field is encoded */
|
|
}
|
|
|
|
if (ai90_mode == 1) {
|
|
/* Alphanumeric mode */
|
|
do {
|
|
if ((source[read_posn] >= '0') && (source[read_posn] <= '9')) {
|
|
bin_append(source[read_posn] - 43, 5, binary_string);
|
|
}
|
|
|
|
if ((source[read_posn] >= 'A') && (source[read_posn] <= 'Z')) {
|
|
bin_append(source[read_posn] - 33, 6, binary_string);
|
|
}
|
|
|
|
switch (source[read_posn]) {
|
|
case '[':
|
|
bin_append(15, 5, binary_string);
|
|
break;
|
|
case '*':
|
|
bin_append(58, 6, binary_string);
|
|
break;
|
|
case ',':
|
|
bin_append(59, 6, binary_string);
|
|
break;
|
|
case '-':
|
|
bin_append(60, 6, binary_string);
|
|
break;
|
|
case '.':
|
|
bin_append(61, 6, binary_string);
|
|
break;
|
|
case '/':
|
|
bin_append(62, 6, binary_string);
|
|
break;
|
|
}
|
|
|
|
read_posn++;
|
|
} while ((source[read_posn - 1] != '[') && (source[read_posn - 1] != '\0'));
|
|
}
|
|
|
|
read_posn += (2 * ai_crop);
|
|
|
|
/* The compressed data field has been processed if appropriate - the
|
|
rest of the data (if any) goes into a general-purpose data compaction field */
|
|
|
|
j = 0;
|
|
if (fnc1_latch == 1) {
|
|
/* Encodation method "10" has been used but it is not followed by
|
|
AI 10, so a FNC1 character needs to be added */
|
|
general_field[j] = '[';
|
|
j++;
|
|
}
|
|
|
|
for (i = read_posn; i < (int) strlen(source); i++) {
|
|
general_field[j] = source[i];
|
|
j++;
|
|
}
|
|
general_field[j] = '\0';
|
|
|
|
if (strlen(general_field) != 0) {
|
|
alpha_pad = 0;
|
|
}
|
|
|
|
latch = 0;
|
|
for (i = 0; i < (int) strlen(general_field); i++) {
|
|
/* Table 13 - ISO/IEC 646 encodation */
|
|
if ((general_field[i] < ' ') || (general_field[i] > 'z')) {
|
|
general_field_type[i] = INVALID_CHAR;
|
|
latch = 1;
|
|
} else {
|
|
general_field_type[i] = ISOIEC;
|
|
}
|
|
|
|
if (general_field[i] == '#') {
|
|
general_field_type[i] = INVALID_CHAR;
|
|
latch = 1;
|
|
}
|
|
if (general_field[i] == '$') {
|
|
general_field_type[i] = INVALID_CHAR;
|
|
latch = 1;
|
|
}
|
|
if (general_field[i] == '@') {
|
|
general_field_type[i] = INVALID_CHAR;
|
|
latch = 1;
|
|
}
|
|
if (general_field[i] == 92) {
|
|
general_field_type[i] = INVALID_CHAR;
|
|
latch = 1;
|
|
}
|
|
if (general_field[i] == '^') {
|
|
general_field_type[i] = INVALID_CHAR;
|
|
latch = 1;
|
|
}
|
|
if (general_field[i] == 96) {
|
|
general_field_type[i] = INVALID_CHAR;
|
|
latch = 1;
|
|
}
|
|
|
|
/* Table 12 - Alphanumeric encodation */
|
|
if ((general_field[i] >= 'A') && (general_field[i] <= 'Z')) {
|
|
general_field_type[i] = ALPHA_OR_ISO;
|
|
}
|
|
if (general_field[i] == '*') {
|
|
general_field_type[i] = ALPHA_OR_ISO;
|
|
}
|
|
if (general_field[i] == ',') {
|
|
general_field_type[i] = ALPHA_OR_ISO;
|
|
}
|
|
if (general_field[i] == '-') {
|
|
general_field_type[i] = ALPHA_OR_ISO;
|
|
}
|
|
if (general_field[i] == '.') {
|
|
general_field_type[i] = ALPHA_OR_ISO;
|
|
}
|
|
if (general_field[i] == '/') {
|
|
general_field_type[i] = ALPHA_OR_ISO;
|
|
}
|
|
|
|
/* Numeric encodation */
|
|
if ((general_field[i] >= '0') && (general_field[i] <= '9')) {
|
|
general_field_type[i] = ANY_ENC;
|
|
}
|
|
if (general_field[i] == '[') {
|
|
/* FNC1 can be encoded in any system */
|
|
general_field_type[i] = ANY_ENC;
|
|
}
|
|
|
|
}
|
|
|
|
general_field_type[strlen(general_field)] = '\0';
|
|
|
|
if (latch == 1) {
|
|
/* Invalid characters in input data */
|
|
strcpy(symbol->errtxt, "441: Invalid characters in input data");
|
|
return ZINT_ERROR_INVALID_DATA;
|
|
}
|
|
|
|
for (i = 0; i < (int) strlen(general_field); i++) {
|
|
if ((general_field_type[i] == ISOIEC) && (general_field[i + 1] == '[')) {
|
|
general_field_type[i + 1] = ISOIEC;
|
|
}
|
|
}
|
|
|
|
for (i = 0; i < (int) strlen(general_field); i++) {
|
|
if ((general_field_type[i] == ALPHA_OR_ISO) && (general_field[i + 1] == '[')) {
|
|
general_field_type[i + 1] = ALPHA_OR_ISO;
|
|
}
|
|
}
|
|
|
|
latch = general_rules(general_field_type);
|
|
|
|
i = 0;
|
|
do {
|
|
switch (general_field_type[i]) {
|
|
case NUMERIC:
|
|
|
|
if (i != 0) {
|
|
if ((general_field_type[i - 1] != NUMERIC) && (general_field[i - 1] != '[')) {
|
|
bin_append(0, 3, binary_string); /* Numeric latch */
|
|
}
|
|
}
|
|
|
|
if (general_field[i] != '[') {
|
|
d1 = ctoi(general_field[i]);
|
|
} else {
|
|
d1 = 10;
|
|
}
|
|
|
|
if (general_field[i + 1] != '[') {
|
|
d2 = ctoi(general_field[i + 1]);
|
|
} else {
|
|
d2 = 10;
|
|
}
|
|
|
|
bin_append((11 * d1) + d2 + 8, 7, binary_string);
|
|
|
|
i += 2;
|
|
break;
|
|
|
|
case ALPHA:
|
|
|
|
if (i != 0) {
|
|
if ((general_field_type[i - 1] == NUMERIC) || (general_field[i - 1] == '[')) {
|
|
bin_append(0, 4, binary_string); /* Alphanumeric latch */
|
|
}
|
|
if (general_field_type[i - 1] == ISOIEC) {
|
|
bin_append(4, 5, binary_string); /* ISO/IEC 646 latch */
|
|
}
|
|
}
|
|
|
|
if ((general_field[i] >= '0') && (general_field[i] <= '9')) {
|
|
bin_append(general_field[i] - 43, 5, binary_string);
|
|
}
|
|
|
|
if ((general_field[i] >= 'A') && (general_field[i] <= 'Z')) {
|
|
bin_append(general_field[i] - 33, 6, binary_string);
|
|
}
|
|
|
|
switch (general_field[i]) {
|
|
case '[':
|
|
bin_append(15, 5, binary_string);
|
|
break;
|
|
case '*':
|
|
bin_append(58, 6, binary_string);
|
|
break;
|
|
case ',':
|
|
bin_append(59, 6, binary_string);
|
|
break;
|
|
case '-':
|
|
bin_append(60, 6, binary_string);
|
|
break;
|
|
case '.':
|
|
bin_append(61, 6, binary_string);
|
|
break;
|
|
case '/':
|
|
bin_append(62, 6, binary_string);
|
|
break;
|
|
}
|
|
|
|
i++;
|
|
break;
|
|
|
|
case ISOIEC:
|
|
|
|
if (i != 0) {
|
|
if ((general_field_type[i - 1] == NUMERIC) || (general_field[i - 1] == '[')) {
|
|
bin_append(0, 4, binary_string); /* Alphanumeric latch */
|
|
bin_append(4, 5, binary_string); /* ISO/IEC 646 latch */
|
|
}
|
|
if (general_field_type[i - 1] == ALPHA) {
|
|
bin_append(4, 5, binary_string);; /* ISO/IEC 646 latch */
|
|
}
|
|
}
|
|
|
|
if ((general_field[i] >= '0') && (general_field[i] <= '9')) {
|
|
bin_append(general_field[i] - 43, 5, binary_string);
|
|
}
|
|
|
|
if ((general_field[i] >= 'A') && (general_field[i] <= 'Z')) {
|
|
bin_append(general_field[i] - 1, 7, binary_string);
|
|
}
|
|
|
|
if ((general_field[i] >= 'a') && (general_field[i] <= 'z')) {
|
|
bin_append(general_field[i] - 7, 7, binary_string);
|
|
}
|
|
|
|
if (general_field[i] == '[') strcat(binary_string, "01111"); /* FNC1/Numeric latch */
|
|
if (general_field[i] == '!') strcat(binary_string, "11101000"); /* exclamation mark */
|
|
if (general_field[i] == 34) strcat(binary_string, "11101001"); /* quotation mark */
|
|
if (general_field[i] == 37) strcat(binary_string, "11101010"); /* percent sign */
|
|
if (general_field[i] == '&') strcat(binary_string, "11101011"); /* ampersand */
|
|
if (general_field[i] == 39) strcat(binary_string, "11101100"); /* apostrophe */
|
|
if (general_field[i] == '(') strcat(binary_string, "11101101"); /* left parenthesis */
|
|
if (general_field[i] == ')') strcat(binary_string, "11101110"); /* right parenthesis */
|
|
if (general_field[i] == '*') strcat(binary_string, "11101111"); /* asterisk */
|
|
if (general_field[i] == '+') strcat(binary_string, "11110000"); /* plus sign */
|
|
if (general_field[i] == ',') strcat(binary_string, "11110001"); /* comma */
|
|
if (general_field[i] == '-') strcat(binary_string, "11110010"); /* minus or hyphen */
|
|
if (general_field[i] == '.') strcat(binary_string, "11110011"); /* period or full stop */
|
|
if (general_field[i] == '/') strcat(binary_string, "11110100"); /* slash or solidus */
|
|
if (general_field[i] == ':') strcat(binary_string, "11110101"); /* colon */
|
|
if (general_field[i] == ';') strcat(binary_string, "11110110"); /* semicolon */
|
|
if (general_field[i] == '<') strcat(binary_string, "11110111"); /* less-than sign */
|
|
if (general_field[i] == '=') strcat(binary_string, "11111000"); /* equals sign */
|
|
if (general_field[i] == '>') strcat(binary_string, "11111001"); /* greater-than sign */
|
|
if (general_field[i] == '?') strcat(binary_string, "11111010"); /* question mark */
|
|
if (general_field[i] == '_') strcat(binary_string, "11111011"); /* underline or low line */
|
|
if (general_field[i] == ' ') strcat(binary_string, "11111100"); /* space */
|
|
|
|
i++;
|
|
break;
|
|
}
|
|
} while (i + latch < (int) strlen(general_field));
|
|
|
|
binary_length = (int)strlen(binary_string);
|
|
switch (cc_mode) {
|
|
case 1:
|
|
target_bitsize = calc_padding_cca(binary_length, *(cc_width));
|
|
break;
|
|
case 2:
|
|
target_bitsize = calc_padding_ccb(binary_length, *(cc_width));
|
|
break;
|
|
case 3:
|
|
target_bitsize = calc_padding_ccc(binary_length, cc_width, lin_width, ecc);
|
|
break;
|
|
}
|
|
|
|
if (target_bitsize == 0) {
|
|
strcpy(symbol->errtxt, "442: Input too long for selected 2d component");
|
|
return ZINT_ERROR_TOO_LONG;
|
|
}
|
|
|
|
remainder = target_bitsize - binary_length;
|
|
|
|
if (latch == 1) {
|
|
i = 0;
|
|
/* There is still one more numeric digit to encode */
|
|
|
|
if ((remainder >= 4) && (remainder <= 6)) {
|
|
bin_append(ctoi(general_field[i]) + 1, 4, binary_string);
|
|
} else {
|
|
bin_append((11 * ctoi(general_field[i])) + 18, 7, binary_string);
|
|
/* This may push the symbol up to the next size */
|
|
}
|
|
}
|
|
|
|
if (strlen(binary_string) > 11805) { /* (2361 * 5) */
|
|
strcpy(symbol->errtxt, "443: Input too long");
|
|
return ZINT_ERROR_TOO_LONG;
|
|
}
|
|
|
|
|
|
binary_length = (int)strlen(binary_string);
|
|
switch (cc_mode) {
|
|
case 1:
|
|
target_bitsize = calc_padding_cca(binary_length, *(cc_width));
|
|
break;
|
|
case 2:
|
|
target_bitsize = calc_padding_ccb(binary_length, *(cc_width));
|
|
break;
|
|
case 3:
|
|
target_bitsize = calc_padding_ccc(binary_length, cc_width, lin_width, ecc);
|
|
break;
|
|
}
|
|
|
|
if (target_bitsize == 0) {
|
|
strcpy(symbol->errtxt, "444: Input too long for selected 2d component");
|
|
return ZINT_ERROR_TOO_LONG;
|
|
}
|
|
|
|
if (binary_length < target_bitsize) {
|
|
/* Now add padding to binary string */
|
|
if (alpha_pad == 1) {
|
|
strcat(binary_string, "11111");
|
|
alpha_pad = 0;
|
|
/* Extra FNC1 character required after Alpha encodation (section 5.2.3) */
|
|
}
|
|
|
|
if ((strlen(general_field) != 0) && (general_field_type[strlen(general_field) - 1] == NUMERIC)) {
|
|
strcat(binary_string, "0000");
|
|
}
|
|
|
|
while (strlen(binary_string) < (unsigned int) target_bitsize) {
|
|
strcat(binary_string, "00100");
|
|
}
|
|
|
|
if (strlen(binary_string) > (unsigned int) target_bitsize) {
|
|
binary_string[target_bitsize] = '\0';
|
|
}
|
|
}
|
|
|
|
return 0;
|
|
}
|
|
|
|
int linear_dummy_run(unsigned char *source, int length) {
|
|
struct zint_symbol *dummy;
|
|
int error_number;
|
|
int linear_width;
|
|
|
|
dummy = ZBarcode_Create();
|
|
dummy->symbology = BARCODE_EAN128_CC;
|
|
dummy->option_1 = 3;
|
|
error_number = ean_128(dummy, source, length);
|
|
linear_width = dummy->width;
|
|
ZBarcode_Delete(dummy);
|
|
|
|
if (error_number == 0) {
|
|
return linear_width;
|
|
} else {
|
|
return 0;
|
|
}
|
|
}
|
|
|
|
int composite(struct zint_symbol *symbol, unsigned char source[], int length) {
|
|
int error_number, cc_mode, cc_width, ecc_level;
|
|
int j, i, k;
|
|
unsigned int rs = length + 1;
|
|
unsigned int bs = 20 * rs;
|
|
unsigned int pri_len;
|
|
#ifndef _MSC_VER
|
|
char reduced[rs];
|
|
char binary_string[bs];
|
|
#else
|
|
char* reduced = (char*) _alloca(rs);
|
|
char* binary_string = (char*) _alloca(bs);
|
|
#endif
|
|
struct zint_symbol *linear;
|
|
int top_shift, bottom_shift;
|
|
int linear_width = 0;
|
|
|
|
/* Perform sanity checks on input options first */
|
|
error_number = 0;
|
|
pri_len = (int)strlen(symbol->primary);
|
|
if (pri_len == 0) {
|
|
strcpy(symbol->errtxt, "445: No primary (linear) message in 2D composite");
|
|
return ZINT_ERROR_INVALID_OPTION;
|
|
}
|
|
|
|
if (length > 2990) {
|
|
strcpy(symbol->errtxt, "446: 2D component input data too long");
|
|
return ZINT_ERROR_TOO_LONG;
|
|
}
|
|
|
|
cc_mode = symbol->option_1;
|
|
if ((cc_mode == 3) && (symbol->symbology != BARCODE_EAN128_CC)) {
|
|
/* CC-C can only be used with a GS1-128 linear part */
|
|
strcpy(symbol->errtxt, "447: Invalid mode (CC-C only valid with GS1-128 linear component)");
|
|
return ZINT_ERROR_INVALID_OPTION;
|
|
}
|
|
|
|
error_number = gs1_verify(symbol, source, length, reduced);
|
|
if (error_number != 0) {
|
|
return error_number;
|
|
}
|
|
|
|
if (symbol->symbology == BARCODE_EAN128_CC) {
|
|
/* Do a test run of encoding the linear component to establish its width */
|
|
linear_width = linear_dummy_run((unsigned char *) symbol->primary, pri_len);
|
|
if (linear_width == 0) {
|
|
strcpy(symbol->errtxt, "448: Invalid data");
|
|
return ZINT_ERROR_INVALID_DATA;
|
|
}
|
|
}
|
|
|
|
switch (symbol->symbology) {
|
|
/* Determine width of 2D component according to ISO/IEC 24723 Table 1 */
|
|
case BARCODE_EANX_CC:
|
|
switch (pri_len) {
|
|
case 7: /* EAN-8 */
|
|
case 10: /* EAN-8 + 2 */
|
|
case 13: /* EAN-8 + 5 */
|
|
cc_width = 3;
|
|
break;
|
|
case 12: /* EAN-13 */
|
|
case 15: /* EAN-13 + 2 */
|
|
case 18: /* EAN-13 + 5 */
|
|
cc_width = 4;
|
|
break;
|
|
}
|
|
break;
|
|
case BARCODE_EAN128_CC: cc_width = 4;
|
|
break;
|
|
case BARCODE_RSS14_CC: cc_width = 4;
|
|
break;
|
|
case BARCODE_RSS_LTD_CC: cc_width = 3;
|
|
break;
|
|
case BARCODE_RSS_EXP_CC: cc_width = 4;
|
|
break;
|
|
case BARCODE_UPCA_CC: cc_width = 4;
|
|
break;
|
|
case BARCODE_UPCE_CC: cc_width = 2;
|
|
break;
|
|
case BARCODE_RSS14STACK_CC: cc_width = 2;
|
|
break;
|
|
case BARCODE_RSS14_OMNI_CC: cc_width = 2;
|
|
break;
|
|
case BARCODE_RSS_EXPSTACK_CC: cc_width = 4;
|
|
break;
|
|
}
|
|
|
|
memset(binary_string, 0, bs);
|
|
|
|
if (cc_mode < 1 || cc_mode > 3) {
|
|
cc_mode = 1;
|
|
}
|
|
|
|
if (cc_mode == 1) {
|
|
i = cc_binary_string(symbol, reduced, binary_string, cc_mode, &cc_width, &ecc_level, linear_width);
|
|
if (i == ZINT_ERROR_TOO_LONG) {
|
|
cc_mode = 2;
|
|
}
|
|
}
|
|
|
|
if (cc_mode == 2) {
|
|
/* If the data didn't fit into CC-A it is recalculated for CC-B */
|
|
i = cc_binary_string(symbol, reduced, binary_string, cc_mode, &cc_width, &ecc_level, linear_width);
|
|
if (i == ZINT_ERROR_TOO_LONG) {
|
|
if (symbol->symbology != BARCODE_EAN128_CC) {
|
|
return ZINT_ERROR_TOO_LONG;
|
|
} else {
|
|
cc_mode = 3;
|
|
}
|
|
}
|
|
}
|
|
|
|
if (cc_mode == 3) {
|
|
/* If the data didn't fit in CC-B (and linear part is GS1-128) it is recalculated for CC-C */
|
|
i = cc_binary_string(symbol, reduced, binary_string, cc_mode, &cc_width, &ecc_level, linear_width);
|
|
if (i == ZINT_ERROR_TOO_LONG) {
|
|
return ZINT_ERROR_TOO_LONG;
|
|
}
|
|
}
|
|
|
|
switch (cc_mode) {
|
|
/* Note that ecc_level is only relevant to CC-C */
|
|
case 1: error_number = cc_a(symbol, binary_string, cc_width);
|
|
break;
|
|
case 2: error_number = cc_b(symbol, binary_string, cc_width);
|
|
break;
|
|
case 3: error_number = cc_c(symbol, binary_string, cc_width, ecc_level);
|
|
break;
|
|
}
|
|
|
|
if (error_number != 0) {
|
|
return ZINT_ERROR_ENCODING_PROBLEM;
|
|
}
|
|
|
|
/* 2D component done, now calculate linear component */
|
|
linear = ZBarcode_Create(); /* Symbol contains the 2D component and Linear contains the rest */
|
|
|
|
linear->symbology = symbol->symbology;
|
|
|
|
if (linear->symbology != BARCODE_EAN128_CC) {
|
|
/* Set the "component linkage" flag in the linear component */
|
|
linear->option_1 = 2;
|
|
} else {
|
|
/* GS1-128 needs to know which type of 2D component is used */
|
|
linear->option_1 = cc_mode;
|
|
}
|
|
|
|
switch (symbol->symbology) {
|
|
case BARCODE_EANX_CC: error_number = eanx(linear, (unsigned char *) symbol->primary, pri_len);
|
|
break;
|
|
case BARCODE_EAN128_CC: error_number = ean_128(linear, (unsigned char *) symbol->primary, pri_len);
|
|
break;
|
|
case BARCODE_RSS14_CC: error_number = rss14(linear, (unsigned char *) symbol->primary, pri_len);
|
|
break;
|
|
case BARCODE_RSS_LTD_CC: error_number = rsslimited(linear, (unsigned char *) symbol->primary, pri_len);
|
|
break;
|
|
case BARCODE_RSS_EXP_CC: error_number = rssexpanded(linear, (unsigned char *) symbol->primary, pri_len);
|
|
break;
|
|
case BARCODE_UPCA_CC: error_number = eanx(linear, (unsigned char *) symbol->primary, pri_len);
|
|
break;
|
|
case BARCODE_UPCE_CC: error_number = eanx(linear, (unsigned char *) symbol->primary, pri_len);
|
|
break;
|
|
case BARCODE_RSS14STACK_CC: error_number = rss14(linear, (unsigned char *) symbol->primary, pri_len);
|
|
break;
|
|
case BARCODE_RSS14_OMNI_CC: error_number = rss14(linear, (unsigned char *) symbol->primary, pri_len);
|
|
break;
|
|
case BARCODE_RSS_EXPSTACK_CC: error_number = rssexpanded(linear, (unsigned char *) symbol->primary, pri_len);
|
|
break;
|
|
}
|
|
|
|
if (error_number != 0) {
|
|
strcpy(symbol->errtxt, linear->errtxt);
|
|
strcat(symbol->errtxt, " in linear component ");
|
|
ZBarcode_Delete(linear);
|
|
return error_number;
|
|
}
|
|
|
|
/* Merge the linear component with the 2D component */
|
|
|
|
top_shift = 0;
|
|
bottom_shift = 0;
|
|
|
|
switch (symbol->symbology) {
|
|
/* Determine horizontal alignment (according to section 12.3) */
|
|
case BARCODE_EANX_CC:
|
|
switch (pri_len) {
|
|
case 7: /* EAN-8 */
|
|
case 10: /* EAN-8 + 2 */
|
|
case 13: /* EAN-8 + 5 */
|
|
bottom_shift = 13;
|
|
break;
|
|
case 12: /* EAN-13 */
|
|
case 15: /* EAN-13 + 2 */
|
|
case 18: /* EAN-13 + 5 */
|
|
bottom_shift = 2;
|
|
break;
|
|
}
|
|
break;
|
|
case BARCODE_EAN128_CC: if (cc_mode == 3) {
|
|
bottom_shift = 7;
|
|
}
|
|
break;
|
|
case BARCODE_RSS14_CC: bottom_shift = 4;
|
|
break;
|
|
case BARCODE_RSS_LTD_CC: bottom_shift = 9;
|
|
break;
|
|
case BARCODE_RSS_EXP_CC: k = 1;
|
|
while ((!(module_is_set(linear, 1, k - 1))) && module_is_set(linear, 1, k)) {
|
|
k++;
|
|
}
|
|
top_shift = k;
|
|
break;
|
|
case BARCODE_UPCA_CC: bottom_shift = 2;
|
|
break;
|
|
case BARCODE_UPCE_CC: bottom_shift = 2;
|
|
break;
|
|
case BARCODE_RSS14STACK_CC: top_shift = 1;
|
|
break;
|
|
case BARCODE_RSS14_OMNI_CC: top_shift = 1;
|
|
break;
|
|
case BARCODE_RSS_EXPSTACK_CC: k = 1;
|
|
while ((!(module_is_set(linear, 1, k - 1))) && module_is_set(linear, 1, k)) {
|
|
k++;
|
|
}
|
|
top_shift = k;
|
|
break;
|
|
}
|
|
|
|
if (top_shift != 0) {
|
|
/* Move the 2d component of the symbol horizontally */
|
|
for (i = 0; i <= symbol->rows; i++) {
|
|
for (j = (symbol->width + top_shift); j >= top_shift; j--) {
|
|
if (module_is_set(symbol, i, j - top_shift)) {
|
|
set_module(symbol, i, j);
|
|
} else {
|
|
unset_module(symbol, i, j);
|
|
}
|
|
}
|
|
for (j = 0; j < top_shift; j++) {
|
|
unset_module(symbol, i, j);
|
|
}
|
|
}
|
|
}
|
|
|
|
/* Merge linear and 2D components into one structure */
|
|
for (i = 0; i <= linear->rows; i++) {
|
|
symbol->row_height[symbol->rows + i] = linear->row_height[i];
|
|
for (j = 0; j <= linear->width; j++) {
|
|
if (module_is_set(linear, i, j)) {
|
|
set_module(symbol, i + symbol->rows, j + bottom_shift);
|
|
} else {
|
|
unset_module(symbol, i + symbol->rows, j + bottom_shift);
|
|
}
|
|
}
|
|
}
|
|
if ((linear->width + bottom_shift) > symbol->width) {
|
|
symbol->width = linear->width + bottom_shift;
|
|
}
|
|
if ((symbol->width + top_shift) > symbol->width) {
|
|
symbol->width += top_shift;
|
|
}
|
|
symbol->rows += linear->rows;
|
|
ustrcpy(symbol->text, (unsigned char *) linear->text);
|
|
|
|
ZBarcode_Delete(linear);
|
|
|
|
return error_number;
|
|
}
|
|
|
|
|