Marlin/Marlin/utf_mapper.h
2017-04-21 21:42:41 -05:00

477 lines
18 KiB
C
Raw Blame History

This file contains invisible Unicode characters

This file contains invisible Unicode characters that are indistinguishable to humans but may be processed differently by a computer. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

/**
* Marlin 3D Printer Firmware
* Copyright (C) 2016 MarlinFirmware [https://github.com/MarlinFirmware/Marlin]
*
* Based on Sprinter and grbl.
* Copyright (C) 2011 Camiel Gubbels / Erik van der Zalm
*
* This program is free software: you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation, either version 3 of the License, or
* (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*
*/
#ifndef UTF_MAPPER_H
#define UTF_MAPPER_H
#include "language.h"
#if ENABLED(DOGLCD)
#define HARDWARE_CHAR_OUT u8g.print
#else
#define HARDWARE_CHAR_OUT lcd.write
#endif
#if DISABLED(SIMULATE_ROMFONT) && ENABLED(DOGLCD)
#if ENABLED(DISPLAY_CHARSET_ISO10646_1) \
|| ENABLED(DISPLAY_CHARSET_ISO10646_5) \
|| ENABLED(DISPLAY_CHARSET_ISO10646_KANA) \
|| ENABLED(DISPLAY_CHARSET_ISO10646_GREEK) \
|| ENABLED(DISPLAY_CHARSET_ISO10646_TR)
#define MAPPER_ONE_TO_ONE
#endif
#else // SIMULATE_ROMFONT
#if DISPLAY_CHARSET_HD44780 == JAPANESE
#if ENABLED(MAPPER_C2C3)
const PROGMEM uint8_t utf_recode[] =
{ // 0 1 2 3 4 5 6 7 8 9 a b c d e f This is fair for symbols
0x20,0x3F,0xEC,0xED,0x3F,0x5C,0x7C,0x3F,0x22,0x63,0x61,0x7F,0x3F,0x3F,0x52,0xB0, // c2a
//' ' ¢ £ ­ l " c a « R
0xDF,0x3F,0x32,0x33,0x27,0xE4,0xF1,0xA5,0x2C,0x31,0xDF,0x7E,0x3F,0x3F,0x3F,0x3F, // c2b but relatively bad for letters.
// ° 2 3 ` N p . , 1 ° »
0x3F,0x3F,0x3F,0x3F,0xE1,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F, // c38
// ä
0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0xEF,0x78,0x3F,0x3F,0x3F,0x3F,0xF5,0x3F,0x3F,0xE2, // c39 missing characters display as '?'
// ö x ü ß
0x3F,0x3F,0x3F,0x3F,0xE1,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F,0x3F, // c3a
// ä
0x3F,0xEE,0x3F,0x3F,0x3F,0x3F,0xEF,0xFD,0x3F,0x3F,0x3F,0x3F,0xF5,0x3F,0x3F,0x3F // c3b
// n ö ÷ ü
};
#elif ENABLED(MAPPER_E382E383)
const PROGMEM uint8_t utf_recode[] =
{ // 0 1 2 3 4 5 6 7 8 9 a b c d e f
0x3D,0xB1,0xB1,0xA8,0xB2,0xA9,0xB3,0xAA,0xB4,0xAB,0xB5,0xB6,0xB6,0xB7,0xB7,0xB8, // e382a Please test and correct
// = ア ア ィ イ ゥ ウ ェ エ ォ オ ガ ガ キ キ ク
0xB8,0xB9,0xB9,0xBA,0xBA,0xBB,0xBB,0xBC,0xBC,0xBD,0xBD,0xBE,0xBE,0xBF,0xBF,0xC0, // e382b
// ク ケ ケ コ コ サ サ シ シ ス ス セ セ ソ ソ タ
0xC0,0xC1,0xC1,0xC2,0xC2,0xC2,0xC3,0xC3,0xC4,0xC4,0xC5,0xC6,0xC7,0xC8,0xC9,0xCA, // e3838
// タ チ チ ッ ッ ッ テ テ ト ト ナ ニ ヌ ネ
0xCA,0xCA,0xCB,0xCB,0xCB,0xCC,0xCC,0xCC,0xCD,0xCD,0xCD,0xCE,0xCE,0xCE,0xCF,0xD0, // e3839
// ハ ハ ヒ ヒ ヒ フ フ フ ヘ ヘ ヘ ホ ホ ホ マ ミ
0xD1,0xD2,0xD3,0xD4,0xD4,0xD5,0xD5,0xAE,0xD6,0xD7,0xD8,0xD9,0xDA,0xDB,0xDC,0xDC, // e383a
// ム メ モ ャ ャ ユ ユ ョ ヨ ラ リ ル レ ロ ワ ワ
0xEC,0xA7,0xA6,0xDD,0xCC,0x3F,0x3F,0x3F,0x3F,0x3F,0xA6,0xA5,0xB0,0xA4,0xA4,0x3F // e383b
// ヰ ヱ ヲ ン フ ? ? ? ? ? ヲ ・ ー ヽ ヽ ?
};
#elif ENABLED(MAPPER_D0D1)
#error "Cyrillic on a JAPANESE display makes no sense. There are no matching symbols."
#endif
#elif DISPLAY_CHARSET_HD44780 == WESTERN
#if ENABLED(MAPPER_C2C3)
const PROGMEM uint8_t utf_recode[] =
{ // 0 1 2 3 4 5 6 7 8 9 a b c d e f This is relative complete.
0x20,0xA1,0xA2,0xA3,0xA4,0xA5,0xA6,0xA7,0x22,0xA9,0xAA,0xAB,0x3F,0x3F,0xAE,0x3F, // c2a ¡¢£¤¥¦§¨©ª«¬­®¯
//' ' ¡ ¢ £ ¤ ¥ ¦ § " © ª « ? ? ® ?
0xB0,0xB1,0xB2,0xB3,0x27,0xB5,0xB6,0xB7,0x2C,0xB9,0xBA,0xBB,0xBC,0xBD,0xBE,0xBF, // c2b °±²³´µ¶·¸¹º»¼½¾¿
// ° ± ³ ² ? µ ¶ · , ¹ º » ¼ ½ ¾ ¿
0xC0,0xC1,0xC2,0xC3,0xC4,0xC5,0xC6,0xC7,0xC8,0xC9,0xCA,0xCB,0xCC,0xCD,0xCE,0xCF, // c38 ÀÁÃÄÅÆÇÈÉÊËÌÍÎÏ
// À Á Â Ã Ä Å Æ Ç È É Ê Ë Ì Í Î Ï
0xD0,0xD1,0xD2,0xD3,0xD4,0xD5,0xD6,0xD7,0xD8,0xD9,0xDA,0xDB,0xDC,0xDD,0xDE,0xDF, // c39 ÐÑÓÔÕÖ×ØÙÚÛÜÝÞß
// Ð Ñ Ò Ó Ô Õ Ö × Ø Ù Ú Û Ü Ý Þ ß
0xE0,0xE1,0xE2,0xE3,0xE4,0xE5,0xE6,0xE7,0xE8,0xE9,0xEA,0xEB,0xEC,0xED,0xEE,0xEF, // c3a àáãäåæçèéêëìíîï
// à á â ã ä å æ ç è é ê ë ì í î ï
0xF0,0xF1,0xF2,0xF3,0xF4,0xF5,0xF6,0xF7,0xF8,0xF9,0xFA,0xFB,0xFC,0xFD,0xFE,0xFF // c3b ðñóôõö÷øùúûüýþÿ
// ð ñ ò ó ô õ ö ÷ ø ù ú û ü ý þ ÿ
};
#elif ENABLED(MAPPER_D0D1)
#define MAPPER_D0D1_MOD
const PROGMEM uint8_t utf_recode[] =
{//0 1 2 3 4 5 6 7 8 9 a b c d e f
0x41,0x80,0x42,0x92,0x81,0x45,0x82,0x83,0x84,0x85,0x4B,0x86,0x4D,0x48,0x4F,0x87, // d0a
// A Б B Г Д E Ж З И Й K Л M H O П
0x50,0x43,0x54,0x88,0xD8,0x58,0x89,0x8A,0x8B,0x8C,0x8D,0x8E,0x62,0x8F,0xAC,0xAD, // d0b
// P C T У Ф X Ч ч Ш Щ Ъ Ы b Э Ю Я
0x61,0x36,0x42,0x92,0x81,0x65,0x82,0xB3,0x84,0x85,0x6B,0x86,0x4D,0x48,0x6F,0x87, // d18
// a 6 B Г Д e Ж ³ И Й k Л M H o П
0x70,0x63,0x54,0x79,0xD8,0x78,0x89,0x8A,0x8B,0x8C,0x8D,0x8E,0x62,0x8F,0xAC,0xAD // d19
// p c T y Ф x Ч ч Ш Щ Ъ Ы b Э Ю Я
};
#elif ENABLED(MAPPER_E382E383)
#error "Katakana on a WESTERN display makes no sense. There are no matching symbols."
#endif
#elif DISPLAY_CHARSET_HD44780 == CYRILLIC
#if ENABLED(MAPPER_D0D1)
#define MAPPER_D0D1_MOD
// it is a Russian alphabet translation
// except 0401 --> 0xA2 = Ё, 0451 --> 0xB5 = ё
const PROGMEM uint8_t utf_recode[] =
{ 0x41,0xA0,0x42,0xA1,0xE0,0x45,0xA3,0xA4, // unicode U+0400 to U+047f
// A Б->Ё B Г Д E Ж З // 0 Ѐ Ё Ђ Ѓ Є Ѕ І Ї
0xA5,0xA6,0x4B,0xA7,0x4D,0x48,0x4F,0xA8, // Ј Љ Њ Ћ Ќ Ѝ Ў Џ
// И Й K Л M H O П // 1 А Б В Г Д Е Ж З
0x50,0x43,0x54,0xA9,0xAA,0x58,0xE1,0xAB, // И Й К Л М Н О П
// P C T У Ф X Ч ч // 2 Р С Т У Ф Х Г Ч
0xAC,0xE2,0xAD,0xAE,0x62,0xAF,0xB0,0xB1, // Ш Щ Ъ Ы Ь Э Ю Я
// Ш Щ Ъ Ы b Э Ю Я // 3 а б в г д е ж з
0x61,0xB2,0xB3,0xB4,0xE3,0x65,0xB6,0xB7, // и й к л м н о п
// a б->ё в г д e ж з // 4 р с т у ф х ц ч
0xB8,0xB9,0xBA,0xBB,0xBC,0xBD,0x6F,0xBE, // ш щ ъ ы ь э ю я
// и й к л м н o п // 5 ѐ ё ђ ѓ є ѕ і ї
0x70,0x63,0xBF,0x79,0xE4,0x78,0xE5,0xC0, // ј љ њ ћ ќ ѝ ў џ
// p c т y ф x ц ч // 6 Ѡ ѡ Ѣ ѣ Ѥ ѥ Ѧ ѧ
0xC1,0xE6,0xC2,0xC3,0xC4,0xC5,0xC6,0xC7 // Ѫ ѩ Ѫ ѫ Ѭ ѭ Ѯ ѯ
// ш щ ъ ы ь э ю я // 7 Ѱ ѱ Ѳ ѳ Ѵ ѵ Ѷ ѷ
}; // ѻ ѹ Ѻ ѻ Ѽ ѽ Ѿ ѿ
#elif ENABLED(MAPPER_C2C3)
#error "Western languages on a CYRILLIC display makes no sense. There are no matching symbols."
#elif ENABLED(MAPPER_E382E383)
#error "Katakana on a CYRILLIC display makes no sense. There are no matching symbols."
#endif
#else
#error "Something went wrong in the setting of DISPLAY_CHARSET_HD44780"
#endif // DISPLAY_CHARSET_HD44780
#endif // SIMULATE_ROMFONT
#if ENABLED(MAPPER_C2C3)
char charset_mapper(const char c) {
static uint8_t utf_hi_char; // UTF-8 high part
static bool seen_c2 = false;
uint8_t d = c;
if ( d >= 0x80u ) { // UTF-8 handling
if ( (d >= 0xC0u) && (!seen_c2) ) {
utf_hi_char = d - 0xC2u;
seen_c2 = true;
return 0;
}
else if (seen_c2) {
d &= 0x3Fu;
#ifndef MAPPER_ONE_TO_ONE
HARDWARE_CHAR_OUT((char)pgm_read_byte_near(utf_recode + d + (utf_hi_char << 6) - 0x20));
#else
HARDWARE_CHAR_OUT((char)(0x80u + (utf_hi_char << 6) + d)) ;
#endif
}
else {
HARDWARE_CHAR_OUT('?');
}
}
else {
HARDWARE_CHAR_OUT((char) c );
}
seen_c2 = false;
return 1;
}
#elif ENABLED(MAPPER_C2C3_TR)
// the C2C3-mapper extended for the 6 altered symbols from C4 and C5 range.
char charset_mapper(const char c) {
static uint8_t utf_hi_char; // UTF-8 high part
static bool seen_c2 = false,
seen_c4 = false,
seen_c5 = false;
uint8_t d = c;
if ( d >= 0x80u ) { // UTF-8 handling
if ( d == 0xC4u ) {seen_c4 = true; return 0;}
else if ( d == 0xC5u ) {seen_c5 = true; return 0;}
else if ( (d >= 0xC0u) && (!seen_c2) ) {
utf_hi_char = d - 0xC2u;
seen_c2 = true;
return 0;
}
else if (seen_c4) {
switch(d) {
case 0x9Eu: d = 0xD0u; break;
case 0x9Fu: d = 0xF0u; break;
case 0xB0u: d = 0xDDu; break;
case 0xB1u: d = 0xFDu; break;
default: d = '?';
}
HARDWARE_CHAR_OUT((char)d) ;
}
else if (seen_c5) {
switch(d) {
case 0x9Eu: d = 0xDEu; break;
case 0x9Fu: d = 0xFEu; break;
default: d = '?';
}
HARDWARE_CHAR_OUT((char)d) ;
}
else if (seen_c2) {
d &= 0x3Fu;
#ifndef MAPPER_ONE_TO_ONE
HARDWARE_CHAR_OUT((char)pgm_read_byte_near(utf_recode + d + (utf_hi_char << 6) - 0x20));
#else
HARDWARE_CHAR_OUT((char)(0x80u + (utf_hi_char << 6) + d)) ;
#endif
}
else {
HARDWARE_CHAR_OUT('?');
}
}
else {
HARDWARE_CHAR_OUT((char) c );
}
seen_c2 = seen_c4 = seen_c5 = false;
return 1;
}
#elif ENABLED(MAPPER_CECF)
char charset_mapper(const char c) {
static uint8_t utf_hi_char; // UTF-8 high part
static bool seen_ce = false;
uint8_t d = c;
if ( d >= 0x80 ) { // UTF-8 handling
if ( (d >= 0xC0) && (!seen_ce) ) {
utf_hi_char = d - 0xCE;
seen_ce = true;
return 0;
}
else if (seen_ce) {
d &= 0x3F;
#ifndef MAPPER_ONE_TO_ONE
HARDWARE_CHAR_OUT((char)pgm_read_byte_near(utf_recode + d + (utf_hi_char << 6) - 0x20));
#else
HARDWARE_CHAR_OUT((char)(0x80 + (utf_hi_char << 6) + d)) ;
#endif
}
else {
HARDWARE_CHAR_OUT('?');
}
}
else {
HARDWARE_CHAR_OUT((char)c);
}
seen_ce = false;
return 1;
}
#elif ENABLED(MAPPER_CECF)
char charset_mapper(const char c) {
static uint8_t utf_hi_char; // UTF-8 high part
static bool seen_ce = false;
uint8_t d = c;
if (d >= 0x80) { // UTF-8 handling
if (d >= 0xC0 && !seen_ce) {
utf_hi_char = d - 0xCE;
seen_ce = true;
return 0;
}
else if (seen_ce) {
d &= 0x3F;
#ifndef MAPPER_ONE_TO_ONE
HARDWARE_CHAR_OUT((char)pgm_read_byte_near(utf_recode + d + (utf_hi_char << 6) - 0x20));
#else
HARDWARE_CHAR_OUT((char)(0x80 + (utf_hi_char << 6) + d)) ;
#endif
}
else {
HARDWARE_CHAR_OUT('?');
}
}
else {
HARDWARE_CHAR_OUT((char) c );
}
seen_ce = false;
return 1;
}
#elif ENABLED(MAPPER_D0D1_MOD)
char charset_mapper(const char c) {
// it is a Russian alphabet translation
// except 0401 --> 0xA2 = Ё, 0451 --> 0xB5 = ё
static uint8_t utf_hi_char; // UTF-8 high part
static bool seen_d5 = false;
uint8_t d = c;
if (d >= 0x80) { // UTF-8 handling
if (d >= 0xD0 && !seen_d5) {
utf_hi_char = d - 0xD0;
seen_d5 = true;
return 0;
}
else if (seen_d5) {
d &= 0x3F;
if (!utf_hi_char && d == 1) {
HARDWARE_CHAR_OUT((char) 0xA2); // Ё
}
else if (utf_hi_char == 1 && d == 0x11) {
HARDWARE_CHAR_OUT((char)0xB5); // ё
}
else {
HARDWARE_CHAR_OUT((char)pgm_read_byte_near(utf_recode + d + (utf_hi_char << 6) - 0x10));
}
}
else {
HARDWARE_CHAR_OUT('?');
}
}
else {
HARDWARE_CHAR_OUT((char) c );
}
seen_d5 = false;
return 1;
}
#elif ENABLED(MAPPER_D0D1)
char charset_mapper(const char c) {
static uint8_t utf_hi_char; // UTF-8 high part
static bool seen_d5 = false;
uint8_t d = c;
if (d >= 0x80u) { // UTF-8 handling
if (d >= 0xD0u && !seen_d5) {
utf_hi_char = d - 0xD0u;
seen_d5 = true;
return 0;
}
else if (seen_d5) {
d &= 0x3Fu;
#ifndef MAPPER_ONE_TO_ONE
HARDWARE_CHAR_OUT((char)pgm_read_byte_near(utf_recode + d + (utf_hi_char << 6) - 0x20));
#else
HARDWARE_CHAR_OUT((char)(0xA0u + (utf_hi_char << 6) + d)) ;
#endif
}
else {
HARDWARE_CHAR_OUT('?');
}
}
else {
HARDWARE_CHAR_OUT((char) c );
}
seen_d5 = false;
return 1;
}
#elif ENABLED(MAPPER_E382E383)
char charset_mapper(const char c) {
static uint8_t utf_hi_char; // UTF-8 high part
static bool seen_e3 = false,
seen_82_83 = false;
uint8_t d = c;
if (d >= 0x80) { // UTF-8 handling
if (d == 0xE3 && !seen_e3) {
seen_e3 = true;
return 0; // eat 0xE3
}
else if (d >= 0x82 && seen_e3 && !seen_82_83) {
utf_hi_char = d - 0x82;
seen_82_83 = true;
return 0;
}
else if (seen_e3 && seen_82_83) {
d &= 0x3F;
#ifndef MAPPER_ONE_TO_ONE
HARDWARE_CHAR_OUT((char)pgm_read_byte_near(utf_recode + d + (utf_hi_char << 6) - 0x20));
#else
HARDWARE_CHAR_OUT((char)(0x80 + (utf_hi_char << 6) + d)) ;
#endif
}
else {
HARDWARE_CHAR_OUT((char) '?' );
}
}
else {
HARDWARE_CHAR_OUT((char) c );
}
seen_e3 = false;
seen_82_83 = false;
return 1;
}
#elif ENABLED(MAPPER_C3C4C5_PL)
//Ą c4 84 = 80
//ą c4 85 = 81
//Ć c4 86 = 82
//ć c4 87 = 83
//Ę c4 98 = 84
//ę c4 99 = 85
//Ł c5 81 = 86
//ł c5 82 = 87
//Ń c5 83 = 88
//ń c5 84 = 89
//Ó c3 93 = 8a
//ó c3 b3 = 8b
//Ś c5 9a = 8c
//ś c5 9b = 8d
//Ź c5 b9 = 8e
//ź c5 ba = 8f
//Ż c5 bb = 90
//ż c5 bc = 91
char charset_mapper(const char c) {
static uint8_t utf_hi_char; // UTF-8 high part
static bool seen_c3 = false,
seen_c4 = false,
seen_c5 = false;
uint8_t d = c;
if ( d >= 0x80u ) { // UTF-8 handling
if ( d == 0xC4u ) {seen_c4 = true; return 0;}
else if ( d == 0xC5u ) {seen_c5 = true; return 0;}
else if ( d == 0xC3u ) {seen_c3 = true; return 0;}
else if (seen_c4) {
switch(d) {
case 0x84u ... 0x87u: d -= 4; break; //Ą - ć
case 0x98u ... 0x99u: d -= 20; break; //Ę i ę
default: d = '?';
}
HARDWARE_CHAR_OUT((char)d) ;
}
else if (seen_c5) {
switch(d) {
case 0x81u ... 0x84u: d += 5; break; //Ł - ń
case 0x9Au ... 0x9Bu: d -= 0x0Eu; break; //Ś i ś
case 0xB9u ... 0xBCu: d -= 0x2Bu; break; //Ź - ż
default: d = '?';
}
HARDWARE_CHAR_OUT((char)d) ;
}
else if (seen_c3) {
switch(d) {
case 0x93u: d = 0x8Au; break; //Ó
case 0xB3u: d = 0x8Bu; break; //ó
d = '?';
}
HARDWARE_CHAR_OUT((char)d) ;
}
}
else {
HARDWARE_CHAR_OUT((char) c );
}
seen_c3 = seen_c4 = seen_c5 = false;
return 1;
}
#else
#define MAPPER_NON
char charset_mapper(const char c) {
HARDWARE_CHAR_OUT( c );
return 1;
}
#endif // code mappers
#endif // UTF_MAPPER_H