utf8.c

#include "utf8.h"

#include <stdint.h>

const char unicode_box_to_cp437[128] = {
/* 0x2500 ─ */ '\xC4', // 196  single horizontal
/* 0x2501 ━ */ '\xC4', // heavy → single
/* 0x2502 │ */ '\xB3', // 179  single vertical
/* 0x2503 ┃ */ '\xB3', // heavy → single
/* 0x2504 ┄ */ '\xC4', // light dashed horizontal → solid single
/* 0x2505 ┅ */ '\xC4', // heavy dashed horizontal → solid single
/* 0x2506 ┆ */ '\xB3', // light dashed vertical → solid single
/* 0x2507 ┇ */ '\xB3', // heavy dashed vertical → solid single
/* 0x2508 ┈ */ '\xC4', // longer light dashed horiz → solid
/* 0x2509 ┉ */ '\xC4',
/* 0x250A ┊ */ '\xB3', // longer light dashed vert → solid
/* 0x250B ┋ */ '\xB3',
/* 0x250C ┌ */ '\xDA', // 218  single down-right
/* 0x250D ┍ */ '\xDA',
/* 0x250E ┎ */ '\xDA',
/* 0x250F ┏ */ '\xDA',
/* 0x2510 ┐ */ '\xBF', // 191  single down-left
/* 0x2511 ┑ */ '\xBF',
/* 0x2512 ┒ */ '\xBF',
/* 0x2513 ┓ */ '\xBF',
/* 0x2514 └ */ '\xC0', // 192  single up-right
/* 0x2515 ┕ */ '\xC0',
/* 0x2516 ┖ */ '\xC0',
/* 0x2517 ┗ */ '\xC0',
/* 0x2518 ┘ */ '\xD9', // 217  single up-left
/* 0x2519 ┙ */ '\xD9',
/* 0x251A ┚ */ '\xD9',
/* 0x251B ┛ */ '\xD9',
/* 0x251C ├ */ '\xC3', // 195  single right + vertical
/* 0x251D ┝ */ '\xC3',
/* 0x251E ┞ */ '\xC3',
/* 0x251F ┟ */ '\xC3',
/* 0x2520 ┠ */ '\xC3',
/* 0x2521 ┡ */ '\xC3',
/* 0x2522 ┢ */ '\xC3',
/* 0x2523 ┣ */ '\xC3',
/* 0x2524 ┤ */ '\xB4', // 180  single left + vertical
/* 0x2525 ┥ */ '\xB4',
/* 0x2526 ┦ */ '\xB4',
/* 0x2527 ┧ */ '\xB4',
/* 0x2528 ┨ */ '\xB4',
/* 0x2529 ┩ */ '\xB4',
/* 0x252A ┪ */ '\xB4',
/* 0x252B ┫ */ '\xB4',
/* 0x252C ┬ */ '\xC2', // 194  single horizontal + down
/* 0x252D ┭ */ '\xC2',
/* 0x252E ┮ */ '\xC2',
/* 0x252F ┯ */ '\xC2',
/* 0x2530 ┰ */ '\xC2',
/* 0x2531 ┱ */ '\xC2',
/* 0x2532 ┲ */ '\xC2',
/* 0x2533 ┳ */ '\xC2',
/* 0x2534 ┴ */ '\xC1', // 193  single horizontal + up
/* 0x2535 ┵ */ '\xC1',
/* 0x2536 ┶ */ '\xC1',
/* 0x2537 ┷ */ '\xC1',
/* 0x2538 ┸ */ '\xC1',
/* 0x2539 ┹ */ '\xC1',
/* 0x253A ┺ */ '\xC1',
/* 0x253B ┻ */ '\xC1',
/* 0x253C ┼ */ '\xC5', // 197  single cross
/* 0x253D ┽ */ '\xC5',
/* 0x253E ┾ */ '\xC5',
/* 0x253F ┿ */ '\xC5',
/* 0x2540 ╀ */ '\xC5',
/* 0x2541 ╁ */ '\xC5',
/* 0x2542 ╂ */ '\xC5',
/* 0x2543 ╃ */ '\xC5',
/* 0x2544 ╄ */ '\xC5',
/* 0x2545 ╅ */ '\xC5',
/* 0x2546 ╆ */ '\xC5',
/* 0x2547 ╇ */ '\xC5',
/* 0x2548 ╈ */ '\xC5',
/* 0x2549 ╉ */ '\xC5',
/* 0x254A ╊ */ '\xC5',
/* 0x254B ╋ */ '\xC5',
/* 0x254C ╌ */ '\xC4', // dashed double horiz → solid single horiz
/* 0x254D ╍ */ '\xC4',
/* 0x254E ╎ */ '\xB3', // dashed double vert → solid single vert
/* 0x254F ╏ */ '\xB3',
/* 0x2550 ═ */ '\xCD', // 205  double horizontal
/* 0x2551 ║ */ '\xBA', // 186  double vertical
/* 0x2552 ╒ */ '\xD5', // 213
/* 0x2553 ╓ */ '\xB4', // best approximation
/* 0x2554 ╔ */ '\xC9', // 201  double down-right
/* 0x2555 ╕ */ '\xB8', // 184
/* 0x2556 ╖ */ '\xB4',
/* 0x2557 ╗ */ '\xBB', // 187  double down-left
/* 0x2558 ╘ */ '\xBE', // 190
/* 0x2559 ╙ */ '\xC8', // approx
/* 0x255A ╚ */ '\xC8', // 200  double up-right
/* 0x255B ╛ */ '\xBC', // 188
/* 0x255C ╜ */ '\xBC',
/* 0x255D ╝ */ '\xBC', // 188  double up-left
/* 0x255E ╞ */ '\xCC', // 204
/* 0x255F ╟ */ '\xCC',
/* 0x2560 ╠ */ '\xCC', // 204  double right + vertical
/* 0x2561 ╡ */ '\xB4', // approx
/* 0x2562 ╢ */ '\xB6',
/* 0x2563 ╣ */ '\xB9', // 185  double left + vertical
/* 0x2564 ╤ */ '\xD1', // 209
/* 0x2565 ╥ */ '\xD1',
/* 0x2566 ╦ */ '\xCB', // 203  double horiz + down
/* 0x2567 ╧ */ '\xCF', // 207
/* 0x2568 ╨ */ '\xCF',
/* 0x2569 ╩ */ '\xCA', // 202  double horiz + up
/* 0x256A ╪ */ '\xCE', // 206
/* 0x256B ╫ */ '\xCE',
/* 0x256C ╬ */ '\xCE', // 206  double cross
/* 0x256D ╭ */ '\xDA', // rounded → single corner
/* 0x256E ╮ */ '\xBF',
/* 0x256F ╯ */ '\xD9',
/* 0x2570 ╰ */ '\xC0',
/* 0x2571 ╱ */ '/', // diagonal
/* 0x2572 ╲ */ '\\',
/* 0x2573 ╳ */ 'X', // closest common substitute
/* 0x2574 ╴ */ '\xC4', // horizontal stub
/* 0x2575 ╵ */ '\xB3', // vertical stub
/* 0x2576 ╶ */ '\xC4',
/* 0x2577 ╷ */ '\xB3',
/* 0x2578 ╸ */ '\xC4',
/* 0x2579 ╹ */ '\xB3',
/* 0x257A ╺ */ '\xC4',
/* 0x257B ╻ */ '\xB3',
/* 0x257C ╼ */ '\xC4',
/* 0x257D ╽ */ '\xB3',
/* 0x257E ╾ */ '\xC4',
/* 0x257F ╿ */ '\xB3'
};

int utf8_to_cp437(unsigned char *data, const int size, uint32_t *accumulator, int *pending_bytes) {
	unsigned char *ascii = data;
	int length = 0;

	for (int i = 0; i < size; i++) {
		const unsigned char c = data[i];

		if (*pending_bytes == 0) {
			if (c < 0x80) {
				ascii[length++] = c;
			}
			else if ((c & 0xE0) == 0xC0) {
				// 2-byte sequence (110xxxxx)
				*accumulator = (uint32_t)(c & 0x1F);
				*pending_bytes = 1;
			}
			else if ((c & 0xF0) == 0xE0) {
				// 3-byte sequence (1110xxxx)
				*accumulator = (uint32_t)(c & 0x0F);
				*pending_bytes = 2;
			}
			else if ((c & 0xF8) == 0xF0) {
				// 4-byte sequence (11110xxx)
				*accumulator = (uint32_t)(c & 0x07);
				*pending_bytes = 3;
			}
			else {
				// invalid start byte
				ascii[length++] = '?';
			}
		}
		else {
			// continuation byte expected (10xxxxxx)
			if ((c & 0xC0) == 0x80) {
				// valid continuation
				*accumulator = (*accumulator << 6) | (c & 0x3F);

				if (--*pending_bytes == 0) {
					unsigned char ch = '?';
					if (*accumulator >= unicode_box_to_cp437_first && *accumulator <= unicode_box_to_cp437_last) {
						ch = unicode_box_to_cp437[*accumulator - unicode_box_to_cp437_first];
					}
					ascii[length++] = ch;
					*accumulator = 0;
				}
			}
			else {
				// invalid continuation → recovery
				*pending_bytes = 0;
				*accumulator = 0;
				ascii[length++] = '?';
				i--; // re-process this byte as start of new sequence
			}
		}
	}

	return length;
}