mirror of
https://github.com/n64decomp/sm64.git
synced 2024-11-14 05:15:09 -05:00
486 lines
13 KiB
C
486 lines
13 KiB
C
#include <ctype.h>
|
|
#include <errno.h>
|
|
#include <stdarg.h>
|
|
#include <stdint.h>
|
|
#include <stdlib.h>
|
|
#include <stdio.h>
|
|
#include <string.h>
|
|
|
|
#include "hashtable.h"
|
|
#include "utf8.h"
|
|
|
|
#define ARRAY_COUNT(arr) (sizeof(arr) / sizeof(arr[0]))
|
|
|
|
struct CharmapEntry
|
|
{
|
|
uint32_t unicode[3];
|
|
int length; // length of the unicode array. TODO: use dynamic memory allocation
|
|
int bytesCount;
|
|
uint8_t bytes[2]; // bytes to convert unicode array to, (e.g. 'A' = 0x0A)
|
|
};
|
|
|
|
static struct HashTable *charmap;
|
|
|
|
static void fatal_error(const char *msgfmt, ...)
|
|
{
|
|
va_list args;
|
|
|
|
fputs("error: ", stderr);
|
|
|
|
va_start(args, msgfmt);
|
|
vfprintf(stderr, msgfmt, args);
|
|
va_end(args);
|
|
|
|
fputc('\n', stderr);
|
|
|
|
exit(1);
|
|
}
|
|
|
|
static void parse_error(const char *filename, int lineNum, const char *msgfmt, ...)
|
|
{
|
|
va_list args;
|
|
|
|
fprintf(stderr, "%s: line %i: ", filename, lineNum);
|
|
|
|
va_start(args, msgfmt);
|
|
vfprintf(stderr, msgfmt, args);
|
|
va_end(args);
|
|
|
|
fputc('\n', stderr);
|
|
|
|
exit(1);
|
|
}
|
|
|
|
// Reads the whole file and returns a null-terminated buffer with its contents
|
|
void *read_text_file(const char *filename)
|
|
{
|
|
FILE *file = fopen(filename, "rb");
|
|
uint8_t *buffer;
|
|
size_t size;
|
|
|
|
if (file == NULL)
|
|
fatal_error("failed to open file '%s' for reading: %s", filename, strerror(errno));
|
|
|
|
// get size
|
|
fseek(file, 0, SEEK_END);
|
|
size = ftell(file);
|
|
|
|
// allocate buffer
|
|
buffer = malloc(size + 1);
|
|
|
|
// read file
|
|
fseek(file, 0, SEEK_SET);
|
|
if (fread(buffer, size, 1, file) != 1)
|
|
fatal_error("error reading from file '%s': %s", filename, strerror(errno));
|
|
|
|
// null-terminate the buffer
|
|
buffer[size] = 0;
|
|
|
|
fclose(file);
|
|
|
|
return buffer;
|
|
}
|
|
|
|
static char *skip_whitespace(char *str)
|
|
{
|
|
while (isspace(*str))
|
|
str++;
|
|
return str;
|
|
}
|
|
|
|
// null terminates the current line and returns a pointer to the next line
|
|
static char *line_split(char *str)
|
|
{
|
|
while (*str != '\n')
|
|
{
|
|
if (*str == 0)
|
|
return str; // end of string
|
|
str++;
|
|
}
|
|
*str = 0; // terminate line
|
|
return str + 1;
|
|
}
|
|
|
|
static char *parse_number(const char *str, unsigned int *num)
|
|
{
|
|
char *endptr;
|
|
unsigned int n = strtol(str, &endptr, 0);
|
|
|
|
*num = n;
|
|
if (endptr > str)
|
|
return endptr;
|
|
else
|
|
return NULL;
|
|
}
|
|
|
|
static int is_identifier_char(char c)
|
|
{
|
|
return isalnum(c) || c == '_';
|
|
}
|
|
|
|
static int get_escape_char(int c)
|
|
{
|
|
const uint8_t escapeTable[] =
|
|
{
|
|
['a'] = '\a',
|
|
['b'] = '\b',
|
|
['f'] = '\f',
|
|
['n'] = '\n',
|
|
['r'] = '\r',
|
|
['t'] = '\t',
|
|
['v'] = '\v',
|
|
['\\'] = '\\',
|
|
['\''] = '\'',
|
|
['"'] = '"',
|
|
};
|
|
|
|
if ((unsigned int)c < ARRAY_COUNT(escapeTable) && escapeTable[c] != 0)
|
|
return escapeTable[c];
|
|
else
|
|
return 0;
|
|
}
|
|
|
|
static void read_charmap(const char *filename)
|
|
{
|
|
char *filedata = read_text_file(filename);
|
|
char *line = filedata;
|
|
int lineNum = 1;
|
|
|
|
while (line[0] != 0)
|
|
{
|
|
char *nextLine = line_split(line);
|
|
|
|
struct CharmapEntry entry;
|
|
|
|
line = skip_whitespace(line);
|
|
if (line[0] != 0 && line[0] != '#') // ignore empty lines and comments
|
|
{
|
|
int len = 0;
|
|
/* Read Character */
|
|
|
|
// opening quote
|
|
if (*line != '\'')
|
|
parse_error(filename, lineNum, "expected '");
|
|
line++;
|
|
|
|
// perform analysis of charmap entry, we are in the quote
|
|
while(1)
|
|
{
|
|
if(*line == '\'')
|
|
{
|
|
line++;
|
|
break;
|
|
}
|
|
else if(len == ARRAY_COUNT(entry.unicode))
|
|
{
|
|
// TODO: Use dynamic memory allocation so this is unnecessary.
|
|
parse_error(filename, lineNum, "string limit exceeded");
|
|
}
|
|
else if (*line == '\\')
|
|
{
|
|
line++; // advance to get the character being escaped
|
|
if (*line == '\r')
|
|
line++;
|
|
if (*line == '\n')
|
|
{
|
|
// Backslash at end of line is ignored
|
|
continue;
|
|
}
|
|
entry.unicode[len] = get_escape_char(*line);
|
|
if (entry.unicode[len] == 0)
|
|
parse_error(filename, lineNum, "unknown escape sequence \\%c", *line);
|
|
line++; // increment again to get past the escape sequence.
|
|
}
|
|
else
|
|
{
|
|
line = utf8_decode(line, &entry.unicode[len]);
|
|
if (line == NULL)
|
|
parse_error(filename, lineNum, "invalid UTF8");
|
|
}
|
|
len++;
|
|
}
|
|
entry.length = len;
|
|
|
|
// equals sign
|
|
line = skip_whitespace(line);
|
|
if (*line != '=')
|
|
parse_error(filename, lineNum, "expected = after character \\%c", *line);
|
|
line++;
|
|
|
|
entry.bytesCount = 0;
|
|
|
|
// value
|
|
while (1)
|
|
{
|
|
uint32_t value;
|
|
|
|
if (entry.bytesCount >= 2)
|
|
parse_error(filename, lineNum, "more than 2 values specified");
|
|
|
|
line = skip_whitespace(line);
|
|
|
|
line = parse_number(line, &value);
|
|
if (line == NULL)
|
|
parse_error(filename, lineNum, "expected number after =");
|
|
if (value > 0xFF)
|
|
parse_error(filename, lineNum, "0x%X is larger than 1 byte", value);
|
|
|
|
entry.bytes[entry.bytesCount] = value;
|
|
entry.bytesCount++;
|
|
|
|
line = skip_whitespace(line);
|
|
if (*line == 0)
|
|
break;
|
|
if (*line != ',')
|
|
parse_error(filename, lineNum, "junk at end of line");
|
|
line++;
|
|
}
|
|
|
|
if (hashtable_query(charmap, &entry) != NULL)
|
|
parse_error(filename, lineNum, "entry for character already exists");
|
|
hashtable_insert(charmap, &entry);
|
|
}
|
|
|
|
line = nextLine;
|
|
lineNum++;
|
|
}
|
|
|
|
free(filedata);
|
|
}
|
|
|
|
static int count_line_num(const char *start, const char *pos)
|
|
{
|
|
const char *c;
|
|
int lineNum = 1;
|
|
|
|
for (c = start; c < pos; c++)
|
|
{
|
|
if (*c == '\n')
|
|
lineNum++;
|
|
}
|
|
return lineNum;
|
|
}
|
|
|
|
static char *convert_string(char *pos, FILE *fout, const char *inputFileName, char *start, int uncompressed)
|
|
{
|
|
int hasString = 0;
|
|
|
|
while (1)
|
|
{
|
|
pos = skip_whitespace(pos);
|
|
if (*pos == ')')
|
|
{
|
|
if (hasString)
|
|
break;
|
|
else
|
|
parse_error(inputFileName, count_line_num(start, pos), "expected quoted string after '_('");
|
|
}
|
|
else if (*pos != '"')
|
|
parse_error(inputFileName, count_line_num(start, pos), "unexpected character '%c'", *pos);
|
|
pos++;
|
|
|
|
hasString = 1;
|
|
|
|
// convert quoted string
|
|
while (*pos != '"')
|
|
{
|
|
struct CharmapEntry input;
|
|
struct CharmapEntry *last_valid_entry = NULL;
|
|
struct CharmapEntry *entry;
|
|
int i, c;
|
|
int length = 0;
|
|
char* last_valid_pos = NULL;
|
|
|
|
// safely erase the unicode area before use
|
|
memset(input.unicode, 0, sizeof (input.unicode));
|
|
input.length = 0;
|
|
|
|
// Find a charmap entry of longest length possible starting from this position
|
|
while (*pos != '"')
|
|
{
|
|
if ((uncompressed && length == 1) || length == ARRAY_COUNT(entry->unicode))
|
|
{
|
|
// Stop searching after length 3; we only support strings of lengths up
|
|
// to that right now. Unless uncompressed is set, in which we ignore multi
|
|
// texts by discarding entries longer than 1.
|
|
break;
|
|
}
|
|
|
|
if (*pos == 0)
|
|
parse_error(inputFileName, count_line_num(start, pos), "EOF in string literal");
|
|
if (*pos == '\\')
|
|
{
|
|
pos++;
|
|
c = get_escape_char(*pos);
|
|
if (c == 0)
|
|
parse_error(inputFileName, count_line_num(start, pos), "unknown escape sequence \\%c", *pos);
|
|
input.unicode[length] = c;
|
|
pos++;
|
|
}
|
|
else
|
|
{
|
|
pos = utf8_decode(pos, &input.unicode[length]);
|
|
if (pos == NULL)
|
|
parse_error(inputFileName, count_line_num(start, pos), "invalid unicode encountered in file");
|
|
}
|
|
length++;
|
|
input.length = length;
|
|
|
|
entry = hashtable_query(charmap, &input);
|
|
if (entry != NULL)
|
|
{
|
|
last_valid_entry = entry;
|
|
last_valid_pos = pos;
|
|
}
|
|
}
|
|
|
|
entry = last_valid_entry;
|
|
pos = last_valid_pos;
|
|
if (entry == NULL)
|
|
parse_error(inputFileName, count_line_num(start, pos), "no charmap entry for U+%X", input.unicode[0]);
|
|
for (i = 0; i < entry->bytesCount; i++)
|
|
fprintf(fout, "0x%02X,", entry->bytes[i]);
|
|
}
|
|
pos++; // skip over closing '"'
|
|
}
|
|
pos++; // skip over closing ')'
|
|
fputs("0xFF", fout);
|
|
return pos;
|
|
}
|
|
|
|
static void convert_file(const char *infilename, const char *outfilename)
|
|
{
|
|
char *in = read_text_file(infilename);
|
|
FILE *fout = fopen(outfilename, "wb");
|
|
|
|
if (fout == NULL)
|
|
fatal_error("failed to open file '%s' for writing: %s", strerror(errno));
|
|
|
|
char *start = in;
|
|
char *end = in;
|
|
char *pos = in;
|
|
|
|
while (1)
|
|
{
|
|
if (*pos == 0) // end of file
|
|
goto eof;
|
|
|
|
// check for comment
|
|
if (*pos == '/')
|
|
{
|
|
pos++;
|
|
// skip over // comment
|
|
if (*pos == '/')
|
|
{
|
|
pos++;
|
|
// skip over next newline
|
|
while (*pos != '\n')
|
|
{
|
|
if (*pos == 0)
|
|
goto eof;
|
|
pos++;
|
|
}
|
|
pos++;
|
|
}
|
|
// skip over /* */ comment
|
|
else if (*pos == '*')
|
|
{
|
|
pos++;
|
|
while (*pos != '*' && pos[1] != '/')
|
|
{
|
|
if (*pos == 0)
|
|
goto eof;
|
|
pos++;
|
|
}
|
|
pos += 2;
|
|
}
|
|
}
|
|
// skip over normal string literal
|
|
else if (*pos == '"')
|
|
{
|
|
pos++;
|
|
while (*pos != '"')
|
|
{
|
|
if (*pos == 0)
|
|
goto eof;
|
|
if (*pos == '\\')
|
|
pos++;
|
|
pos++;
|
|
}
|
|
pos++;
|
|
}
|
|
// check for _( sequence
|
|
else if ((*pos == '_') && (pos == in || !is_identifier_char(pos[-1])))
|
|
{
|
|
int uncompressed = 0;
|
|
end = pos;
|
|
pos++;
|
|
if (*pos == '_') // an extra _ signifies uncompressed strings. Enable uncompressed flag
|
|
{
|
|
pos++;
|
|
uncompressed = 1;
|
|
}
|
|
if (*pos == '(')
|
|
{
|
|
pos++;
|
|
fwrite(start, end - start, 1, fout);
|
|
pos = convert_string(pos, fout, infilename, in, uncompressed);
|
|
start = pos;
|
|
}
|
|
}
|
|
else
|
|
{
|
|
pos++;
|
|
}
|
|
}
|
|
|
|
eof:
|
|
fwrite(start, pos - start, 1, fout);
|
|
fclose(fout);
|
|
free(in);
|
|
}
|
|
|
|
static unsigned int charmap_hash(const void *value)
|
|
{
|
|
const struct CharmapEntry* entry = value;
|
|
unsigned int ret = 0;
|
|
for (int i = 0; i < entry->length; i++)
|
|
ret = ret * 17 + entry->unicode[i];
|
|
return ret;
|
|
}
|
|
|
|
static int charmap_cmp(const void *a, const void *b)
|
|
{
|
|
const struct CharmapEntry *ea = a;
|
|
const struct CharmapEntry *eb = b;
|
|
if (ea->length != eb->length)
|
|
return 0;
|
|
for(int i = 0; i < ea->length; i++)
|
|
if(ea->unicode[i] != eb->unicode[i])
|
|
return 0;
|
|
return 1;
|
|
}
|
|
|
|
static void usage(const char *execName)
|
|
{
|
|
fprintf(stderr, "Usage: %s CHARMAP INPUT OUTPUT\n", execName);
|
|
}
|
|
|
|
int main(int argc, char **argv)
|
|
{
|
|
if (argc != 4)
|
|
{
|
|
usage(argv[0]);
|
|
return 1;
|
|
}
|
|
|
|
charmap = hashtable_new(charmap_hash, charmap_cmp, 256, sizeof(struct CharmapEntry));
|
|
|
|
read_charmap(argv[1]);
|
|
convert_file(argv[2], argv[3]);
|
|
|
|
hashtable_free(charmap);
|
|
|
|
return 0;
|
|
}
|