michael@0: /*
michael@0:  * Copyright 1996, 1997, 1998 Computing Research Labs,
michael@0:  * New Mexico State University
michael@0:  *
michael@0:  * Permission is hereby granted, free of charge, to any person obtaining a
michael@0:  * copy of this software and associated documentation files (the "Software"),
michael@0:  * to deal in the Software without restriction, including without limitation
michael@0:  * the rights to use, copy, modify, merge, publish, distribute, sublicense,
michael@0:  * and/or sell copies of the Software, and to permit persons to whom the
michael@0:  * Software is furnished to do so, subject to the following conditions:
michael@0:  *
michael@0:  * The above copyright notice and this permission notice shall be included in
michael@0:  * all copies or substantial portions of the Software.
michael@0:  *
michael@0:  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
michael@0:  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
michael@0:  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
michael@0:  * THE COMPUTING RESEARCH LAB OR NEW MEXICO STATE UNIVERSITY BE LIABLE FOR ANY
michael@0:  * CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT
michael@0:  * OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR
michael@0:  * THE USE OR OTHER DEALINGS IN THE SOFTWARE.
michael@0:  */
michael@0: #ifndef lint
michael@0: #ifdef __GNUC__
michael@0: static char rcsid[] __attribute__ ((unused)) = "$Id: ucgendat.c,v 1.1 1999/01/08 00:19:21 ftang%netscape.com Exp $";
michael@0: #else
michael@0: static char rcsid[] = "$Id: ucgendat.c,v 1.1 1999/01/08 00:19:21 ftang%netscape.com Exp $";
michael@0: #endif
michael@0: #endif
michael@0: 
michael@0: #include <stdio.h>
michael@0: #include <stdlib.h>
michael@0: #include <string.h>
michael@0: #ifndef WIN32
michael@0: #include <unistd.h>
michael@0: #endif
michael@0: 
michael@0: #define ishdigit(cc) (((cc) >= '0' && (cc) <= '9') ||\
michael@0:                       ((cc) >= 'A' && (cc) <= 'F') ||\
michael@0:                       ((cc) >= 'a' && (cc) <= 'f'))
michael@0: 
michael@0: /*
michael@0:  * A header written to the output file with the byte-order-mark and the number
michael@0:  * of property nodes.
michael@0:  */
michael@0: static unsigned short hdr[2] = {0xfeff, 0};
michael@0: 
michael@0: #define NUMPROPS 49
michael@0: #define NEEDPROPS (NUMPROPS + (4 - (NUMPROPS & 3)))
michael@0: 
michael@0: typedef struct {
michael@0:     char *name;
michael@0:     int len;
michael@0: } _prop_t;
michael@0: 
michael@0: /*
michael@0:  * List of properties expected to be found in the Unicode Character Database
michael@0:  * including some implementation specific properties.
michael@0:  *
michael@0:  * The implementation specific properties are:
michael@0:  * Cm = Composed (can be decomposed)
michael@0:  * Nb = Non-breaking
michael@0:  * Sy = Symmetric (has left and right forms)
michael@0:  * Hd = Hex digit
michael@0:  * Qm = Quote marks
michael@0:  * Mr = Mirroring
michael@0:  * Ss = Space, other
michael@0:  * Cp = Defined character
michael@0:  */
michael@0: static _prop_t props[NUMPROPS] = {
michael@0:     {"Mn", 2}, {"Mc", 2}, {"Me", 2}, {"Nd", 2}, {"Nl", 2}, {"No", 2},
michael@0:     {"Zs", 2}, {"Zl", 2}, {"Zp", 2}, {"Cc", 2}, {"Cf", 2}, {"Cs", 2},
michael@0:     {"Co", 2}, {"Cn", 2}, {"Lu", 2}, {"Ll", 2}, {"Lt", 2}, {"Lm", 2},
michael@0:     {"Lo", 2}, {"Pc", 2}, {"Pd", 2}, {"Ps", 2}, {"Pe", 2}, {"Po", 2},
michael@0:     {"Sm", 2}, {"Sc", 2}, {"Sk", 2}, {"So", 2}, {"L",  1}, {"R",  1},
michael@0:     {"EN", 2}, {"ES", 2}, {"ET", 2}, {"AN", 2}, {"CS", 2}, {"B",  1},
michael@0:     {"S",  1}, {"WS", 2}, {"ON", 2},
michael@0:     {"Cm", 2}, {"Nb", 2}, {"Sy", 2}, {"Hd", 2}, {"Qm", 2}, {"Mr", 2},
michael@0:     {"Ss", 2}, {"Cp", 2}, {"Pi", 2}, {"Pf", 2}
michael@0: };
michael@0: 
michael@0: typedef struct {
michael@0:     unsigned long *ranges;
michael@0:     unsigned short used;
michael@0:     unsigned short size;
michael@0: } _ranges_t;
michael@0: 
michael@0: static _ranges_t proptbl[NUMPROPS];
michael@0: 
michael@0: /*
michael@0:  * Make sure this array is sized to be on a 4-byte boundary at compile time.
michael@0:  */
michael@0: static unsigned short propcnt[NEEDPROPS];
michael@0: 
michael@0: /*
michael@0:  * Array used to collect a decomposition before adding it to the decomposition
michael@0:  * table.
michael@0:  */
michael@0: static unsigned long dectmp[64];
michael@0: static unsigned long dectmp_size;
michael@0: 
michael@0: typedef struct {
michael@0:     unsigned long code;
michael@0:     unsigned short size;
michael@0:     unsigned short used;
michael@0:     unsigned long *decomp;
michael@0: } _decomp_t;
michael@0: 
michael@0: /*
michael@0:  * List of decomposition.  Created and expanded in order as the characters are
michael@0:  * encountered.
michael@0:  */
michael@0: static _decomp_t *decomps;
michael@0: static unsigned long decomps_used;
michael@0: static unsigned long decomps_size;
michael@0: 
michael@0: /*
michael@0:  * Types and lists for handling lists of case mappings.
michael@0:  */
michael@0: typedef struct {
michael@0:     unsigned long key;
michael@0:     unsigned long other1;
michael@0:     unsigned long other2;
michael@0: } _case_t;
michael@0: 
michael@0: static _case_t *upper;
michael@0: static _case_t *lower;
michael@0: static _case_t *title;
michael@0: static unsigned long upper_used;
michael@0: static unsigned long upper_size;
michael@0: static unsigned long lower_used;
michael@0: static unsigned long lower_size;
michael@0: static unsigned long title_used;
michael@0: static unsigned long title_size;
michael@0: 
michael@0: /*
michael@0:  * Array used to collect case mappings before adding them to a list.
michael@0:  */
michael@0: static unsigned long cases[3];
michael@0: 
michael@0: /*
michael@0:  * An array to hold ranges for combining classes.
michael@0:  */
michael@0: static unsigned long *ccl;
michael@0: static unsigned long ccl_used;
michael@0: static unsigned long ccl_size;
michael@0: 
michael@0: /*
michael@0:  * Structures for handling numbers.
michael@0:  */
michael@0: typedef struct {
michael@0:     unsigned long code;
michael@0:     unsigned long idx;
michael@0: } _codeidx_t;
michael@0: 
michael@0: typedef struct {
michael@0:     short numerator;
michael@0:     short denominator;
michael@0: } _num_t;
michael@0: 
michael@0: /*
michael@0:  * Arrays to hold the mapping of codes to numbers.
michael@0:  */
michael@0: static _codeidx_t *ncodes;
michael@0: static unsigned long ncodes_used;
michael@0: static unsigned long ncodes_size;
michael@0: 
michael@0: static _num_t *nums;
michael@0: static unsigned long nums_used;
michael@0: static unsigned long nums_size;
michael@0: 
michael@0: /*
michael@0:  * Array for holding numbers.
michael@0:  */
michael@0: static _num_t *nums;
michael@0: static unsigned long nums_used;
michael@0: static unsigned long nums_size;
michael@0: 
michael@0: static void
michael@0: #ifdef __STDC__
michael@0: add_range(unsigned long start, unsigned long end, char *p1, char *p2)
michael@0: #else
michael@0: add_range(start, end, p1, p2)
michael@0: unsigned long start, end;
michael@0: char *p1, *p2;
michael@0: #endif
michael@0: {
michael@0:     int i, j, k, len;
michael@0:     _ranges_t *rlp;
michael@0:     char *name;
michael@0: 
michael@0:     for (k = 0; k < 2; k++) {
michael@0:         if (k == 0) {
michael@0:             name = p1;
michael@0:             len = 2;
michael@0:         } else {
michael@0:             if (p2 == 0)
michael@0:               break;
michael@0: 
michael@0:             name = p2;
michael@0:             len = 1;
michael@0:         }
michael@0: 
michael@0:         for (i = 0; i < NUMPROPS; i++) {
michael@0:             if (props[i].len == len && memcmp(props[i].name, name, len) == 0)
michael@0:               break;
michael@0:         }
michael@0: 
michael@0:         if (i == NUMPROPS)
michael@0:           continue;
michael@0: 
michael@0:         rlp = &proptbl[i];
michael@0: 
michael@0:         /*
michael@0:          * Resize the range list if necessary.
michael@0:          */
michael@0:         if (rlp->used == rlp->size) {
michael@0:             if (rlp->size == 0)
michael@0:               rlp->ranges = (unsigned long *)
michael@0:                   malloc(sizeof(unsigned long) << 3);
michael@0:             else
michael@0:               rlp->ranges = (unsigned long *)
michael@0:                   realloc((char *) rlp->ranges,
michael@0:                           sizeof(unsigned long) * (rlp->size + 8));
michael@0:             rlp->size += 8;
michael@0:         }
michael@0: 
michael@0:         /*
michael@0:          * If this is the first code for this property list, just add it
michael@0:          * and return.
michael@0:          */
michael@0:         if (rlp->used == 0) {
michael@0:             rlp->ranges[0] = start;
michael@0:             rlp->ranges[1] = end;
michael@0:             rlp->used += 2;
michael@0:             continue;
michael@0:         }
michael@0: 
michael@0:         /*
michael@0:          * Optimize the case of adding the range to the end.
michael@0:          */
michael@0:         j = rlp->used - 1;
michael@0:         if (start > rlp->ranges[j]) {
michael@0:             j = rlp->used;
michael@0:             rlp->ranges[j++] = start;
michael@0:             rlp->ranges[j++] = end;
michael@0:             rlp->used = j;
michael@0:             continue;
michael@0:         }
michael@0: 
michael@0:         /*
michael@0:          * Need to locate the insertion point.
michael@0:          */
michael@0:         for (i = 0;
michael@0:              i < rlp->used && start > rlp->ranges[i + 1] + 1; i += 2) ;
michael@0: 
michael@0:         /*
michael@0:          * If the start value lies in the current range, then simply set the
michael@0:          * new end point of the range to the end value passed as a parameter.
michael@0:          */
michael@0:         if (rlp->ranges[i] <= start && start <= rlp->ranges[i + 1] + 1) {
michael@0:             rlp->ranges[i + 1] = end;
michael@0:             return;
michael@0:         }
michael@0: 
michael@0:         /*
michael@0:          * Shift following values up by two.
michael@0:          */
michael@0:         for (j = rlp->used; j > i; j -= 2) {
michael@0:             rlp->ranges[j] = rlp->ranges[j - 2];
michael@0:             rlp->ranges[j + 1] = rlp->ranges[j - 1];
michael@0:         }
michael@0: 
michael@0:         /*
michael@0:          * Add the new range at the insertion point.
michael@0:          */
michael@0:         rlp->ranges[i] = start;
michael@0:         rlp->ranges[i + 1] = end;
michael@0:         rlp->used += 2;
michael@0:     }
michael@0: }
michael@0: 
michael@0: static void
michael@0: #ifdef __STDC__
michael@0: ordered_range_insert(unsigned long c, char *name, int len)
michael@0: #else
michael@0: ordered_range_insert(c, name, len)
michael@0: unsigned long c;
michael@0: char *name;
michael@0: int len;
michael@0: #endif
michael@0: {
michael@0:     int i, j;
michael@0:     unsigned long s, e;
michael@0:     _ranges_t *rlp;
michael@0: 
michael@0:     if (len == 0)
michael@0:       return;
michael@0: 
michael@0:     for (i = 0; i < NUMPROPS; i++) {
michael@0:         if (props[i].len == len && memcmp(props[i].name, name, len) == 0)
michael@0:           break;
michael@0:     }
michael@0: 
michael@0:     if (i == NUMPROPS)
michael@0:       return;
michael@0: 
michael@0:     /*
michael@0:      * Have a match, so insert the code in order.
michael@0:      */
michael@0:     rlp = &proptbl[i];
michael@0: 
michael@0:     /*
michael@0:      * Resize the range list if necessary.
michael@0:      */
michael@0:     if (rlp->used == rlp->size) {
michael@0:         if (rlp->size == 0)
michael@0:           rlp->ranges = (unsigned long *)
michael@0:               malloc(sizeof(unsigned long) << 3);
michael@0:         else
michael@0:           rlp->ranges = (unsigned long *)
michael@0:               realloc((char *) rlp->ranges,
michael@0:                       sizeof(unsigned long) * (rlp->size + 8));
michael@0:         rlp->size += 8;
michael@0:     }
michael@0: 
michael@0:     /*
michael@0:      * If this is the first code for this property list, just add it
michael@0:      * and return.
michael@0:      */
michael@0:     if (rlp->used == 0) {
michael@0:         rlp->ranges[0] = rlp->ranges[1] = c;
michael@0:         rlp->used += 2;
michael@0:         return;
michael@0:     }
michael@0: 
michael@0:     /*
michael@0:      * Optimize the cases of extending the last range and adding new ranges to
michael@0:      * the end.
michael@0:      */
michael@0:     j = rlp->used - 1;
michael@0:     e = rlp->ranges[j];
michael@0:     s = rlp->ranges[j - 1];
michael@0: 
michael@0:     if (c == e + 1) {
michael@0:         /*
michael@0:          * Extend the last range.
michael@0:          */
michael@0:         rlp->ranges[j] = c;
michael@0:         return;
michael@0:     }
michael@0: 
michael@0:     if (c > e + 1) {
michael@0:         /*
michael@0:          * Start another range on the end.
michael@0:          */
michael@0:         j = rlp->used;
michael@0:         rlp->ranges[j] = rlp->ranges[j + 1] = c;
michael@0:         rlp->used += 2;
michael@0:         return;
michael@0:     }
michael@0: 
michael@0:     if (c >= s)
michael@0:       /*
michael@0:        * The code is a duplicate of a code in the last range, so just return.
michael@0:        */
michael@0:       return;
michael@0: 
michael@0:     /*
michael@0:      * The code should be inserted somewhere before the last range in the
michael@0:      * list.  Locate the insertion point.
michael@0:      */
michael@0:     for (i = 0;
michael@0:          i < rlp->used && c > rlp->ranges[i + 1] + 1; i += 2) ;
michael@0: 
michael@0:     s = rlp->ranges[i];
michael@0:     e = rlp->ranges[i + 1];
michael@0: 
michael@0:     if (c == e + 1)
michael@0:       /*
michael@0:        * Simply extend the current range.
michael@0:        */
michael@0:       rlp->ranges[i + 1] = c;
michael@0:     else if (c < s) {
michael@0:         /*
michael@0:          * Add a new entry before the current location.  Shift all entries
michael@0:          * before the current one up by one to make room.
michael@0:          */
michael@0:         for (j = rlp->used; j > i; j -= 2) {
michael@0:             rlp->ranges[j] = rlp->ranges[j - 2];
michael@0:             rlp->ranges[j + 1] = rlp->ranges[j - 1];
michael@0:         }
michael@0:         rlp->ranges[i] = rlp->ranges[i + 1] = c;
michael@0: 
michael@0:         rlp->used += 2;
michael@0:     }
michael@0: }
michael@0: 
michael@0: static void
michael@0: #ifdef __STDC__
michael@0: add_decomp(unsigned long code)
michael@0: #else
michael@0: add_decomp(code)
michael@0: unsigned long code;
michael@0: #endif
michael@0: {
michael@0:     unsigned long i, j, size;
michael@0: 
michael@0:     /*
michael@0:      * Add the code to the composite property.
michael@0:      */
michael@0:     ordered_range_insert(code, "Cm", 2);
michael@0: 
michael@0:     /*
michael@0:      * Locate the insertion point for the code.
michael@0:      */
michael@0:     for (i = 0; i < decomps_used && code > decomps[i].code; i++) ;
michael@0: 
michael@0:     /*
michael@0:      * Allocate space for a new decomposition.
michael@0:      */
michael@0:     if (decomps_used == decomps_size) {
michael@0:         if (decomps_size == 0)
michael@0:           decomps = (_decomp_t *) malloc(sizeof(_decomp_t) << 3);
michael@0:         else
michael@0:           decomps = (_decomp_t *)
michael@0:               realloc((char *) decomps,
michael@0:                       sizeof(_decomp_t) * (decomps_size + 8));
michael@0:         (void) memset((char *) (decomps + decomps_size), 0,
michael@0:                       sizeof(_decomp_t) << 3);
michael@0:         decomps_size += 8;
michael@0:     }
michael@0: 
michael@0:     if (i < decomps_used && code != decomps[i].code) {
michael@0:         /*
michael@0:          * Shift the decomps up by one if the codes don't match.
michael@0:          */
michael@0:         for (j = decomps_used; j > i; j--)
michael@0:           (void) memcpy((char *) &decomps[j], (char *) &decomps[j - 1],
michael@0:                         sizeof(_decomp_t));
michael@0:     }
michael@0: 
michael@0:     /*
michael@0:      * Insert or replace a decomposition.
michael@0:      */
michael@0:     size = dectmp_size + (4 - (dectmp_size & 3));
michael@0:     if (decomps[i].size < size) {
michael@0:         if (decomps[i].size == 0)
michael@0:           decomps[i].decomp = (unsigned long *)
michael@0:               malloc(sizeof(unsigned long) * size);
michael@0:         else
michael@0:           decomps[i].decomp = (unsigned long *)
michael@0:               realloc((char *) decomps[i].decomp,
michael@0:                       sizeof(unsigned long) * size);
michael@0:         decomps[i].size = size;
michael@0:     }
michael@0: 
michael@0:     if (decomps[i].code != code)
michael@0:       decomps_used++;
michael@0: 
michael@0:     decomps[i].code = code;
michael@0:     decomps[i].used = dectmp_size;
michael@0:     (void) memcpy((char *) decomps[i].decomp, (char *) dectmp,
michael@0:                   sizeof(unsigned long) * dectmp_size);
michael@0: 
michael@0: }
michael@0: 
michael@0: static void
michael@0: #ifdef __STDC__
michael@0: add_title(unsigned long code)
michael@0: #else
michael@0: add_title(code)
michael@0: unsigned long code;
michael@0: #endif
michael@0: {
michael@0:     unsigned long i, j;
michael@0: 
michael@0:     /*
michael@0:      * Always map the code to itself.
michael@0:      */
michael@0:     cases[2] = code;
michael@0: 
michael@0:     if (title_used == title_size) {
michael@0:         if (title_size == 0)
michael@0:           title = (_case_t *) malloc(sizeof(_case_t) << 3);
michael@0:         else
michael@0:           title = (_case_t *) realloc((char *) title,
michael@0:                                       sizeof(_case_t) * (title_size + 8));
michael@0:         title_size += 8;
michael@0:     }
michael@0: 
michael@0:     /*
michael@0:      * Locate the insertion point.
michael@0:      */
michael@0:     for (i = 0; i < title_used && code > title[i].key; i++) ;
michael@0: 
michael@0:     if (i < title_used) {
michael@0:         /*
michael@0:          * Shift the array up by one.
michael@0:          */
michael@0:         for (j = title_used; j > i; j--)
michael@0:           (void) memcpy((char *) &title[j], (char *) &title[j - 1],
michael@0:                         sizeof(_case_t));
michael@0:     }
michael@0: 
michael@0:     title[i].key = cases[2];    /* Title */
michael@0:     title[i].other1 = cases[0]; /* Upper */
michael@0:     title[i].other2 = cases[1]; /* Lower */
michael@0: 
michael@0:     title_used++;
michael@0: }
michael@0: 
michael@0: static void
michael@0: #ifdef __STDC__
michael@0: add_upper(unsigned long code)
michael@0: #else
michael@0: add_upper(code)
michael@0: unsigned long code;
michael@0: #endif
michael@0: {
michael@0:     unsigned long i, j;
michael@0: 
michael@0:     /*
michael@0:      * Always map the code to itself.
michael@0:      */
michael@0:     cases[0] = code;
michael@0: 
michael@0:     /*
michael@0:      * If the title case character is not present, then make it the same as
michael@0:      * the upper case.
michael@0:      */
michael@0:     if (cases[2] == 0)
michael@0:       cases[2] = code;
michael@0: 
michael@0:     if (upper_used == upper_size) {
michael@0:         if (upper_size == 0)
michael@0:           upper = (_case_t *) malloc(sizeof(_case_t) << 3);
michael@0:         else
michael@0:           upper = (_case_t *) realloc((char *) upper,
michael@0:                                       sizeof(_case_t) * (upper_size + 8));
michael@0:         upper_size += 8;
michael@0:     }
michael@0: 
michael@0:     /*
michael@0:      * Locate the insertion point.
michael@0:      */
michael@0:     for (i = 0; i < upper_used && code > upper[i].key; i++) ;
michael@0: 
michael@0:     if (i < upper_used) {
michael@0:         /*
michael@0:          * Shift the array up by one.
michael@0:          */
michael@0:         for (j = upper_used; j > i; j--)
michael@0:           (void) memcpy((char *) &upper[j], (char *) &upper[j - 1],
michael@0:                         sizeof(_case_t));
michael@0:     }
michael@0: 
michael@0:     upper[i].key = cases[0];    /* Upper */
michael@0:     upper[i].other1 = cases[1]; /* Lower */
michael@0:     upper[i].other2 = cases[2]; /* Title */
michael@0: 
michael@0:     upper_used++;
michael@0: }
michael@0: 
michael@0: static void
michael@0: #ifdef __STDC__
michael@0: add_lower(unsigned long code)
michael@0: #else
michael@0: add_lower(code)
michael@0: unsigned long code;
michael@0: #endif
michael@0: {
michael@0:     unsigned long i, j;
michael@0: 
michael@0:     /*
michael@0:      * Always map the code to itself.
michael@0:      */
michael@0:     cases[1] = code;
michael@0: 
michael@0:     /*
michael@0:      * If the title case character is empty, then make it the same as the
michael@0:      * upper case.
michael@0:      */
michael@0:     if (cases[2] == 0)
michael@0:       cases[2] = cases[0];
michael@0: 
michael@0:     if (lower_used == lower_size) {
michael@0:         if (lower_size == 0)
michael@0:           lower = (_case_t *) malloc(sizeof(_case_t) << 3);
michael@0:         else
michael@0:           lower = (_case_t *) realloc((char *) lower,
michael@0:                                       sizeof(_case_t) * (lower_size + 8));
michael@0:         lower_size += 8;
michael@0:     }
michael@0: 
michael@0:     /*
michael@0:      * Locate the insertion point.
michael@0:      */
michael@0:     for (i = 0; i < lower_used && code > lower[i].key; i++) ;
michael@0: 
michael@0:     if (i < lower_used) {
michael@0:         /*
michael@0:          * Shift the array up by one.
michael@0:          */
michael@0:         for (j = lower_used; j > i; j--)
michael@0:           (void) memcpy((char *) &lower[j], (char *) &lower[j - 1],
michael@0:                         sizeof(_case_t));
michael@0:     }
michael@0: 
michael@0:     lower[i].key = cases[1];    /* Lower */
michael@0:     lower[i].other1 = cases[0]; /* Upper */
michael@0:     lower[i].other2 = cases[2]; /* Title */
michael@0: 
michael@0:     lower_used++;
michael@0: }
michael@0: 
michael@0: static void
michael@0: #ifdef __STDC__
michael@0: ordered_ccl_insert(unsigned long c, unsigned long ccl_code)
michael@0: #else
michael@0: ordered_ccl_insert(c, ccl_code)
michael@0: unsigned long c, ccl_code;
michael@0: #endif
michael@0: {
michael@0:     unsigned long i, j;
michael@0: 
michael@0:     if (ccl_used == ccl_size) {
michael@0:         if (ccl_size == 0)
michael@0:           ccl = (unsigned long *) malloc(sizeof(unsigned long) * 24);
michael@0:         else
michael@0:           ccl = (unsigned long *)
michael@0:               realloc((char *) ccl, sizeof(unsigned long) * (ccl_size + 24));
michael@0:         ccl_size += 24;
michael@0:     }
michael@0: 
michael@0:     /*
michael@0:      * Optimize adding the first item.
michael@0:      */
michael@0:     if (ccl_used == 0) {
michael@0:         ccl[0] = ccl[1] = c;
michael@0:         ccl[2] = ccl_code;
michael@0:         ccl_used += 3;
michael@0:         return;
michael@0:     }
michael@0: 
michael@0:     /*
michael@0:      * Handle the special case of extending the range on the end.  This
michael@0:      * requires that the combining class codes are the same.
michael@0:      */
michael@0:     if (ccl_code == ccl[ccl_used - 1] && c == ccl[ccl_used - 2] + 1) {
michael@0:         ccl[ccl_used - 2] = c;
michael@0:         return;
michael@0:     }
michael@0: 
michael@0:     /*
michael@0:      * Handle the special case of adding another range on the end.
michael@0:      */
michael@0:     if (c > ccl[ccl_used - 2] + 1 ||
michael@0:         (c == ccl[ccl_used - 2] + 1 && ccl_code != ccl[ccl_used - 1])) {
michael@0:         ccl[ccl_used++] = c;
michael@0:         ccl[ccl_used++] = c;
michael@0:         ccl[ccl_used++] = ccl_code;
michael@0:         return;
michael@0:     }
michael@0: 
michael@0:     /*
michael@0:      * Locate either the insertion point or range for the code.
michael@0:      */
michael@0:     for (i = 0; i < ccl_used && c > ccl[i + 1] + 1; i += 3) ;
michael@0: 
michael@0:     if (ccl_code == ccl[i + 2] && c == ccl[i + 1] + 1) {
michael@0:         /*
michael@0:          * Extend an existing range.
michael@0:          */
michael@0:         ccl[i + 1] = c;
michael@0:         return;
michael@0:     } else if (c < ccl[i]) {
michael@0:         /*
michael@0:          * Start a new range before the current location.
michael@0:          */
michael@0:         for (j = ccl_used; j > i; j -= 3) {
michael@0:             ccl[j] = ccl[j - 3];
michael@0:             ccl[j - 1] = ccl[j - 4];
michael@0:             ccl[j - 2] = ccl[j - 5];
michael@0:         }
michael@0:         ccl[i] = ccl[i + 1] = c;
michael@0:         ccl[i + 2] = ccl_code;
michael@0:     }
michael@0: }
michael@0: 
michael@0: /*
michael@0:  * Adds a number if it does not already exist and returns an index value
michael@0:  * multiplied by 2.
michael@0:  */
michael@0: static unsigned long
michael@0: #ifdef __STDC__
michael@0: make_number(short num, short denom)
michael@0: #else
michael@0: make_number(num, denom)
michael@0: short num, denom;
michael@0: #endif
michael@0: {
michael@0:     unsigned long n;
michael@0: 
michael@0:     /*
michael@0:      * Determine if the number already exists.
michael@0:      */
michael@0:     for (n = 0; n < nums_used; n++) {
michael@0:         if (nums[n].numerator == num && nums[n].denominator == denom)
michael@0:           return n << 1;
michael@0:     }
michael@0: 
michael@0:     if (nums_used == nums_size) {
michael@0:         if (nums_size == 0)
michael@0:           nums = (_num_t *) malloc(sizeof(_num_t) << 3);
michael@0:         else
michael@0:           nums = (_num_t *) realloc((char *) nums,
michael@0:                                     sizeof(_num_t) * (nums_size + 8));
michael@0:         nums_size += 8;
michael@0:     }
michael@0: 
michael@0:     n = nums_used++;
michael@0:     nums[n].numerator = num;
michael@0:     nums[n].denominator = denom;
michael@0: 
michael@0:     return n << 1;
michael@0: }
michael@0: 
michael@0: static void
michael@0: #ifdef __STDC__
michael@0: add_number(unsigned long code, short num, short denom)
michael@0: #else
michael@0: add_number(code, num, denom)
michael@0: unsigned long code;
michael@0: short num, denom;
michael@0: #endif
michael@0: {
michael@0:     unsigned long i, j;
michael@0: 
michael@0:     /*
michael@0:      * Insert the code in order.
michael@0:      */
michael@0:     for (i = 0; i < ncodes_used && code > ncodes[i].code; i++) ;
michael@0: 
michael@0:     /*
michael@0:      * Handle the case of the codes matching and simply replace the number
michael@0:      * that was there before.
michael@0:      */
michael@0:     if (ncodes_used > 0 && code == ncodes[i].code) {
michael@0:         ncodes[i].idx = make_number(num, denom);
michael@0:         return;
michael@0:     }
michael@0: 
michael@0:     /*
michael@0:      * Resize the array if necessary.
michael@0:      */
michael@0:     if (ncodes_used == ncodes_size) {
michael@0:         if (ncodes_size == 0)
michael@0:           ncodes = (_codeidx_t *) malloc(sizeof(_codeidx_t) << 3);
michael@0:         else
michael@0:           ncodes = (_codeidx_t *)
michael@0:               realloc((char *) ncodes, sizeof(_codeidx_t) * (ncodes_size + 8));
michael@0: 
michael@0:         ncodes_size += 8;
michael@0:     }
michael@0: 
michael@0:     /*
michael@0:      * Shift things around to insert the code if necessary.
michael@0:      */
michael@0:     if (i < ncodes_used) {
michael@0:         for (j = ncodes_used; j > i; j--) {
michael@0:             ncodes[j].code = ncodes[j - 1].code;
michael@0:             ncodes[j].idx = ncodes[j - 1].idx;
michael@0:         }
michael@0:     }
michael@0:     ncodes[i].code = code;
michael@0:     ncodes[i].idx = make_number(num, denom);
michael@0: 
michael@0:     ncodes_used++;
michael@0: }
michael@0: 
michael@0: /*
michael@0:  * This routine assumes that the line is a valid Unicode Character Database
michael@0:  * entry.
michael@0:  */
michael@0: static void
michael@0: #ifdef __STDC__
michael@0: read_cdata(FILE *in)
michael@0: #else
michael@0: read_cdata(in)
michael@0: FILE *in;
michael@0: #endif
michael@0: {
michael@0:     unsigned long i, lineno, skip, code, ccl_code;
michael@0:     short wnum, neg, number[2];
michael@0:     char line[512], *s, *e;
michael@0: 
michael@0:     lineno = skip = 0;
michael@0:     while (fscanf(in, "%[^\n]\n", line) != EOF) {
michael@0:         lineno++;
michael@0: 
michael@0:         /*
michael@0:          * Skip blank lines and lines that start with a '#'.
michael@0:          */
michael@0:         if (line[0] == 0 || line[0] == '#')
michael@0:           continue;
michael@0: 
michael@0:         /*
michael@0:          * If lines need to be skipped, do it here.
michael@0:          */
michael@0:         if (skip) {
michael@0:             skip--;
michael@0:             continue;
michael@0:         }
michael@0: 
michael@0:         /*
michael@0:          * Collect the code.  The code can be up to 6 hex digits in length to
michael@0:          * allow surrogates to be specified.
michael@0:          */
michael@0:         for (s = line, i = code = 0; *s != ';' && i < 6; i++, s++) {
michael@0:             code <<= 4;
michael@0:             if (*s >= '0' && *s <= '9')
michael@0:               code += *s - '0';
michael@0:             else if (*s >= 'A' && *s <= 'F')
michael@0:               code += (*s - 'A') + 10;
michael@0:             else if (*s >= 'a' && *s <= 'f')
michael@0:               code += (*s - 'a') + 10;
michael@0:         }
michael@0: 
michael@0:         /*
michael@0:          * Handle the following special cases:
michael@0:          * 1. 4E00-9FA5 CJK Ideographs.
michael@0:          * 2. AC00-D7A3 Hangul Syllables.
michael@0:          * 3. D800-DFFF Surrogates.
michael@0:          * 4. E000-F8FF Private Use Area.
michael@0:          * 5. F900-FA2D Han compatibility.
michael@0:          */
michael@0:         switch (code) {
michael@0:           case 0x4e00:
michael@0:             /*
michael@0:              * The Han ideographs.
michael@0:              */
michael@0:             add_range(0x4e00, 0x9fff, "Lo", "L");
michael@0: 
michael@0:             /*
michael@0:              * Add the characters to the defined category.
michael@0:              */
michael@0:             add_range(0x4e00, 0x9fa5, "Cp", 0);
michael@0: 
michael@0:             skip = 1;
michael@0:             break;
michael@0:           case 0xac00:
michael@0:             /*
michael@0:              * The Hangul syllables.
michael@0:              */
michael@0:             add_range(0xac00, 0xd7a3, "Lo", "L");
michael@0: 
michael@0:             /*
michael@0:              * Add the characters to the defined category.
michael@0:              */
michael@0:             add_range(0xac00, 0xd7a3, "Cp", 0);
michael@0: 
michael@0:             skip = 1;
michael@0:             break;
michael@0:           case 0xd800:
michael@0:             /*
michael@0:              * Make a range of all surrogates and assume some default
michael@0:              * properties.
michael@0:              */
michael@0:             add_range(0x010000, 0x10ffff, "Cs", "L");
michael@0:             skip = 5;
michael@0:             break;
michael@0:           case 0xe000:
michael@0:             /*
michael@0:              * The Private Use area.  Add with a default set of properties.
michael@0:              */
michael@0:             add_range(0xe000, 0xf8ff, "Co", "L");
michael@0:             skip = 1;
michael@0:             break;
michael@0:           case 0xf900:
michael@0:             /*
michael@0:              * The CJK compatibility area.
michael@0:              */
michael@0:             add_range(0xf900, 0xfaff, "Lo", "L");
michael@0: 
michael@0:             /*
michael@0:              * Add the characters to the defined category.
michael@0:              */
michael@0:             add_range(0xf900, 0xfaff, "Cp", 0);
michael@0: 
michael@0:             skip = 1;
michael@0:         }
michael@0: 
michael@0:         if (skip)
michael@0:           continue;
michael@0: 
michael@0:         /*
michael@0:          * Add the code to the defined category.
michael@0:          */
michael@0:         ordered_range_insert(code, "Cp", 2);
michael@0: 
michael@0:         /*
michael@0:          * Locate the first character property field.
michael@0:          */
michael@0:         for (i = 0; *s != 0 && i < 2; s++) {
michael@0:             if (*s == ';')
michael@0:               i++;
michael@0:         }
michael@0:         for (e = s; *e && *e != ';'; e++) ;
michael@0:     
michael@0:         ordered_range_insert(code, s, e - s);
michael@0: 
michael@0:         /*
michael@0:          * Locate the combining class code.
michael@0:          */
michael@0:         for (s = e; *s != 0 && i < 3; s++) {
michael@0:             if (*s == ';')
michael@0:               i++;
michael@0:         }
michael@0: 
michael@0:         /*
michael@0:          * Convert the combining class code from decimal.
michael@0:          */
michael@0:         for (ccl_code = 0, e = s; *e && *e != ';'; e++)
michael@0:           ccl_code = (ccl_code * 10) + (*e - '0');
michael@0: 
michael@0:         /*
michael@0:          * Add the code if it not 0.
michael@0:          */
michael@0:         if (ccl_code != 0)
michael@0:           ordered_ccl_insert(code, ccl_code);
michael@0: 
michael@0:         /*
michael@0:          * Locate the second character property field.
michael@0:          */
michael@0:         for (s = e; *s != 0 && i < 4; s++) {
michael@0:             if (*s == ';')
michael@0:               i++;
michael@0:         }
michael@0:         for (e = s; *e && *e != ';'; e++) ;
michael@0: 
michael@0:         ordered_range_insert(code, s, e - s);
michael@0: 
michael@0:         /*
michael@0:          * Check for a decomposition.
michael@0:          */
michael@0:         s = ++e;
michael@0:         if (*s != ';' && *s != '<') {
michael@0:             /*
michael@0:              * Collect the codes of the decomposition.
michael@0:              */
michael@0:             for (dectmp_size = 0; *s != ';'; ) {
michael@0:                 /*
michael@0:                  * Skip all leading non-hex digits.
michael@0:                  */
michael@0:                 while (!ishdigit(*s))
michael@0:                   s++;
michael@0: 
michael@0:                 for (dectmp[dectmp_size] = 0; ishdigit(*s); s++) {
michael@0:                     dectmp[dectmp_size] <<= 4;
michael@0:                     if (*s >= '0' && *s <= '9')
michael@0:                       dectmp[dectmp_size] += *s - '0';
michael@0:                     else if (*s >= 'A' && *s <= 'F')
michael@0:                       dectmp[dectmp_size] += (*s - 'A') + 10;
michael@0:                     else if (*s >= 'a' && *s <= 'f')
michael@0:                       dectmp[dectmp_size] += (*s - 'a') + 10;
michael@0:                 }
michael@0:                 dectmp_size++;
michael@0:             }
michael@0: 
michael@0:             /*
michael@0:              * If there is more than one code in the temporary decomposition
michael@0:              * array, then add the character with its decomposition.
michael@0:              */
michael@0:             if (dectmp_size > 1)
michael@0:               add_decomp(code);
michael@0:         }
michael@0: 
michael@0:         /*
michael@0:          * Skip to the number field.
michael@0:          */
michael@0:         for (i = 0; i < 3 && *s; s++) {
michael@0:             if (*s == ';')
michael@0:               i++;
michael@0:         }
michael@0: 
michael@0:         /*
michael@0:          * Scan the number in.
michael@0:          */
michael@0:         number[0] = number[1] = 0;
michael@0:         for (e = s, neg = wnum = 0; *e && *e != ';'; e++) {
michael@0:             if (*e == '-') {
michael@0:                 neg = 1;
michael@0:                 continue;
michael@0:             }
michael@0: 
michael@0:             if (*e == '/') {
michael@0:                 /*
michael@0:                  * Move the the denominator of the fraction.
michael@0:                  */
michael@0:                 if (neg)
michael@0:                   number[wnum] *= -1;
michael@0:                 neg = 0;
michael@0:                 e++;
michael@0:                 wnum++;
michael@0:             }
michael@0:             number[wnum] = (number[wnum] * 10) + (*e - '0');
michael@0:         }
michael@0: 
michael@0:         if (e > s) {
michael@0:             /*
michael@0:              * Adjust the denominator in case of integers and add the number.
michael@0:              */
michael@0:             if (wnum == 0)
michael@0:               number[1] = number[0];
michael@0: 
michael@0:             add_number(code, number[0], number[1]);
michael@0:         }
michael@0: 
michael@0:         /*
michael@0:          * Skip to the start of the possible case mappings.
michael@0:          */
michael@0:         for (s = e, i = 0; i < 4 && *s; s++) {
michael@0:             if (*s == ';')
michael@0:               i++;
michael@0:         }
michael@0: 
michael@0:         /*
michael@0:          * Collect the case mappings.
michael@0:          */
michael@0:         cases[0] = cases[1] = cases[2] = 0;
michael@0:         for (i = 0; i < 3; i++) {
michael@0:             while (ishdigit(*s)) {
michael@0:                 cases[i] <<= 4;
michael@0:                 if (*s >= '0' && *s <= '9')
michael@0:                   cases[i] += *s - '0';
michael@0:                 else if (*s >= 'A' && *s <= 'F')
michael@0:                   cases[i] += (*s - 'A') + 10;
michael@0:                 else if (*s >= 'a' && *s <= 'f')
michael@0:                   cases[i] += (*s - 'a') + 10;
michael@0:                 s++;
michael@0:             }
michael@0:             if (*s == ';')
michael@0:               s++;
michael@0:         }
michael@0:         if (cases[0] && cases[1])
michael@0:           /*
michael@0:            * Add the upper and lower mappings for a title case character.
michael@0:            */
michael@0:           add_title(code);
michael@0:         else if (cases[1])
michael@0:           /*
michael@0:            * Add the lower and title case mappings for the upper case
michael@0:            * character.
michael@0:            */
michael@0:           add_upper(code);
michael@0:         else if (cases[0])
michael@0:           /*
michael@0:            * Add the upper and title case mappings for the lower case
michael@0:            * character.
michael@0:            */
michael@0:           add_lower(code);
michael@0:     }
michael@0: }
michael@0: 
michael@0: static _decomp_t *
michael@0: #ifdef __STDC__
michael@0: find_decomp(unsigned long code)
michael@0: #else
michael@0: find_decomp(code)
michael@0: unsigned long code;
michael@0: #endif
michael@0: {
michael@0:     long l, r, m;
michael@0: 
michael@0:     l = 0;
michael@0:     r = decomps_used - 1;
michael@0:     while (l <= r) {
michael@0:         m = (l + r) >> 1;
michael@0:         if (code > decomps[m].code)
michael@0:           l = m + 1;
michael@0:         else if (code < decomps[m].code)
michael@0:           r = m - 1;
michael@0:         else
michael@0:           return &decomps[m];
michael@0:     }
michael@0:     return 0;
michael@0: }
michael@0: 
michael@0: static void
michael@0: #ifdef __STDC__
michael@0: decomp_it(_decomp_t *d)
michael@0: #else
michael@0: decomp_it(d)
michael@0: _decomp_t *d;
michael@0: #endif
michael@0: {
michael@0:     unsigned long i;
michael@0:     _decomp_t *dp;
michael@0: 
michael@0:     for (i = 0; i < d->used; i++) {
michael@0:         if ((dp = find_decomp(d->decomp[i])) != 0)
michael@0:           decomp_it(dp);
michael@0:         else
michael@0:           dectmp[dectmp_size++] = d->decomp[i];
michael@0:     }
michael@0: }
michael@0: 
michael@0: /*
michael@0:  * Expand all decompositions by recursively decomposing each character
michael@0:  * in the decomposition.
michael@0:  */
michael@0: static void
michael@0: #ifdef __STDC__
michael@0: expand_decomp(void)
michael@0: #else
michael@0: expand_decomp()
michael@0: #endif
michael@0: {
michael@0:     unsigned long i;
michael@0: 
michael@0:     for (i = 0; i < decomps_used; i++) {
michael@0:         dectmp_size = 0;
michael@0:         decomp_it(&decomps[i]);
michael@0:         if (dectmp_size > 0)
michael@0:           add_decomp(decomps[i].code);
michael@0:     }
michael@0: }
michael@0: 
michael@0: static void
michael@0: #ifdef __STDC__
michael@0: write_cdata(char *opath)
michael@0: #else
michael@0: write_cdata(opath)
michael@0: char *opath;
michael@0: #endif
michael@0: {
michael@0:     FILE *out;
michael@0:     unsigned long i, idx, bytes, nprops;
michael@0:     unsigned short casecnt[2];
michael@0:     char path[BUFSIZ];
michael@0: 
michael@0:     /*****************************************************************
michael@0:      *
michael@0:      * Generate the ctype data.
michael@0:      *
michael@0:      *****************************************************************/
michael@0: 
michael@0:     /*
michael@0:      * Open the ctype.dat file.
michael@0:      */
michael@0:     sprintf(path, "%s/ctype.dat", opath);
michael@0:     if ((out = fopen(path, "wb")) == 0)
michael@0:       return;
michael@0: 
michael@0:     /*
michael@0:      * Collect the offsets for the properties.  The offsets array is
michael@0:      * on a 4-byte boundary to keep things efficient for architectures
michael@0:      * that need such a thing.
michael@0:      */
michael@0:     for (i = idx = 0; i < NUMPROPS; i++) {
michael@0:         propcnt[i] = (proptbl[i].used != 0) ? idx : 0xffff;
michael@0:         idx += proptbl[i].used;
michael@0:     }
michael@0: 
michael@0:     /*
michael@0:      * Add the sentinel index which is used by the binary search as the upper
michael@0:      * bound for a search.
michael@0:      */
michael@0:     propcnt[i] = idx;
michael@0: 
michael@0:     /*
michael@0:      * Record the actual number of property lists.  This may be different than
michael@0:      * the number of offsets actually written because of aligning on a 4-byte
michael@0:      * boundary.
michael@0:      */
michael@0:     hdr[1] = NUMPROPS;
michael@0: 
michael@0:     /*
michael@0:      * Calculate the byte count needed and pad the property counts array to a
michael@0:      * 4-byte boundary.
michael@0:      */
michael@0:     if ((bytes = sizeof(unsigned short) * (NUMPROPS + 1)) & 3)
michael@0:       bytes += 4 - (bytes & 3);
michael@0:     nprops = bytes / sizeof(unsigned short);
michael@0:     bytes += sizeof(unsigned long) * idx;
michael@0:         
michael@0:     /*
michael@0:      * Write the header.
michael@0:      */
michael@0:     fwrite((char *) hdr, sizeof(unsigned short), 2, out);
michael@0: 
michael@0:     /*
michael@0:      * Write the byte count.
michael@0:      */
michael@0:     fwrite((char *) &bytes, sizeof(unsigned long), 1, out);
michael@0: 
michael@0:     /*
michael@0:      * Write the property list counts.
michael@0:      */
michael@0:     fwrite((char *) propcnt, sizeof(unsigned short), nprops, out);
michael@0: 
michael@0:     /*
michael@0:      * Write the property lists.
michael@0:      */
michael@0:     for (i = 0; i < NUMPROPS; i++) {
michael@0:         if (proptbl[i].used > 0)
michael@0:           fwrite((char *) proptbl[i].ranges, sizeof(unsigned long),
michael@0:                  proptbl[i].used, out);
michael@0:     }
michael@0: 
michael@0:     fclose(out);
michael@0: 
michael@0:     /*****************************************************************
michael@0:      *
michael@0:      * Generate the case mapping data.
michael@0:      *
michael@0:      *****************************************************************/
michael@0: 
michael@0:     /*
michael@0:      * Open the case.dat file.
michael@0:      */
michael@0:     sprintf(path, "%s/case.dat", opath);
michael@0:     if ((out = fopen(path, "wb")) == 0)
michael@0:       return;
michael@0: 
michael@0:     /*
michael@0:      * Write the case mapping tables.
michael@0:      */
michael@0:     hdr[1] = upper_used + lower_used + title_used;
michael@0:     casecnt[0] = upper_used;
michael@0:     casecnt[1] = lower_used;
michael@0: 
michael@0:     /*
michael@0:      * Write the header.
michael@0:      */
michael@0:     fwrite((char *) hdr, sizeof(unsigned short), 2, out);
michael@0: 
michael@0:     /*
michael@0:      * Write the upper and lower case table sizes.
michael@0:      */
michael@0:     fwrite((char *) casecnt, sizeof(unsigned short), 2, out);
michael@0: 
michael@0:     if (upper_used > 0)
michael@0:       /*
michael@0:        * Write the upper case table.
michael@0:        */
michael@0:       fwrite((char *) upper, sizeof(_case_t), upper_used, out);
michael@0: 
michael@0:     if (lower_used > 0)
michael@0:       /*
michael@0:        * Write the lower case table.
michael@0:        */
michael@0:       fwrite((char *) lower, sizeof(_case_t), lower_used, out);
michael@0: 
michael@0:     if (title_used > 0)
michael@0:       /*
michael@0:        * Write the title case table.
michael@0:        */
michael@0:       fwrite((char *) title, sizeof(_case_t), title_used, out);
michael@0: 
michael@0:     fclose(out);
michael@0: 
michael@0:     /*****************************************************************
michael@0:      *
michael@0:      * Generate the decomposition data.
michael@0:      *
michael@0:      *****************************************************************/
michael@0: 
michael@0:     /*
michael@0:      * Fully expand all decompositions before generating the output file.
michael@0:      */
michael@0:     expand_decomp();
michael@0: 
michael@0:     /*
michael@0:      * Open the decomp.dat file.
michael@0:      */
michael@0:     sprintf(path, "%s/decomp.dat", opath);
michael@0:     if ((out = fopen(path, "wb")) == 0)
michael@0:       return;
michael@0: 
michael@0:     hdr[1] = decomps_used;
michael@0: 
michael@0:     /*
michael@0:      * Write the header.
michael@0:      */
michael@0:     fwrite((char *) hdr, sizeof(unsigned short), 2, out);
michael@0: 
michael@0:     /*
michael@0:      * Write a temporary byte count which will be calculated as the
michael@0:      * decompositions are written out.
michael@0:      */
michael@0:     bytes = 0;
michael@0:     fwrite((char *) &bytes, sizeof(unsigned long), 1, out);
michael@0: 
michael@0:     if (decomps_used) {
michael@0:         /*
michael@0:          * Write the list of decomp nodes.
michael@0:          */
michael@0:         for (i = idx = 0; i < decomps_used; i++) {
michael@0:             fwrite((char *) &decomps[i].code, sizeof(unsigned long), 1, out);
michael@0:             fwrite((char *) &idx, sizeof(unsigned long), 1, out);
michael@0:             idx += decomps[i].used;
michael@0:         }
michael@0: 
michael@0:         /*
michael@0:          * Write the sentinel index as the last decomp node.
michael@0:          */
michael@0:         fwrite((char *) &idx, sizeof(unsigned long), 1, out);
michael@0: 
michael@0:         /*
michael@0:          * Write the decompositions themselves.
michael@0:          */
michael@0:         for (i = 0; i < decomps_used; i++)
michael@0:           fwrite((char *) decomps[i].decomp, sizeof(unsigned long),
michael@0:                  decomps[i].used, out);
michael@0: 
michael@0:         /*
michael@0:          * Seek back to the beginning and write the byte count.
michael@0:          */
michael@0:         bytes = (sizeof(unsigned long) * idx) +
michael@0:             (sizeof(unsigned long) * ((hdr[1] << 1) + 1));
michael@0:         fseek(out, sizeof(unsigned short) << 1, 0L);
michael@0:         fwrite((char *) &bytes, sizeof(unsigned long), 1, out);
michael@0: 
michael@0:         fclose(out);
michael@0:     }
michael@0: 
michael@0:     /*****************************************************************
michael@0:      *
michael@0:      * Generate the combining class data.
michael@0:      *
michael@0:      *****************************************************************/
michael@0: 
michael@0:     /*
michael@0:      * Open the cmbcl.dat file.
michael@0:      */
michael@0:     sprintf(path, "%s/cmbcl.dat", opath);
michael@0:     if ((out = fopen(path, "wb")) == 0)
michael@0:       return;
michael@0: 
michael@0:     /*
michael@0:      * Set the number of ranges used.  Each range has a combining class which
michael@0:      * means each entry is a 3-tuple.
michael@0:      */
michael@0:     hdr[1] = ccl_used / 3;
michael@0: 
michael@0:     /*
michael@0:      * Write the header.
michael@0:      */
michael@0:     fwrite((char *) hdr, sizeof(unsigned short), 2, out);
michael@0: 
michael@0:     /*
michael@0:      * Write out the byte count to maintain header size.
michael@0:      */
michael@0:     bytes = ccl_used * sizeof(unsigned long);
michael@0:     fwrite((char *) &bytes, sizeof(unsigned long), 1, out);
michael@0: 
michael@0:     if (ccl_used > 0)
michael@0:       /*
michael@0:        * Write the combining class ranges out.
michael@0:        */
michael@0:       fwrite((char *) ccl, sizeof(unsigned long), ccl_used, out);
michael@0: 
michael@0:     fclose(out);
michael@0: 
michael@0:     /*****************************************************************
michael@0:      *
michael@0:      * Generate the number data.
michael@0:      *
michael@0:      *****************************************************************/
michael@0: 
michael@0:     /*
michael@0:      * Open the num.dat file.
michael@0:      */
michael@0:     sprintf(path, "%s/num.dat", opath);
michael@0:     if ((out = fopen(path, "wb")) == 0)
michael@0:       return;
michael@0: 
michael@0:     /*
michael@0:      * The count part of the header will be the total number of codes that
michael@0:      * have numbers.
michael@0:      */
michael@0:     hdr[1] = (unsigned short) (ncodes_used << 1);
michael@0:     bytes = (ncodes_used * sizeof(_codeidx_t)) + (nums_used * sizeof(_num_t));
michael@0: 
michael@0:     /*
michael@0:      * Write the header.
michael@0:      */
michael@0:     fwrite((char *) hdr, sizeof(unsigned short), 2, out);
michael@0: 
michael@0:     /*
michael@0:      * Write out the byte count to maintain header size.
michael@0:      */
michael@0:     fwrite((char *) &bytes, sizeof(unsigned long), 1, out);
michael@0: 
michael@0:     /*
michael@0:      * Now, if number mappings exist, write them out.
michael@0:      */
michael@0:     if (ncodes_used > 0) {
michael@0:         fwrite((char *) ncodes, sizeof(_codeidx_t), ncodes_used, out);
michael@0:         fwrite((char *) nums, sizeof(_num_t), nums_used, out);
michael@0:     }
michael@0: 
michael@0:     fclose(out);
michael@0: }
michael@0: 
michael@0: void
michael@0: #ifdef __STDC__
michael@0: main(int argc, char *argv[])
michael@0: #else
michael@0: main(argc, argv)
michael@0: int argc;
michael@0: char *argv[];
michael@0: #endif
michael@0: {
michael@0:     FILE *in;
michael@0:     char *prog, *opath;
michael@0: 
michael@0:     if ((prog = strrchr(argv[0], '/')) != 0)
michael@0:       prog++;
michael@0:     else
michael@0:       prog = argv[0];
michael@0: 
michael@0:     opath = 0;
michael@0:     in = stdin;
michael@0: 
michael@0:     argc--;
michael@0:     argv++;
michael@0: 
michael@0:     while (argc > 0) {
michael@0:         if (argv[0][0] == '-' && argv[0][1] == 'o') {
michael@0:             argc--;
michael@0:             argv++;
michael@0:             opath = argv[0];
michael@0:         } else {
michael@0:             if (in != stdin)
michael@0:               fclose(in);
michael@0:             if ((in = fopen(argv[0], "rb")) == 0)
michael@0:               fprintf(stderr, "%s: unable to open ctype file %s\n",
michael@0:                       prog, argv[0]);
michael@0:             else {
michael@0:                 read_cdata(in);
michael@0:                 fclose(in);
michael@0:                 in = 0;
michael@0:             }
michael@0:         }
michael@0:         argc--;
michael@0:         argv++;
michael@0:     }
michael@0: 
michael@0:     if (opath == 0)
michael@0:       opath = ".";
michael@0:     write_cdata(opath);
michael@0: 
michael@0:     exit(0);
michael@0: }