/* $OpenLDAP$ */
/* This work is part of OpenLDAP Software .
*
* Copyright 1998-2005 The OpenLDAP Foundation.
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted only as authorized by the OpenLDAP
* Public License.
*
* A copy of this license is available in file LICENSE in the
* top-level directory of the distribution or, alternatively, at
* .
*/
/* Copyright 2001 Computing Research Labs, New Mexico State University
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* THE COMPUTING RESEARCH LAB OR NEW MEXICO STATE UNIVERSITY BE LIABLE FOR ANY
* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT
* OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR
* THE USE OR OTHER DEALINGS IN THE SOFTWARE.
*/
/* $Id: ucpgba.c,v 1.5 2001/01/02 18:46:20 mleisher Exp $ */
#include "portable.h"
#include
#include
#include "ucdata.h"
#include "ucpgba.h"
/*
* These macros are used while reordering of RTL runs of text for the
* special case of non-spacing characters being in runs of weakly
* directional text. They check for weak and non-spacing, and digits and
* non-spacing.
*/
#define ISWEAKSPECIAL(cc) ucisprop(cc, UC_EN|UC_ES|UC_MN, UC_ET|UC_AN|UC_CS)
#define ISDIGITSPECIAL(cc) ucisprop(cc, UC_ND|UC_MN, 0)
/*
* These macros are used while breaking a string into runs of text in
* different directions. Descriptions:
*
* ISLTR_LTR - Test for members of an LTR run in an LTR context. This looks
* for characters with ltr, non-spacing, weak, and neutral
* properties.
*
* ISRTL_RTL - Test for members of an RTL run in an RTL context. This looks
* for characters with rtl, non-spacing, weak, and neutral
* properties.
*
* ISRTL_NEUTRAL - Test for RTL or neutral characters.
*
* ISWEAK_NEUTRAL - Test for weak or neutral characters.
*/
#define ISLTR_LTR(cc) ucisprop(cc, UC_L|UC_MN|UC_EN|UC_ES,\
UC_ET|UC_CS|UC_B|UC_S|UC_WS|UC_ON)
#define ISRTL_RTL(cc) ucisprop(cc, UC_R|UC_MN|UC_EN|UC_ES,\
UC_ET|UC_AN|UC_CS|UC_B|UC_S|UC_WS|UC_ON)
#define ISRTL_NEUTRAL(cc) ucisprop(cc, UC_R, UC_B|UC_S|UC_WS|UC_ON)
#define ISWEAK_NEUTRAL(cc) ucisprop(cc, UC_EN|UC_ES, \
UC_B|UC_S|UC_WS|UC_ON|UC_ET|UC_AN|UC_CS)
/*
* This table is temporarily hard-coded here until it can be constructed
* automatically somehow.
*/
static unsigned long _symmetric_pairs[] = {
0x0028, 0x0029, 0x0029, 0x0028, 0x003C, 0x003E, 0x003E, 0x003C,
0x005B, 0x005D, 0x005D, 0x005B, 0x007B, 0x007D, 0x007D, 0x007B,
0x2045, 0x2046, 0x2046, 0x2045, 0x207D, 0x207E, 0x207E, 0x207D,
0x208D, 0x208E, 0x208E, 0x208D, 0x3008, 0x3009, 0x3009, 0x3008,
0x300A, 0x300B, 0x300B, 0x300A, 0x300C, 0x300D, 0x300D, 0x300C,
0x300E, 0x300F, 0x300F, 0x300E, 0x3010, 0x3011, 0x3011, 0x3010,
0x3014, 0x3015, 0x3015, 0x3014, 0x3016, 0x3017, 0x3017, 0x3016,
0x3018, 0x3019, 0x3019, 0x3018, 0x301A, 0x301B, 0x301B, 0x301A,
0xFD3E, 0xFD3F, 0xFD3F, 0xFD3E, 0xFE59, 0xFE5A, 0xFE5A, 0xFE59,
0xFE5B, 0xFE5C, 0xFE5C, 0xFE5B, 0xFE5D, 0xFE5E, 0xFE5E, 0xFE5D,
0xFF08, 0xFF09, 0xFF09, 0xFF08, 0xFF3B, 0xFF3D, 0xFF3D, 0xFF3B,
0xFF5B, 0xFF5D, 0xFF5D, 0xFF5B, 0xFF62, 0xFF63, 0xFF63, 0xFF62,
};
static int _symmetric_pairs_size =
sizeof(_symmetric_pairs)/sizeof(_symmetric_pairs[0]);
/*
* This routine looks up the other form of a symmetric pair.
*/
static unsigned long
_ucsymmetric_pair(unsigned long c)
{
int i;
for (i = 0; i < _symmetric_pairs_size; i += 2) {
if (_symmetric_pairs[i] == c)
return _symmetric_pairs[i+1];
}
return c;
}
/*
* This routine creates a new run, copies the text into it, links it into the
* logical text order chain and returns it to the caller to be linked into
* the visual text order chain.
*/
static ucrun_t *
_add_run(ucstring_t *str, unsigned long *src,
unsigned long start, unsigned long end, int direction)
{
long i, t;
ucrun_t *run;
run = (ucrun_t *) malloc(sizeof(ucrun_t));
run->visual_next = run->visual_prev = 0;
run->direction = direction;
run->cursor = ~0;
run->chars = (unsigned long *)
malloc(sizeof(unsigned long) * ((end - start) << 1));
run->positions = run->chars + (end - start);
run->source = src;
run->start = start;
run->end = end;
if (direction == UCPGBA_RTL) {
/*
* Copy the source text into the run in reverse order and select
* replacements for the pairwise punctuation and the <> characters.
*/
for (i = 0, t = end - 1; start < end; start++, t--, i++) {
run->positions[i] = t;
if (ucissymmetric(src[t]) || src[t] == '<' || src[t] == '>')
run->chars[i] = _ucsymmetric_pair(src[t]);
else
run->chars[i] = src[t];
}
} else {
/*
* Copy the source text into the run directly.
*/
for (i = start; i < end; i++) {
run->positions[i - start] = i;
run->chars[i - start] = src[i];
}
}
/*
* Add the run to the logical list for cursor traversal.
*/
if (str->logical_first == 0)
str->logical_first = str->logical_last = run;
else {
run->logical_prev = str->logical_last;
str->logical_last->logical_next = run;
str->logical_last = run;
}
return run;
}
static void
_ucadd_rtl_segment(ucstring_t *str, unsigned long *source, unsigned long start,
unsigned long end)
{
unsigned long s, e;
ucrun_t *run, *lrun;
/*
* This is used to splice runs into strings with overall LTR direction.
* The `lrun' variable will never be NULL because at least one LTR run was
* added before this RTL run.
*/
lrun = str->visual_last;
for (e = s = start; s < end;) {
for (; e < end && ISRTL_NEUTRAL(source[e]); e++) ;
if (e > s) {
run = _add_run(str, source, s, e, UCPGBA_RTL);
/*
* Add the run to the visual list for cursor traversal.
*/
if (str->visual_first != 0) {
if (str->direction == UCPGBA_LTR) {
run->visual_prev = lrun;
run->visual_next = lrun->visual_next;
if (lrun->visual_next != 0)
lrun->visual_next->visual_prev = run;
lrun->visual_next = run;
if (lrun == str->visual_last)
str->visual_last = run;
} else {
run->visual_next = str->visual_first;
str->visual_first->visual_prev = run;
str->visual_first = run;
}
} else
str->visual_first = str->visual_last = run;
}
/*
* Handle digits in a special way. This makes sure the weakly
* directional characters appear on the expected sides of a number
* depending on whether that number is Arabic or not.
*/
for (s = e; e < end && ISWEAKSPECIAL(source[e]); e++) {
if (!ISDIGITSPECIAL(source[e]) &&
(e + 1 == end || !ISDIGITSPECIAL(source[e + 1])))
break;
}
if (e > s) {
run = _add_run(str, source, s, e, UCPGBA_LTR);
/*
* Add the run to the visual list for cursor traversal.
*/
if (str->visual_first != 0) {
if (str->direction == UCPGBA_LTR) {
run->visual_prev = lrun;
run->visual_next = lrun->visual_next;
if (lrun->visual_next != 0)
lrun->visual_next->visual_prev = run;
lrun->visual_next = run;
if (lrun == str->visual_last)
str->visual_last = run;
} else {
run->visual_next = str->visual_first;
str->visual_first->visual_prev = run;
str->visual_first = run;
}
} else
str->visual_first = str->visual_last = run;
}
/*
* Collect all weak non-digit sequences for an RTL segment. These
* will appear as part of the next RTL segment or will be added as
* an RTL segment by themselves.
*/
for (s = e; e < end && ucisweak(source[e]) && !ucisdigit(source[e]);
e++) ;
}
/*
* Capture any weak non-digit sequences that occur at the end of the RTL
* run.
*/
if (e > s) {
run = _add_run(str, source, s, e, UCPGBA_RTL);
/*
* Add the run to the visual list for cursor traversal.
*/
if (str->visual_first != 0) {
if (str->direction == UCPGBA_LTR) {
run->visual_prev = lrun;
run->visual_next = lrun->visual_next;
if (lrun->visual_next != 0)
lrun->visual_next->visual_prev = run;
lrun->visual_next = run;
if (lrun == str->visual_last)
str->visual_last = run;
} else {
run->visual_next = str->visual_first;
str->visual_first->visual_prev = run;
str->visual_first = run;
}
} else
str->visual_first = str->visual_last = run;
}
}
static void
_ucadd_ltr_segment(ucstring_t *str, unsigned long *source, unsigned long start,
unsigned long end)
{
ucrun_t *run;
run = _add_run(str, source, start, end, UCPGBA_LTR);
/*
* Add the run to the visual list for cursor traversal.
*/
if (str->visual_first != 0) {
if (str->direction == UCPGBA_LTR) {
run->visual_prev = str->visual_last;
str->visual_last->visual_next = run;
str->visual_last = run;
} else {
run->visual_next = str->visual_first;
str->visual_first->visual_prev = run;
str->visual_first = run;
}
} else
str->visual_first = str->visual_last = run;
}
ucstring_t *
ucstring_create(unsigned long *source, unsigned long start, unsigned long end,
int default_direction, int cursor_motion)
{
int rtl_first;
unsigned long s, e, ld;
ucstring_t *str;
str = (ucstring_t *) malloc(sizeof(ucstring_t));
/*
* Set the initial values.
*/
str->cursor_motion = cursor_motion;
str->logical_first = str->logical_last = 0;
str->visual_first = str->visual_last = str->cursor = 0;
str->source = source;
str->start = start;
str->end = end;
/*
* If the length of the string is 0, then just return it at this point.
*/
if (start == end)
return str;
/*
* This flag indicates whether the collection loop for RTL is called
* before the LTR loop the first time.
*/
rtl_first = 0;
/*
* Look for the first character in the string that has strong
* directionality.
*/
for (s = start; s < end && !ucisstrong(source[s]); s++) ;
if (s == end)
/*
* If the string contains no characters with strong directionality, use
* the default direction.
*/
str->direction = default_direction;
else
str->direction = ucisrtl(source[s]) ? UCPGBA_RTL : UCPGBA_LTR;
if (str->direction == UCPGBA_RTL)
/*
* Set the flag that causes the RTL collection loop to run first.
*/
rtl_first = 1;
/*
* This loop now separates the string into runs based on directionality.
*/
for (s = e = 0; s < end; s = e) {
if (!rtl_first) {
/*
* Determine the next run of LTR text.
*/
ld = s;
while (e < end && ISLTR_LTR(source[e])) {
if (ucisdigit(source[e]) &&
!(0x660 <= source[e] && source[e] <= 0x669))
ld = e;
e++;
}
if (str->direction != UCPGBA_LTR) {
while (e > ld && ISWEAK_NEUTRAL(source[e - 1]))
e--;
}
/*
* Add the LTR segment to the string.
*/
if (e > s)
_ucadd_ltr_segment(str, source, s, e);
}
/*
* Determine the next run of RTL text.
*/
ld = s = e;
while (e < end && ISRTL_RTL(source[e])) {
if (ucisdigit(source[e]) &&
!(0x660 <= source[e] && source[e] <= 0x669))
ld = e;
e++;
}
if (str->direction != UCPGBA_RTL) {
while (e > ld && ISWEAK_NEUTRAL(source[e - 1]))
e--;
}
/*
* Add the RTL segment to the string.
*/
if (e > s)
_ucadd_rtl_segment(str, source, s, e);
/*
* Clear the flag that allowed the RTL collection loop to run first
* for strings with overall RTL directionality.
*/
rtl_first = 0;
}
/*
* Set up the initial cursor run.
*/
str->cursor = str->logical_first;
if (str != 0)
str->cursor->cursor = (str->cursor->direction == UCPGBA_RTL) ?
str->cursor->end - str->cursor->start : 0;
return str;
}
void
ucstring_free(ucstring_t *s)
{
ucrun_t *l, *r;
if (s == 0)
return;
for (l = 0, r = s->visual_first; r != 0; r = r->visual_next) {
if (r->end > r->start)
free((char *) r->chars);
if (l)
free((char *) l);
l = r;
}
if (l)
free((char *) l);
free((char *) s);
}
int
ucstring_set_cursor_motion(ucstring_t *str, int cursor_motion)
{
int n;
if (str == 0)
return -1;
n = str->cursor_motion;
str->cursor_motion = cursor_motion;
return n;
}
static int
_ucstring_visual_cursor_right(ucstring_t *str, int count)
{
int cnt = count;
unsigned long size;
ucrun_t *cursor;
if (str == 0)
return 0;
cursor = str->cursor;
while (cnt > 0) {
size = cursor->end - cursor->start;
if ((cursor->direction == UCPGBA_RTL && cursor->cursor + 1 == size) ||
cursor->cursor + 1 > size) {
/*
* If the next run is NULL, then the cursor is already on the
* far right end already.
*/
if (cursor->visual_next == 0)
/*
* If movement occured, then report it.
*/
return (cnt != count);
/*
* Move to the next run.
*/
str->cursor = cursor = cursor->visual_next;
cursor->cursor = (cursor->direction == UCPGBA_RTL) ? -1 : 0;
size = cursor->end - cursor->start;
} else
cursor->cursor++;
cnt--;
}
return 1;
}
static int
_ucstring_logical_cursor_right(ucstring_t *str, int count)
{
int cnt = count;
unsigned long size;
ucrun_t *cursor;
if (str == 0)
return 0;
cursor = str->cursor;
while (cnt > 0) {
size = cursor->end - cursor->start;
if (str->direction == UCPGBA_RTL) {
if (cursor->direction == UCPGBA_RTL) {
if (cursor->cursor + 1 == size) {
if (cursor == str->logical_first)
/*
* Already at the beginning of the string.
*/
return (cnt != count);
str->cursor = cursor = cursor->logical_prev;
size = cursor->end - cursor->start;
cursor->cursor = (cursor->direction == UCPGBA_LTR) ?
size : 0;
} else
cursor->cursor++;
} else {
if (cursor->cursor == 0) {
if (cursor == str->logical_first)
/*
* At the beginning of the string already.
*/
return (cnt != count);
str->cursor = cursor = cursor->logical_prev;
size = cursor->end - cursor->start;
cursor->cursor = (cursor->direction == UCPGBA_LTR) ?
size : 0;
} else
cursor->cursor--;
}
} else {
if (cursor->direction == UCPGBA_RTL) {
if (cursor->cursor == 0) {
if (cursor == str->logical_last)
/*
* Already at the end of the string.
*/
return (cnt != count);
str->cursor = cursor = cursor->logical_next;
size = cursor->end - cursor->start;
cursor->cursor = (cursor->direction == UCPGBA_LTR) ?
0 : size - 1;
} else
cursor->cursor--;
} else {
if (cursor->cursor + 1 > size) {
if (cursor == str->logical_last)
/*
* Already at the end of the string.
*/
return (cnt != count);
str->cursor = cursor = cursor->logical_next;
cursor->cursor = (cursor->direction == UCPGBA_LTR) ?
0 : size - 1;
} else
cursor->cursor++;
}
}
cnt--;
}
return 1;
}
int
ucstring_cursor_right(ucstring_t *str, int count)
{
if (str == 0)
return 0;
return (str->cursor_motion == UCPGBA_CURSOR_VISUAL) ?
_ucstring_visual_cursor_right(str, count) :
_ucstring_logical_cursor_right(str, count);
}
static int
_ucstring_visual_cursor_left(ucstring_t *str, int count)
{
int cnt = count;
unsigned long size;
ucrun_t *cursor;
if (str == 0)
return 0;
cursor = str->cursor;
while (cnt > 0) {
size = cursor->end - cursor->start;
if ((cursor->direction == UCPGBA_LTR && cursor->cursor == 0) ||
cursor->cursor - 1 < -1) {
/*
* If the preceding run is NULL, then the cursor is already on the
* far left end already.
*/
if (cursor->visual_prev == 0)
/*
* If movement occured, then report it.
*/
return (cnt != count);
/*
* Move to the previous run.
*/
str->cursor = cursor = cursor->visual_prev;
size = cursor->end - cursor->start;
cursor->cursor = (cursor->direction == UCPGBA_RTL) ?
size : size - 1;
} else
cursor->cursor--;
cnt--;
}
return 1;
}
static int
_ucstring_logical_cursor_left(ucstring_t *str, int count)
{
int cnt = count;
unsigned long size;
ucrun_t *cursor;
if (str == 0)
return 0;
cursor = str->cursor;
while (cnt > 0) {
size = cursor->end - cursor->start;
if (str->direction == UCPGBA_RTL) {
if (cursor->direction == UCPGBA_RTL) {
if (cursor->cursor == -1) {
if (cursor == str->logical_last)
/*
* Already at the end of the string.
*/
return (cnt != count);
str->cursor = cursor = cursor->logical_next;
size = cursor->end - cursor->start;
cursor->cursor = (cursor->direction == UCPGBA_LTR) ?
0 : size - 1;
} else
cursor->cursor--;
} else {
if (cursor->cursor + 1 > size) {
if (cursor == str->logical_last)
/*
* At the end of the string already.
*/
return (cnt != count);
str->cursor = cursor = cursor->logical_next;
size = cursor->end - cursor->start;
cursor->cursor = (cursor->direction == UCPGBA_LTR) ?
0 : size - 1;
} else
cursor->cursor++;
}
} else {
if (cursor->direction == UCPGBA_RTL) {
if (cursor->cursor + 1 == size) {
if (cursor == str->logical_first)
/*
* Already at the beginning of the string.
*/
return (cnt != count);
str->cursor = cursor = cursor->logical_prev;
size = cursor->end - cursor->start;
cursor->cursor = (cursor->direction == UCPGBA_LTR) ?
size : 0;
} else
cursor->cursor++;
} else {
if (cursor->cursor == 0) {
if (cursor == str->logical_first)
/*
* Already at the beginning of the string.
*/
return (cnt != count);
str->cursor = cursor = cursor->logical_prev;
cursor->cursor = (cursor->direction == UCPGBA_LTR) ?
size : 0;
} else
cursor->cursor--;
}
}
cnt--;
}
return 1;
}
int
ucstring_cursor_left(ucstring_t *str, int count)
{
if (str == 0)
return 0;
return (str->cursor_motion == UCPGBA_CURSOR_VISUAL) ?
_ucstring_visual_cursor_left(str, count) :
_ucstring_logical_cursor_left(str, count);
}
void
ucstring_cursor_info(ucstring_t *str, int *direction, unsigned long *position)
{
long c;
unsigned long size;
ucrun_t *cursor;
if (str == 0 || direction == 0 || position == 0)
return;
cursor = str->cursor;
*direction = cursor->direction;
c = cursor->cursor;
size = cursor->end - cursor->start;
if (c == size)
*position = (cursor->direction == UCPGBA_RTL) ?
cursor->start : cursor->positions[c - 1];
else if (c == -1)
*position = (cursor->direction == UCPGBA_RTL) ?
cursor->end : cursor->start;
else
*position = cursor->positions[c];
}