worldspawn/libs/splines/q_shared.cpp

1012 lines
18 KiB
C++

/*
Copyright (C) 1999-2006 Id Software, Inc. and contributors.
For a list of contributors, see the accompanying CONTRIBUTORS file.
This file is part of GtkRadiant.
GtkRadiant 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 2 of the License, or
(at your option) any later version.
GtkRadiant 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 GtkRadiant; if not, write to the Free Software
Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
*/
// q_shared.c -- stateless support routines that are included in each code dll
#include "q_shared.h"
/*
============================================================================
GROWLISTS
============================================================================
*/
// malloc / free all in one place for debugging
extern "C" void *Com_Allocate( int bytes );
extern "C" void Com_Dealloc( void *ptr );
void Com_InitGrowList( growList_t *list, int maxElements ) {
list->maxElements = maxElements;
list->currentElements = 0;
list->elements = (void **)Com_Allocate( list->maxElements * sizeof( void * ) );
}
int Com_AddToGrowList( growList_t *list, void *data ) {
void **old;
if ( list->currentElements != list->maxElements ) {
list->elements[list->currentElements] = data;
return list->currentElements++;
}
// grow, reallocate and move
old = list->elements;
if ( list->maxElements < 0 ) {
Com_Error( ERR_FATAL, "Com_AddToGrowList: maxElements = %i", list->maxElements );
}
if ( list->maxElements == 0 ) {
// initialize the list to hold 100 elements
Com_InitGrowList( list, 100 );
return Com_AddToGrowList( list, data );
}
list->maxElements *= 2;
Com_DPrintf( "Resizing growlist to %i maxElements\n", list->maxElements );
list->elements = (void **)Com_Allocate( list->maxElements * sizeof( void * ) );
if ( !list->elements ) {
Com_Error( ERR_DROP, "Growlist alloc failed" );
}
memcpy( list->elements, old, list->currentElements * sizeof( void * ) );
Com_Dealloc( old );
return Com_AddToGrowList( list, data );
}
void *Com_GrowListElement( const growList_t *list, int index ) {
if ( index < 0 || index >= list->currentElements ) {
Com_Error( ERR_DROP, "Com_GrowListElement: %i out of range of %i",
index, list->currentElements );
}
return list->elements[index];
}
int Com_IndexForGrowListElement( const growList_t *list, const void *element ) {
int i;
for ( i = 0; i < list->currentElements; i++ ) {
if ( list->elements[i] == element ) {
return i;
}
}
return -1;
}
//============================================================================
float Com_Clamp( float min, float max, float value ) {
if ( value < min ) {
return min;
}
if ( value > max ) {
return max;
}
return value;
}
/*
============
Com_StringContains
============
*/
const char *Com_StringContains( const char *str1, const char *str2, int casesensitive ) {
int len, i, j;
len = strlen( str1 ) - strlen( str2 );
for ( i = 0; i <= len; i++, str1++ ) {
for ( j = 0; str2[j]; j++ ) {
if ( casesensitive ) {
if ( str1[j] != str2[j] ) {
break;
}
}
else {
if ( toupper( str1[j] ) != toupper( str2[j] ) ) {
break;
}
}
}
if ( !str2[j] ) {
return str1;
}
}
return NULL;
}
/*
============
Com_Filter
============
*/
int Com_Filter( const char *filter, const char *name, int casesensitive ){
char buf[MAX_TOKEN_CHARS];
const char *ptr;
int i, found;
while ( *filter ) {
if ( *filter == '*' ) {
filter++;
for ( i = 0; *filter; i++ ) {
if ( *filter == '*' || *filter == '?' ) {
break;
}
buf[i] = *filter;
filter++;
}
buf[i] = '\0';
if ( strlen( buf ) ) {
ptr = Com_StringContains( name, buf, casesensitive );
if ( !ptr ) {
return qfalse;
}
name = ptr + strlen( buf );
}
}
else if ( *filter == '?' ) {
filter++;
name++;
}
else if ( *filter == '[' && *( filter + 1 ) == '[' ) {
filter++;
}
else if ( *filter == '[' ) {
filter++;
found = qfalse;
while ( *filter && !found ) {
if ( *filter == ']' && *( filter + 1 ) != ']' ) {
break;
}
if ( *( filter + 1 ) == '-' && *( filter + 2 ) && ( *( filter + 2 ) != ']' || *( filter + 3 ) == ']' ) ) {
if ( casesensitive ) {
if ( *name >= *filter && *name <= *( filter + 2 ) ) {
found = qtrue;
}
}
else {
if ( toupper( *name ) >= toupper( *filter ) &&
toupper( *name ) <= toupper( *( filter + 2 ) ) ) {
found = qtrue;
}
}
filter += 3;
}
else {
if ( casesensitive ) {
if ( *filter == *name ) {
found = qtrue;
}
}
else {
if ( toupper( *filter ) == toupper( *name ) ) {
found = qtrue;
}
}
filter++;
}
}
if ( !found ) {
return qfalse;
}
while ( *filter ) {
if ( *filter == ']' && *( filter + 1 ) != ']' ) {
break;
}
filter++;
}
filter++;
name++;
}
else {
if ( casesensitive ) {
if ( *filter != *name ) {
return qfalse;
}
}
else {
if ( toupper( *filter ) != toupper( *name ) ) {
return qfalse;
}
}
filter++;
name++;
}
}
return qtrue;
}
/*
================
Com_HashString
================
*/
int Com_HashString( const char *fname ) {
int i;
long hash;
char letter;
hash = 0;
i = 0;
while ( fname[i] != '\0' ) {
letter = tolower( fname[i] );
if ( letter == '.' ) {
break; // don't include extension
}
if ( letter == '\\' ) {
letter = '/'; // damn path names
}
hash += (long)( letter ) * ( i + 119 );
i++;
}
hash &= ( FILE_HASH_SIZE - 1 );
return hash;
}
/*
============
Com_SkipPath
============
*/
char *Com_SkipPath( char *pathname ){
char *last;
last = pathname;
while ( *pathname )
{
if ( *pathname == '/' ) {
last = pathname + 1;
}
pathname++;
}
return last;
}
/*
============
Com_StripExtension
============
*/
void Com_StripExtension( const char *in, char *out ) {
while ( *in && *in != '.' ) {
*out++ = *in++;
}
*out = 0;
}
/*
==================
Com_DefaultExtension
==================
*/
void Com_DefaultExtension( char *path, int maxSize, const char *extension ) {
char oldPath[MAX_QPATH];
char *src;
//
// if path doesn't have a .EXT, append extension
// (extension should include the .)
//
src = path + strlen( path ) - 1;
while ( *src != '/' && src != path ) {
if ( *src == '.' ) {
return; // it has an extension
}
src--;
}
Q_strncpyz( oldPath, path, sizeof( oldPath ) );
Com_sprintf( path, maxSize, "%s%s", oldPath, extension );
}
/*
============================================================================
BYTE ORDER FUNCTIONS
============================================================================
*/
// can't just use function pointers, or dll linkage can
// mess up when qcommon is included in multiple places
static short ( *_BigShort )( short l );
static short ( *_LittleShort )( short l );
static int ( *_BigLong )( int l );
static int ( *_LittleLong )( int l );
static float ( *_BigFloat )( float l );
static float ( *_LittleFloat )( float l );
short BigShort( short l ){
return _BigShort( l );
}
short LittleShort( short l ) {
return _LittleShort( l );
}
int BigLong( int l ) {
return _BigLong( l );
}
int LittleLong( int l ) {
return _LittleLong( l );
}
float BigFloat( float l ) {
return _BigFloat( l );
}
float LittleFloat( float l ) {
return _LittleFloat( l );
}
short ShortSwap( short l ){
byte b1,b2;
b1 = l & 255;
b2 = ( l >> 8 ) & 255;
return ( b1 << 8 ) + b2;
}
short ShortNoSwap( short l ){
return l;
}
int LongSwap( int l ){
byte b1,b2,b3,b4;
b1 = l & 255;
b2 = ( l >> 8 ) & 255;
b3 = ( l >> 16 ) & 255;
b4 = ( l >> 24 ) & 255;
return ( (int)b1 << 24 ) + ( (int)b2 << 16 ) + ( (int)b3 << 8 ) + b4;
}
int LongNoSwap( int l ){
return l;
}
float FloatSwap( float f ){
union
{
float f;
byte b[4];
} dat1, dat2;
dat1.f = f;
dat2.b[0] = dat1.b[3];
dat2.b[1] = dat1.b[2];
dat2.b[2] = dat1.b[1];
dat2.b[3] = dat1.b[0];
return dat2.f;
}
float FloatNoSwap( float f ){
return f;
}
/*
================
Swap_Init
================
*/
void Swap_Init( void ){
byte swaptest[2] = {1,0};
// set the byte swapping variables in a portable manner
if ( *(short *)swaptest == 1 ) {
_BigShort = ShortSwap;
_LittleShort = ShortNoSwap;
_BigLong = LongSwap;
_LittleLong = LongNoSwap;
_BigFloat = FloatSwap;
_LittleFloat = FloatNoSwap;
}
else
{
_BigShort = ShortNoSwap;
_LittleShort = ShortSwap;
_BigLong = LongNoSwap;
_LittleLong = LongSwap;
_BigFloat = FloatNoSwap;
_LittleFloat = FloatSwap;
}
}
/*
===============
Com_ParseInfos
===============
*/
int Com_ParseInfos( const char *buf, int max, char infos[][MAX_INFO_STRING] ) {
const char *token;
int count;
char key[MAX_TOKEN_CHARS];
count = 0;
while ( 1 ) {
token = Com_Parse( &buf );
if ( !token[0] ) {
break;
}
if ( strcmp( token, "{" ) ) {
Com_Printf( "Missing { in info file\n" );
break;
}
if ( count == max ) {
Com_Printf( "Max infos exceeded\n" );
break;
}
infos[count][0] = 0;
while ( 1 ) {
token = Com_Parse( &buf );
if ( !token[0] ) {
Com_Printf( "Unexpected end of info file\n" );
break;
}
if ( !strcmp( token, "}" ) ) {
break;
}
Q_strncpyz( key, token, sizeof( key ) );
token = Com_ParseOnLine( &buf );
if ( !token[0] ) {
token = "<NULL>";
}
Info_SetValueForKey( infos[count], key, token );
}
count++;
}
return count;
}
/*
============================================================================
LIBRARY REPLACEMENT FUNCTIONS
============================================================================
*/
int Q_isprint( int c ){
if ( c >= 0x20 && c <= 0x7E ) {
return ( 1 );
}
return ( 0 );
}
int Q_islower( int c ){
if ( c >= 'a' && c <= 'z' ) {
return ( 1 );
}
return ( 0 );
}
int Q_isupper( int c ){
if ( c >= 'A' && c <= 'Z' ) {
return ( 1 );
}
return ( 0 );
}
int Q_isalpha( int c ){
if ( ( c >= 'a' && c <= 'z' ) || ( c >= 'A' && c <= 'Z' ) ) {
return ( 1 );
}
return ( 0 );
}
char* Q_strrchr( const char* string, int c ){
char cc = c;
char *s;
char *sp = (char *)0;
s = (char*)string;
while ( *s )
{
if ( *s == cc ) {
sp = s;
}
s++;
}
if ( cc == 0 ) {
sp = s;
}
return sp;
}
/*
=============
Q_strncpyz
Safe strncpy that ensures a trailing zero
=============
*/
void Q_strncpyz( char *dest, const char *src, std::size_t destsize ) {
if ( !src ) {
Com_Error( ERR_FATAL, "Q_strncpyz: NULL src" );
}
if ( destsize < 1 ) {
Com_Error( ERR_FATAL,"Q_strncpyz: destsize < 1" );
}
strncpy( dest, src, destsize - 1 );
dest[destsize - 1] = 0;
}
int Q_stricmpn( const char *s1, const char *s2, int n ) {
int c1, c2;
do {
c1 = *s1++;
c2 = *s2++;
if ( !n-- ) {
return 0; // strings are equal until end point
}
if ( c1 != c2 ) {
if ( c1 >= 'a' && c1 <= 'z' ) {
c1 -= ( 'a' - 'A' );
}
if ( c2 >= 'a' && c2 <= 'z' ) {
c2 -= ( 'a' - 'A' );
}
if ( c1 != c2 ) {
return c1 < c2 ? -1 : 1;
}
}
} while ( c1 );
return 0; // strings are equal
}
int Q_strncmp( const char *s1, const char *s2, int n ) {
int c1, c2;
do {
c1 = *s1++;
c2 = *s2++;
if ( !n-- ) {
return 0; // strings are equal until end point
}
if ( c1 != c2 ) {
return c1 < c2 ? -1 : 1;
}
} while ( c1 );
return 0; // strings are equal
}
int Q_stricmp( const char *s1, const char *s2 ) {
return Q_stricmpn( s1, s2, 99999 );
}
char *Q_strlwr( char *s1 ) {
char *s;
s = s1;
while ( *s ) {
*s = tolower( *s );
s++;
}
return s1;
}
char *Q_strupr( char *s1 ) {
char *s;
s = s1;
while ( *s ) {
*s = toupper( *s );
s++;
}
return s1;
}
// never goes past bounds or leaves without a terminating 0
void Q_strcat( char *dest, std::size_t size, const char *src ) {
auto l1 = strlen( dest );
if ( l1 >= size ) {
Com_Error( ERR_FATAL, "Q_strcat: already overflowed" );
}
Q_strncpyz( dest + l1, src, size - l1 );
}
int Q_PrintStrlen( const char *string ) {
int len;
const char *p;
if ( !string ) {
return 0;
}
len = 0;
p = string;
while ( *p ) {
if ( Q_IsColorString( p ) ) {
p += 2;
continue;
}
p++;
len++;
}
return len;
}
char *Q_CleanStr( char *string ) {
char* d;
char* s;
int c;
s = string;
d = string;
while ( ( c = *s ) != 0 ) {
if ( Q_IsColorString( s ) ) {
s++;
}
else if ( c >= 0x20 && c <= 0x7E ) {
*d++ = c;
}
s++;
}
*d = '\0';
return string;
}
void QDECL Com_sprintf( char *dest, std::size_t size, const char *fmt, ... ) {
va_list argptr;
char bigbuffer[32000]; // big, but small enough to fit in PPC stack
va_start( argptr,fmt );
int ret = vsprintf( bigbuffer,fmt,argptr );
va_end( argptr );
if ( ret < 0 ) {
Com_Error(ERR_FATAL, "Com_sprintf: vsprintf failed");
}
auto len = static_cast<size_t>(ret);
if ( len >= sizeof( bigbuffer ) ) {
Com_Error( ERR_FATAL, "Com_sprintf: overflowed bigbuffer" );
}
if ( len >= size ) {
Com_Printf( "Com_sprintf: overflow of %i in %i\n", len, size );
}
Q_strncpyz( dest, bigbuffer, size );
}
/*
============
va
does a varargs printf into a temp buffer, so I don't need to have
varargs versions of all text functions.
FIXME: make this buffer size safe someday
============
*/
char *QDECL va( const char *format, ... ) {
va_list argptr;
static char string[2][32000]; // in case va is called by nested functions
static int index = 0;
char *buf;
buf = string[index & 1];
index++;
va_start( argptr, format );
vsprintf( buf, format,argptr );
va_end( argptr );
return buf;
}
/*
=====================================================================
INFO STRINGS
=====================================================================
*/
/*
===============
Info_ValueForKey
Searches the string for the given
key and returns the associated value, or an empty string.
FIXME: overflow check?
===============
*/
const char *Info_ValueForKey( const char *s, const char *key ) {
char pkey[MAX_INFO_KEY];
static char value[2][MAX_INFO_VALUE]; // use two buffers so compares
// work without stomping on each other
static int valueindex = 0;
char *o;
if ( !s || !key ) {
return "";
}
if ( strlen( s ) >= MAX_INFO_STRING ) {
Com_Error( ERR_DROP, "Info_ValueForKey: oversize infostring" );
}
valueindex ^= 1;
if ( *s == '\\' ) {
s++;
}
while ( 1 )
{
o = pkey;
while ( *s != '\\' )
{
if ( !*s ) {
return "";
}
*o++ = *s++;
}
*o = 0;
s++;
o = value[valueindex];
while ( *s != '\\' && *s )
{
*o++ = *s++;
}
*o = 0;
if ( !Q_stricmp( key, pkey ) ) {
return value[valueindex];
}
if ( !*s ) {
break;
}
s++;
}
return "";
}
/*
===================
Info_NextPair
Used to itterate through all the key/value pairs in an info string
===================
*/
void Info_NextPair( const char *( *head ), char key[MAX_INFO_KEY], char value[MAX_INFO_VALUE] ) {
char *o;
const char *s;
s = *head;
if ( *s == '\\' ) {
s++;
}
key[0] = 0;
value[0] = 0;
o = key;
while ( *s != '\\' ) {
if ( !*s ) {
*o = 0;
*head = s;
return;
}
*o++ = *s++;
}
*o = 0;
s++;
o = value;
while ( *s != '\\' && *s ) {
*o++ = *s++;
}
*o = 0;
*head = s;
}
/*
===================
Info_RemoveKey
===================
*/
void Info_RemoveKey( char *s, const char *key ) {
char *start;
char pkey[MAX_INFO_KEY];
char value[MAX_INFO_VALUE];
char *o;
if ( strlen( s ) >= MAX_INFO_STRING ) {
Com_Error( ERR_DROP, "Info_RemoveKey: oversize infostring" );
}
if ( strchr( key, '\\' ) ) {
return;
}
while ( 1 )
{
start = s;
if ( *s == '\\' ) {
s++;
}
o = pkey;
while ( *s != '\\' )
{
if ( !*s ) {
return;
}
*o++ = *s++;
}
*o = 0;
s++;
o = value;
while ( *s != '\\' && *s )
{
if ( !*s ) {
return;
}
*o++ = *s++;
}
*o = 0;
if ( !strcmp( key, pkey ) ) {
strcpy( start, s ); // remove this part
return;
}
if ( !*s ) {
return;
}
}
}
/*
==================
Info_Validate
Some characters are illegal in info strings because they
can mess up the server's parsing
==================
*/
qboolean Info_Validate( const char *s ) {
if ( strchr( s, '\"' ) ) {
return qfalse;
}
if ( strchr( s, ';' ) ) {
return qfalse;
}
return qtrue;
}
/*
==================
Info_SetValueForKey
Changes or adds a key/value pair
==================
*/
void Info_SetValueForKey( char *s, const char *key, const char *value ) {
char newi[MAX_INFO_STRING];
if ( strlen( s ) >= MAX_INFO_STRING ) {
Com_Error( ERR_DROP, "Info_SetValueForKey: oversize infostring" );
}
if ( strchr( key, '\\' ) || strchr( value, '\\' ) ) {
Com_Printf( "Can't use keys or values with a \\\n" );
return;
}
if ( strchr( key, ';' ) || strchr( value, ';' ) ) {
Com_Printf( "Can't use keys or values with a semicolon\n" );
return;
}
if ( strchr( key, '\"' ) || strchr( value, '\"' ) ) {
Com_Printf( "Can't use keys or values with a \"\n" );
return;
}
Info_RemoveKey( s, key );
if ( !value || !strlen( value ) ) {
return;
}
Com_sprintf( newi, sizeof( newi ), "\\%s\\%s", key, value );
if ( strlen( newi ) + strlen( s ) > MAX_INFO_STRING ) {
Com_Printf( "Info string length exceeded\n" );
return;
}
strcat( s, newi );
}
//====================================================================
/*
===============
ParseHex
===============
*/
int ParseHex( const char *text ) {
int value;
int c;
value = 0;
while ( ( c = *text++ ) != 0 ) {
if ( c >= '0' && c <= '9' ) {
value = value * 16 + c - '0';
continue;
}
if ( c >= 'a' && c <= 'f' ) {
value = value * 16 + 10 + c - 'a';
continue;
}
if ( c >= 'A' && c <= 'F' ) {
value = value * 16 + 10 + c - 'A';
continue;
}
}
return value;
}