doom3-bfg/neo/idlib/Lexer.cpp
Daniel Gibson 9eeea7adf3 Fix tons of compiler warnings
mostly -Wreorder, use const char* instead of char* for "static strings",
fix inappropriate usage of NULL (e.g. instead of '\0' or (int)0)
2012-12-11 23:57:05 +01:00

2281 lines
42 KiB
C++

/*
===========================================================================
Doom 3 BFG Edition GPL Source Code
Copyright (C) 1993-2012 id Software LLC, a ZeniMax Media company.
This file is part of the Doom 3 BFG Edition GPL Source Code ("Doom 3 BFG Edition Source Code").
Doom 3 BFG Edition Source Code 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 3 of the License, or
(at your option) any later version.
Doom 3 BFG Edition Source Code 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 Doom 3 BFG Edition Source Code. If not, see <http://www.gnu.org/licenses/>.
In addition, the Doom 3 BFG Edition Source Code is also subject to certain additional terms. You should have received a copy of these additional terms immediately following the terms and conditions of the GNU General Public License which accompanied the Doom 3 BFG Edition Source Code. If not, please request a copy in writing from id Software at the address below.
If you have questions concerning this license or the applicable additional terms, you may contact in writing id Software LLC, c/o ZeniMax Media Inc., Suite 120, Rockville, Maryland 20850 USA.
===========================================================================
*/
#include "precompiled.h"
#pragma hdrstop
#define PUNCTABLE
//longer punctuations first
punctuation_t default_punctuations[] =
{
//binary operators
{">>=", P_RSHIFT_ASSIGN},
{"<<=", P_LSHIFT_ASSIGN},
//
{"...", P_PARMS},
//define merge operator
{"##", P_PRECOMPMERGE}, // pre-compiler
//logic operators
{"&&", P_LOGIC_AND}, // pre-compiler
{"||", P_LOGIC_OR}, // pre-compiler
{">=", P_LOGIC_GEQ}, // pre-compiler
{"<=", P_LOGIC_LEQ}, // pre-compiler
{"==", P_LOGIC_EQ}, // pre-compiler
{"!=", P_LOGIC_UNEQ}, // pre-compiler
//arithmatic operators
{"*=", P_MUL_ASSIGN},
{"/=", P_DIV_ASSIGN},
{"%=", P_MOD_ASSIGN},
{"+=", P_ADD_ASSIGN},
{"-=", P_SUB_ASSIGN},
{"++", P_INC},
{"--", P_DEC},
//binary operators
{"&=", P_BIN_AND_ASSIGN},
{"|=", P_BIN_OR_ASSIGN},
{"^=", P_BIN_XOR_ASSIGN},
{">>", P_RSHIFT}, // pre-compiler
{"<<", P_LSHIFT}, // pre-compiler
//reference operators
{"->", P_POINTERREF},
//C++
{"::", P_CPP1},
{".*", P_CPP2},
//arithmatic operators
{"*", P_MUL}, // pre-compiler
{"/", P_DIV}, // pre-compiler
{"%", P_MOD}, // pre-compiler
{"+", P_ADD}, // pre-compiler
{"-", P_SUB}, // pre-compiler
{"=", P_ASSIGN},
//binary operators
{"&", P_BIN_AND}, // pre-compiler
{"|", P_BIN_OR}, // pre-compiler
{"^", P_BIN_XOR}, // pre-compiler
{"~", P_BIN_NOT}, // pre-compiler
//logic operators
{"!", P_LOGIC_NOT}, // pre-compiler
{">", P_LOGIC_GREATER}, // pre-compiler
{"<", P_LOGIC_LESS}, // pre-compiler
//reference operator
{".", P_REF},
//seperators
{",", P_COMMA}, // pre-compiler
{";", P_SEMICOLON},
//label indication
{":", P_COLON}, // pre-compiler
//if statement
{"?", P_QUESTIONMARK}, // pre-compiler
//embracements
{"(", P_PARENTHESESOPEN}, // pre-compiler
{")", P_PARENTHESESCLOSE}, // pre-compiler
{"{", P_BRACEOPEN}, // pre-compiler
{"}", P_BRACECLOSE}, // pre-compiler
{"[", P_SQBRACKETOPEN},
{"]", P_SQBRACKETCLOSE},
//
{"\\", P_BACKSLASH},
//precompiler operator
{"#", P_PRECOMP}, // pre-compiler
{"$", P_DOLLAR},
{NULL, 0}
};
int default_punctuationtable[256];
int default_nextpunctuation[sizeof( default_punctuations ) / sizeof( punctuation_t )];
int default_setup;
char idLexer::baseFolder[ 256 ];
/*
================
idLexer::CreatePunctuationTable
================
*/
void idLexer::CreatePunctuationTable( const punctuation_t* punctuations )
{
int i, n, lastp;
const punctuation_t* p, *newp;
//get memory for the table
if( punctuations == default_punctuations )
{
idLexer::punctuationtable = default_punctuationtable;
idLexer::nextpunctuation = default_nextpunctuation;
if( default_setup )
{
return;
}
default_setup = true;
i = sizeof( default_punctuations ) / sizeof( punctuation_t );
}
else
{
if( !idLexer::punctuationtable || idLexer::punctuationtable == default_punctuationtable )
{
idLexer::punctuationtable = ( int* ) Mem_Alloc( 256 * sizeof( int ), TAG_IDLIB_LEXER );
}
if( idLexer::nextpunctuation && idLexer::nextpunctuation != default_nextpunctuation )
{
Mem_Free( idLexer::nextpunctuation );
}
for( i = 0; punctuations[i].p; i++ )
{
}
idLexer::nextpunctuation = ( int* ) Mem_Alloc( i * sizeof( int ), TAG_IDLIB_LEXER );
}
memset( idLexer::punctuationtable, 0xFF, 256 * sizeof( int ) );
memset( idLexer::nextpunctuation, 0xFF, i * sizeof( int ) );
//add the punctuations in the list to the punctuation table
for( i = 0; punctuations[i].p; i++ )
{
newp = &punctuations[i];
lastp = -1;
//sort the punctuations in this table entry on length (longer punctuations first)
for( n = idLexer::punctuationtable[( unsigned int ) newp->p[0]]; n >= 0; n = idLexer::nextpunctuation[n] )
{
p = &punctuations[n];
if( strlen( p->p ) < strlen( newp->p ) )
{
idLexer::nextpunctuation[i] = n;
if( lastp >= 0 )
{
idLexer::nextpunctuation[lastp] = i;
}
else
{
idLexer::punctuationtable[( unsigned int ) newp->p[0]] = i;
}
break;
}
lastp = n;
}
if( n < 0 )
{
idLexer::nextpunctuation[i] = -1;
if( lastp >= 0 )
{
idLexer::nextpunctuation[lastp] = i;
}
else
{
idLexer::punctuationtable[( unsigned int ) newp->p[0]] = i;
}
}
}
}
/*
================
idLexer::GetPunctuationFromId
================
*/
const char* idLexer::GetPunctuationFromId( int id )
{
int i;
for( i = 0; idLexer::punctuations[i].p; i++ )
{
if( idLexer::punctuations[i].n == id )
{
return idLexer::punctuations[i].p;
}
}
return "unkown punctuation";
}
/*
================
idLexer::GetPunctuationId
================
*/
int idLexer::GetPunctuationId( const char* p )
{
int i;
for( i = 0; idLexer::punctuations[i].p; i++ )
{
if( !strcmp( idLexer::punctuations[i].p, p ) )
{
return idLexer::punctuations[i].n;
}
}
return 0;
}
/*
================
idLexer::Error
================
*/
void idLexer::Error( const char* str, ... )
{
char text[MAX_STRING_CHARS];
va_list ap;
hadError = true;
if( idLexer::flags & LEXFL_NOERRORS )
{
return;
}
va_start( ap, str );
vsprintf( text, str, ap );
va_end( ap );
if( idLexer::flags & LEXFL_NOFATALERRORS )
{
idLib::common->Warning( "file %s, line %d: %s", idLexer::filename.c_str(), idLexer::line, text );
}
else
{
idLib::common->Error( "file %s, line %d: %s", idLexer::filename.c_str(), idLexer::line, text );
}
}
/*
================
idLexer::Warning
================
*/
void idLexer::Warning( const char* str, ... )
{
char text[MAX_STRING_CHARS];
va_list ap;
if( idLexer::flags & LEXFL_NOWARNINGS )
{
return;
}
va_start( ap, str );
vsprintf( text, str, ap );
va_end( ap );
idLib::common->Warning( "file %s, line %d: %s", idLexer::filename.c_str(), idLexer::line, text );
}
/*
================
idLexer::SetPunctuations
================
*/
void idLexer::SetPunctuations( const punctuation_t* p )
{
#ifdef PUNCTABLE
if( p )
{
idLexer::CreatePunctuationTable( p );
}
else
{
idLexer::CreatePunctuationTable( default_punctuations );
}
#endif //PUNCTABLE
if( p )
{
idLexer::punctuations = p;
}
else
{
idLexer::punctuations = default_punctuations;
}
}
/*
================
idLexer::ReadWhiteSpace
Reads spaces, tabs, C-like comments etc.
When a newline character is found the scripts line counter is increased.
================
*/
int idLexer::ReadWhiteSpace()
{
while( 1 )
{
// skip white space
while( *idLexer::script_p <= ' ' )
{
if( !*idLexer::script_p )
{
return 0;
}
if( *idLexer::script_p == '\n' )
{
idLexer::line++;
}
idLexer::script_p++;
}
// skip comments
if( *idLexer::script_p == '/' )
{
// comments //
if( *( idLexer::script_p + 1 ) == '/' )
{
idLexer::script_p++;
do
{
idLexer::script_p++;
if( !*idLexer::script_p )
{
return 0;
}
}
while( *idLexer::script_p != '\n' );
idLexer::line++;
idLexer::script_p++;
if( !*idLexer::script_p )
{
return 0;
}
continue;
}
// comments /* */
else if( *( idLexer::script_p + 1 ) == '*' )
{
idLexer::script_p++;
while( 1 )
{
idLexer::script_p++;
if( !*idLexer::script_p )
{
return 0;
}
if( *idLexer::script_p == '\n' )
{
idLexer::line++;
}
else if( *idLexer::script_p == '/' )
{
if( *( idLexer::script_p - 1 ) == '*' )
{
break;
}
if( *( idLexer::script_p + 1 ) == '*' )
{
idLexer::Warning( "nested comment" );
}
}
}
idLexer::script_p++;
if( !*idLexer::script_p )
{
return 0;
}
idLexer::script_p++;
if( !*idLexer::script_p )
{
return 0;
}
continue;
}
}
break;
}
return 1;
}
/*
========================
idLexer::SkipWhiteSpace
Reads spaces, tabs, C-like comments etc. When a newline character is found, the scripts line
counter is increased. Returns false if there is no token left to be read.
========================
*/
bool idLexer::SkipWhiteSpace( bool currentLine )
{
while( 1 )
{
assert( script_p <= end_p );
if( script_p == end_p )
{
return false;
}
// skip white space
while( *script_p <= ' ' )
{
if( script_p == end_p )
{
return false;
}
if( !*script_p )
{
return false;
}
if( *script_p == '\n' )
{
line++;
if( currentLine )
{
script_p++;
return true;
}
}
script_p++;
}
// skip comments
if( *script_p == '/' )
{
// comments //
if( *( script_p + 1 ) == '/' )
{
script_p++;
do
{
script_p++;
if( !*script_p )
{
return false;
}
}
while( *script_p != '\n' );
line++;
script_p++;
if( currentLine )
{
return true;
}
if( !*script_p )
{
return false;
}
continue;
}
// comments /* */
else if( *( script_p + 1 ) == '*' )
{
script_p++;
while( 1 )
{
script_p++;
if( !*script_p )
{
return false;
}
if( *script_p == '\n' )
{
line++;
}
else if( *script_p == '/' )
{
if( *( script_p - 1 ) == '*' )
{
break;
}
if( *( script_p + 1 ) == '*' )
{
Warning( "nested comment" );
}
}
}
script_p++;
if( !*script_p )
{
return false;
}
continue;
}
}
break;
}
return true;
}
/*
================
idLexer::ReadEscapeCharacter
================
*/
int idLexer::ReadEscapeCharacter( char* ch )
{
int c, val, i;
// step over the leading '\\'
idLexer::script_p++;
// determine the escape character
switch( *idLexer::script_p )
{
case '\\':
c = '\\';
break;
case 'n':
c = '\n';
break;
case 'r':
c = '\r';
break;
case 't':
c = '\t';
break;
case 'v':
c = '\v';
break;
case 'b':
c = '\b';
break;
case 'f':
c = '\f';
break;
case 'a':
c = '\a';
break;
case '\'':
c = '\'';
break;
case '\"':
c = '\"';
break;
case '\?':
c = '\?';
break;
case 'x':
{
idLexer::script_p++;
for( i = 0, val = 0; ; i++, idLexer::script_p++ )
{
c = *idLexer::script_p;
if( c >= '0' && c <= '9' )
c = c - '0';
else if( c >= 'A' && c <= 'Z' )
c = c - 'A' + 10;
else if( c >= 'a' && c <= 'z' )
c = c - 'a' + 10;
else
break;
val = ( val << 4 ) + c;
}
idLexer::script_p--;
if( val > 0xFF )
{
idLexer::Warning( "too large value in escape character" );
val = 0xFF;
}
c = val;
break;
}
default: //NOTE: decimal ASCII code, NOT octal
{
if( *idLexer::script_p < '0' || *idLexer::script_p > '9' )
{
idLexer::Error( "unknown escape char" );
}
for( i = 0, val = 0; ; i++, idLexer::script_p++ )
{
c = *idLexer::script_p;
if( c >= '0' && c <= '9' )
c = c - '0';
else
break;
val = val * 10 + c;
}
idLexer::script_p--;
if( val > 0xFF )
{
idLexer::Warning( "too large value in escape character" );
val = 0xFF;
}
c = val;
break;
}
}
// step over the escape character or the last digit of the number
idLexer::script_p++;
// store the escape character
*ch = c;
// succesfully read escape character
return 1;
}
/*
================
idLexer::ReadString
Escape characters are interpretted.
Reads two strings with only a white space between them as one string.
================
*/
int idLexer::ReadString( idToken* token, int quote )
{
int tmpline;
const char* tmpscript_p;
char ch;
if( quote == '\"' )
{
token->type = TT_STRING;
}
else
{
token->type = TT_LITERAL;
}
// leading quote
idLexer::script_p++;
while( 1 )
{
// if there is an escape character and escape characters are allowed
if( *idLexer::script_p == '\\' && !( idLexer::flags & LEXFL_NOSTRINGESCAPECHARS ) )
{
if( !idLexer::ReadEscapeCharacter( &ch ) )
{
return 0;
}
token->AppendDirty( ch );
}
// if a trailing quote
else if( *idLexer::script_p == quote )
{
// step over the quote
idLexer::script_p++;
// if consecutive strings should not be concatenated
if( ( idLexer::flags & LEXFL_NOSTRINGCONCAT ) &&
( !( idLexer::flags & LEXFL_ALLOWBACKSLASHSTRINGCONCAT ) || ( quote != '\"' ) ) )
{
break;
}
tmpscript_p = idLexer::script_p;
tmpline = idLexer::line;
// read white space between possible two consecutive strings
if( !idLexer::ReadWhiteSpace() )
{
idLexer::script_p = tmpscript_p;
idLexer::line = tmpline;
break;
}
if( idLexer::flags & LEXFL_NOSTRINGCONCAT )
{
if( *idLexer::script_p != '\\' )
{
idLexer::script_p = tmpscript_p;
idLexer::line = tmpline;
break;
}
// step over the '\\'
idLexer::script_p++;
if( !idLexer::ReadWhiteSpace() || ( *idLexer::script_p != quote ) )
{
idLexer::Error( "expecting string after '\' terminated line" );
return 0;
}
}
// if there's no leading qoute
if( *idLexer::script_p != quote )
{
idLexer::script_p = tmpscript_p;
idLexer::line = tmpline;
break;
}
// step over the new leading quote
idLexer::script_p++;
}
else
{
if( *idLexer::script_p == '\0' )
{
idLexer::Error( "missing trailing quote" );
return 0;
}
if( *idLexer::script_p == '\n' )
{
idLexer::Error( "newline inside string" );
return 0;
}
token->AppendDirty( *idLexer::script_p++ );
}
}
token->data[token->len] = '\0';
if( token->type == TT_LITERAL )
{
if( !( idLexer::flags & LEXFL_ALLOWMULTICHARLITERALS ) )
{
if( token->Length() != 1 )
{
idLexer::Warning( "literal is not one character long" );
}
}
token->subtype = ( *token )[0];
}
else
{
// the sub type is the length of the string
token->subtype = token->Length();
}
return 1;
}
/*
================
idLexer::ReadName
================
*/
int idLexer::ReadName( idToken* token )
{
char c;
token->type = TT_NAME;
do
{
token->AppendDirty( *idLexer::script_p++ );
c = *idLexer::script_p;
}
while( ( c >= 'a' && c <= 'z' ) ||
( c >= 'A' && c <= 'Z' ) ||
( c >= '0' && c <= '9' ) ||
c == '_' ||
// if treating all tokens as strings, don't parse '-' as a seperate token
( ( idLexer::flags & LEXFL_ONLYSTRINGS ) && ( c == '-' ) ) ||
// if special path name characters are allowed
( ( idLexer::flags & LEXFL_ALLOWPATHNAMES ) && ( c == '/' || c == '\\' || c == ':' || c == '.' ) ) );
token->data[token->len] = '\0';
//the sub type is the length of the name
token->subtype = token->Length();
return 1;
}
/*
================
idLexer::CheckString
================
*/
ID_INLINE int idLexer::CheckString( const char* str ) const
{
int i;
for( i = 0; str[i]; i++ )
{
if( idLexer::script_p[i] != str[i] )
{
return false;
}
}
return true;
}
/*
================
idLexer::ReadNumber
================
*/
int idLexer::ReadNumber( idToken* token )
{
int i;
int dot;
char c, c2;
token->type = TT_NUMBER;
token->subtype = 0;
token->intvalue = 0;
token->floatvalue = 0;
c = *idLexer::script_p;
c2 = *( idLexer::script_p + 1 );
if( c == '0' && c2 != '.' )
{
// check for a hexadecimal number
if( c2 == 'x' || c2 == 'X' )
{
token->AppendDirty( *idLexer::script_p++ );
token->AppendDirty( *idLexer::script_p++ );
c = *idLexer::script_p;
while( ( c >= '0' && c <= '9' ) ||
( c >= 'a' && c <= 'f' ) ||
( c >= 'A' && c <= 'F' ) )
{
token->AppendDirty( c );
c = *( ++idLexer::script_p );
}
token->subtype = TT_HEX | TT_INTEGER;
}
// check for a binary number
else if( c2 == 'b' || c2 == 'B' )
{
token->AppendDirty( *idLexer::script_p++ );
token->AppendDirty( *idLexer::script_p++ );
c = *idLexer::script_p;
while( c == '0' || c == '1' )
{
token->AppendDirty( c );
c = *( ++idLexer::script_p );
}
token->subtype = TT_BINARY | TT_INTEGER;
}
// its an octal number
else
{
token->AppendDirty( *idLexer::script_p++ );
c = *idLexer::script_p;
while( c >= '0' && c <= '7' )
{
token->AppendDirty( c );
c = *( ++idLexer::script_p );
}
token->subtype = TT_OCTAL | TT_INTEGER;
}
}
else
{
// decimal integer or floating point number or ip address
dot = 0;
while( 1 )
{
if( c >= '0' && c <= '9' )
{
}
else if( c == '.' )
{
dot++;
}
else
{
break;
}
token->AppendDirty( c );
c = *( ++idLexer::script_p );
}
if( c == 'e' && dot == 0 )
{
//We have scientific notation without a decimal point
dot++;
}
// if a floating point number
if( dot == 1 )
{
token->subtype = TT_DECIMAL | TT_FLOAT;
// check for floating point exponent
if( c == 'e' )
{
//Append the e so that GetFloatValue code works
token->AppendDirty( c );
c = *( ++idLexer::script_p );
if( c == '-' )
{
token->AppendDirty( c );
c = *( ++idLexer::script_p );
}
else if( c == '+' )
{
token->AppendDirty( c );
c = *( ++idLexer::script_p );
}
while( c >= '0' && c <= '9' )
{
token->AppendDirty( c );
c = *( ++idLexer::script_p );
}
}
// check for floating point exception infinite 1.#INF or indefinite 1.#IND or NaN
else if( c == '#' )
{
c2 = 4;
if( CheckString( "INF" ) )
{
token->subtype |= TT_INFINITE;
}
else if( CheckString( "IND" ) )
{
token->subtype |= TT_INDEFINITE;
}
else if( CheckString( "NAN" ) )
{
token->subtype |= TT_NAN;
}
else if( CheckString( "QNAN" ) )
{
token->subtype |= TT_NAN;
c2++;
}
else if( CheckString( "SNAN" ) )
{
token->subtype |= TT_NAN;
c2++;
}
for( i = 0; i < c2; i++ )
{
token->AppendDirty( c );
c = *( ++idLexer::script_p );
}
while( c >= '0' && c <= '9' )
{
token->AppendDirty( c );
c = *( ++idLexer::script_p );
}
if( !( idLexer::flags & LEXFL_ALLOWFLOATEXCEPTIONS ) )
{
token->AppendDirty( 0 ); // zero terminate for c_str
idLexer::Error( "parsed %s", token->c_str() );
}
}
}
else if( dot > 1 )
{
if( !( idLexer::flags & LEXFL_ALLOWIPADDRESSES ) )
{
idLexer::Error( "more than one dot in number" );
return 0;
}
if( dot != 3 )
{
idLexer::Error( "ip address should have three dots" );
return 0;
}
token->subtype = TT_IPADDRESS;
}
else
{
token->subtype = TT_DECIMAL | TT_INTEGER;
}
}
if( token->subtype & TT_FLOAT )
{
if( c > ' ' )
{
// single-precision: float
if( c == 'f' || c == 'F' )
{
token->subtype |= TT_SINGLE_PRECISION;
idLexer::script_p++;
}
// extended-precision: long double
else if( c == 'l' || c == 'L' )
{
token->subtype |= TT_EXTENDED_PRECISION;
idLexer::script_p++;
}
// default is double-precision: double
else
{
token->subtype |= TT_DOUBLE_PRECISION;
}
}
else
{
token->subtype |= TT_DOUBLE_PRECISION;
}
}
else if( token->subtype & TT_INTEGER )
{
if( c > ' ' )
{
// default: signed long
for( i = 0; i < 2; i++ )
{
// long integer
if( c == 'l' || c == 'L' )
{
token->subtype |= TT_LONG;
}
// unsigned integer
else if( c == 'u' || c == 'U' )
{
token->subtype |= TT_UNSIGNED;
}
else
{
break;
}
c = *( ++idLexer::script_p );
}
}
}
else if( token->subtype & TT_IPADDRESS )
{
if( c == ':' )
{
token->AppendDirty( c );
c = *( ++idLexer::script_p );
while( c >= '0' && c <= '9' )
{
token->AppendDirty( c );
c = *( ++idLexer::script_p );
}
token->subtype |= TT_IPPORT;
}
}
token->data[token->len] = '\0';
return 1;
}
/*
================
idLexer::ReadPunctuation
================
*/
int idLexer::ReadPunctuation( idToken* token )
{
int l, n, i;
const char* p;
const punctuation_t* punc;
#ifdef PUNCTABLE
for( n = idLexer::punctuationtable[( unsigned int ) * ( idLexer::script_p )]; n >= 0; n = idLexer::nextpunctuation[n] )
{
punc = &( idLexer::punctuations[n] );
#else
int i;
for( i = 0; idLexer::punctuations[i].p; i++ )
{
punc = &idLexer::punctuations[i];
#endif
p = punc->p;
// check for this punctuation in the script
for( l = 0; p[l] && idLexer::script_p[l]; l++ )
{
if( idLexer::script_p[l] != p[l] )
{
break;
}
}
if( !p[l] )
{
//
token->EnsureAlloced( l + 1, false );
for( i = 0; i <= l; i++ )
{
token->data[i] = p[i];
}
token->len = l;
//
idLexer::script_p += l;
token->type = TT_PUNCTUATION;
// sub type is the punctuation id
token->subtype = punc->n;
return 1;
}
}
return 0;
}
/*
================
idLexer::ReadToken
================
*/
int idLexer::ReadToken( idToken* token )
{
int c;
if( !loaded )
{
idLib::common->Error( "idLexer::ReadToken: no file loaded" );
return 0;
}
if( script_p == NULL )
{
return 0;
}
// if there is a token available (from unreadToken)
if( tokenavailable )
{
tokenavailable = 0;
*token = idLexer::token;
return 1;
}
// save script pointer
lastScript_p = script_p;
// save line counter
lastline = line;
// clear the token stuff
token->data[0] = '\0';
token->len = 0;
// start of the white space
whiteSpaceStart_p = script_p;
token->whiteSpaceStart_p = script_p;
// read white space before token
if( !ReadWhiteSpace() )
{
return 0;
}
// end of the white space
idLexer::whiteSpaceEnd_p = script_p;
token->whiteSpaceEnd_p = script_p;
// line the token is on
token->line = line;
// number of lines crossed before token
token->linesCrossed = line - lastline;
// clear token flags
token->flags = 0;
c = *idLexer::script_p;
// if we're keeping everything as whitespace deliminated strings
if( idLexer::flags & LEXFL_ONLYSTRINGS )
{
// if there is a leading quote
if( c == '\"' || c == '\'' )
{
if( !idLexer::ReadString( token, c ) )
{
return 0;
}
}
else if( !idLexer::ReadName( token ) )
{
return 0;
}
}
// if there is a number
else if( ( c >= '0' && c <= '9' ) ||
( c == '.' && ( *( idLexer::script_p + 1 ) >= '0' && *( idLexer::script_p + 1 ) <= '9' ) ) )
{
if( !idLexer::ReadNumber( token ) )
{
return 0;
}
// if names are allowed to start with a number
if( idLexer::flags & LEXFL_ALLOWNUMBERNAMES )
{
c = *idLexer::script_p;
if( ( c >= 'a' && c <= 'z' ) || ( c >= 'A' && c <= 'Z' ) || c == '_' )
{
if( !idLexer::ReadName( token ) )
{
return 0;
}
}
}
}
// if there is a leading quote
else if( c == '\"' || c == '\'' )
{
if( !idLexer::ReadString( token, c ) )
{
return 0;
}
}
// if there is a name
else if( ( c >= 'a' && c <= 'z' ) || ( c >= 'A' && c <= 'Z' ) || c == '_' )
{
if( !idLexer::ReadName( token ) )
{
return 0;
}
}
// names may also start with a slash when pathnames are allowed
else if( ( idLexer::flags & LEXFL_ALLOWPATHNAMES ) && ( ( c == '/' || c == '\\' ) || c == '.' ) )
{
if( !idLexer::ReadName( token ) )
{
return 0;
}
}
// check for punctuations
else if( !idLexer::ReadPunctuation( token ) )
{
idLexer::Error( "unknown punctuation %c", c );
return 0;
}
// succesfully read a token
return 1;
}
/*
================
idLexer::ExpectTokenString
================
*/
int idLexer::ExpectTokenString( const char* string )
{
idToken token;
if( !idLexer::ReadToken( &token ) )
{
idLexer::Error( "couldn't find expected '%s'", string );
return 0;
}
if( token != string )
{
idLexer::Error( "expected '%s' but found '%s'", string, token.c_str() );
return 0;
}
return 1;
}
/*
================
idLexer::ExpectTokenType
================
*/
int idLexer::ExpectTokenType( int type, int subtype, idToken* token )
{
idStr str;
if( !idLexer::ReadToken( token ) )
{
idLexer::Error( "couldn't read expected token" );
return 0;
}
if( token->type != type )
{
switch( type )
{
case TT_STRING:
str = "string";
break;
case TT_LITERAL:
str = "literal";
break;
case TT_NUMBER:
str = "number";
break;
case TT_NAME:
str = "name";
break;
case TT_PUNCTUATION:
str = "punctuation";
break;
default:
str = "unknown type";
break;
}
idLexer::Error( "expected a %s but found '%s'", str.c_str(), token->c_str() );
return 0;
}
if( token->type == TT_NUMBER )
{
if( ( token->subtype & subtype ) != subtype )
{
str.Clear();
if( subtype & TT_DECIMAL ) str = "decimal ";
if( subtype & TT_HEX ) str = "hex ";
if( subtype & TT_OCTAL ) str = "octal ";
if( subtype & TT_BINARY ) str = "binary ";
if( subtype & TT_UNSIGNED ) str += "unsigned ";
if( subtype & TT_LONG ) str += "long ";
if( subtype & TT_FLOAT ) str += "float ";
if( subtype & TT_INTEGER ) str += "integer ";
str.StripTrailing( ' ' );
idLexer::Error( "expected %s but found '%s'", str.c_str(), token->c_str() );
return 0;
}
}
else if( token->type == TT_PUNCTUATION )
{
if( subtype < 0 )
{
idLexer::Error( "BUG: wrong punctuation subtype" );
return 0;
}
if( token->subtype != subtype )
{
idLexer::Error( "expected '%s' but found '%s'", GetPunctuationFromId( subtype ), token->c_str() );
return 0;
}
}
return 1;
}
/*
================
idLexer::ExpectAnyToken
================
*/
int idLexer::ExpectAnyToken( idToken* token )
{
if( !idLexer::ReadToken( token ) )
{
idLexer::Error( "couldn't read expected token" );
return 0;
}
else
{
return 1;
}
}
/*
================
idLexer::CheckTokenString
================
*/
int idLexer::CheckTokenString( const char* string )
{
idToken tok;
if( !ReadToken( &tok ) )
{
return 0;
}
// if the given string is available
if( tok == string )
{
return 1;
}
// unread token
script_p = lastScript_p;
line = lastline;
return 0;
}
/*
================
idLexer::CheckTokenType
================
*/
int idLexer::CheckTokenType( int type, int subtype, idToken* token )
{
idToken tok;
if( !ReadToken( &tok ) )
{
return 0;
}
// if the type matches
if( tok.type == type && ( tok.subtype & subtype ) == subtype )
{
*token = tok;
return 1;
}
// unread token
script_p = lastScript_p;
line = lastline;
return 0;
}
/*
================
idLexer::PeekTokenString
================
*/
int idLexer::PeekTokenString( const char* string )
{
idToken tok;
if( !ReadToken( &tok ) )
{
return 0;
}
// unread token
script_p = lastScript_p;
line = lastline;
// if the given string is available
if( tok == string )
{
return 1;
}
return 0;
}
/*
================
idLexer::PeekTokenType
================
*/
int idLexer::PeekTokenType( int type, int subtype, idToken* token )
{
idToken tok;
if( !ReadToken( &tok ) )
{
return 0;
}
// unread token
script_p = lastScript_p;
line = lastline;
// if the type matches
if( tok.type == type && ( tok.subtype & subtype ) == subtype )
{
*token = tok;
return 1;
}
return 0;
}
/*
================
idLexer::SkipUntilString
================
*/
int idLexer::SkipUntilString( const char* string )
{
idToken token;
while( idLexer::ReadToken( &token ) )
{
if( token == string )
{
return 1;
}
}
return 0;
}
/*
================
idLexer::SkipRestOfLine
================
*/
int idLexer::SkipRestOfLine()
{
idToken token;
while( idLexer::ReadToken( &token ) )
{
if( token.linesCrossed )
{
idLexer::script_p = lastScript_p;
idLexer::line = lastline;
return 1;
}
}
return 0;
}
/*
=================
idLexer::SkipBracedSection
Skips until a matching close brace is found.
Internal brace depths are properly skipped.
=================
*/
int idLexer::SkipBracedSection( bool parseFirstBrace )
{
idToken token;
int depth;
depth = parseFirstBrace ? 0 : 1;
do
{
if( !ReadToken( &token ) )
{
return false;
}
if( token.type == TT_PUNCTUATION )
{
if( token == "{" )
{
depth++;
}
else if( token == "}" )
{
depth--;
}
}
}
while( depth );
return true;
}
/*
================
idLexer::UnreadToken
================
*/
void idLexer::UnreadToken( const idToken* token )
{
if( idLexer::tokenavailable )
{
idLib::common->FatalError( "idLexer::unreadToken, unread token twice\n" );
}
idLexer::token = *token;
idLexer::tokenavailable = 1;
}
/*
================
idLexer::ReadTokenOnLine
================
*/
int idLexer::ReadTokenOnLine( idToken* token )
{
idToken tok;
if( !idLexer::ReadToken( &tok ) )
{
idLexer::script_p = lastScript_p;
idLexer::line = lastline;
return false;
}
// if no lines were crossed before this token
if( !tok.linesCrossed )
{
*token = tok;
return true;
}
// restore our position
idLexer::script_p = lastScript_p;
idLexer::line = lastline;
token->Clear();
return false;
}
/*
================
idLexer::ReadRestOfLine
================
*/
const char* idLexer::ReadRestOfLine( idStr& out )
{
while( 1 )
{
if( *idLexer::script_p == '\n' )
{
idLexer::line++;
break;
}
if( !*idLexer::script_p )
{
break;
}
if( *idLexer::script_p <= ' ' )
{
out += " ";
}
else
{
out += *idLexer::script_p;
}
idLexer::script_p++;
}
out.Strip( ' ' );
return out.c_str();
}
/*
================
idLexer::ParseInt
================
*/
int idLexer::ParseInt()
{
idToken token;
if( !idLexer::ReadToken( &token ) )
{
idLexer::Error( "couldn't read expected integer" );
return 0;
}
if( token.type == TT_PUNCTUATION && token == "-" )
{
idLexer::ExpectTokenType( TT_NUMBER, TT_INTEGER, &token );
return -( ( signed int ) token.GetIntValue() );
}
else if( token.type != TT_NUMBER || token.subtype == TT_FLOAT )
{
idLexer::Error( "expected integer value, found '%s'", token.c_str() );
}
return token.GetIntValue();
}
/*
================
idLexer::ParseBool
================
*/
bool idLexer::ParseBool()
{
idToken token;
if( !idLexer::ExpectTokenType( TT_NUMBER, 0, &token ) )
{
idLexer::Error( "couldn't read expected boolean" );
return false;
}
return ( token.GetIntValue() != 0 );
}
/*
================
idLexer::ParseFloat
================
*/
float idLexer::ParseFloat( bool* errorFlag )
{
idToken token;
if( errorFlag )
{
*errorFlag = false;
}
if( !idLexer::ReadToken( &token ) )
{
if( errorFlag )
{
idLexer::Warning( "couldn't read expected floating point number" );
*errorFlag = true;
}
else
{
idLexer::Error( "couldn't read expected floating point number" );
}
return 0;
}
if( token.type == TT_PUNCTUATION && token == "-" )
{
idLexer::ExpectTokenType( TT_NUMBER, 0, &token );
return -token.GetFloatValue();
}
else if( token.type != TT_NUMBER )
{
if( errorFlag )
{
idLexer::Warning( "expected float value, found '%s'", token.c_str() );
*errorFlag = true;
}
else
{
idLexer::Error( "expected float value, found '%s'", token.c_str() );
}
}
return token.GetFloatValue();
}
/*
================
idLexer::Parse1DMatrix
================
*/
int idLexer::Parse1DMatrix( int x, float* m )
{
int i;
if( !idLexer::ExpectTokenString( "(" ) )
{
return false;
}
for( i = 0; i < x; i++ )
{
m[i] = idLexer::ParseFloat();
}
if( !idLexer::ExpectTokenString( ")" ) )
{
return false;
}
return true;
}
/*
================
idLexer::Parse2DMatrix
================
*/
int idLexer::Parse2DMatrix( int y, int x, float* m )
{
int i;
if( !idLexer::ExpectTokenString( "(" ) )
{
return false;
}
for( i = 0; i < y; i++ )
{
if( !idLexer::Parse1DMatrix( x, m + i * x ) )
{
return false;
}
}
if( !idLexer::ExpectTokenString( ")" ) )
{
return false;
}
return true;
}
/*
================
idLexer::Parse3DMatrix
================
*/
int idLexer::Parse3DMatrix( int z, int y, int x, float* m )
{
int i;
if( !idLexer::ExpectTokenString( "(" ) )
{
return false;
}
for( i = 0 ; i < z; i++ )
{
if( !idLexer::Parse2DMatrix( y, x, m + i * x * y ) )
{
return false;
}
}
if( !idLexer::ExpectTokenString( ")" ) )
{
return false;
}
return true;
}
/*
=================
idParser::ParseBracedSection
The next token should be an open brace.
Parses until a matching close brace is found.
Maintains exact characters between braces.
FIXME: this should use ReadToken and replace the token white space with correct indents and newlines
=================
*/
const char* idLexer::ParseBracedSectionExact( idStr& out, int tabs )
{
int depth;
bool doTabs;
bool skipWhite;
out.Empty();
if( !idLexer::ExpectTokenString( "{" ) )
{
return out.c_str( );
}
out = "{";
depth = 1;
skipWhite = false;
doTabs = tabs >= 0;
while( depth && *idLexer::script_p )
{
char c = *( idLexer::script_p++ );
switch( c )
{
case '\t':
case ' ':
{
if( skipWhite )
{
continue;
}
break;
}
case '\n':
{
if( doTabs )
{
skipWhite = true;
out += c;
continue;
}
break;
}
case '{':
{
depth++;
tabs++;
break;
}
case '}':
{
depth--;
tabs--;
break;
}
}
if( skipWhite )
{
int i = tabs;
if( c == '{' )
{
i--;
}
skipWhite = false;
for( ; i > 0; i-- )
{
out += '\t';
}
}
out += c;
}
return out.c_str();
}
/*
=================
idLexer::ParseBracedSection
The next token should be an open brace.
Parses until a matching close brace is found.
Internal brace depths are properly skipped.
=================
*/
const char* idLexer::ParseBracedSection( idStr& out )
{
idToken token;
int i, depth;
out.Empty();
if( !idLexer::ExpectTokenString( "{" ) )
{
return out.c_str();
}
out = "{";
depth = 1;
do
{
if( !idLexer::ReadToken( &token ) )
{
Error( "missing closing brace" );
return out.c_str();
}
// if the token is on a new line
for( i = 0; i < token.linesCrossed; i++ )
{
out += "\r\n";
}
if( token.type == TT_PUNCTUATION )
{
if( token[0] == '{' )
{
depth++;
}
else if( token[0] == '}' )
{
depth--;
}
}
if( token.type == TT_STRING )
{
out += "\"" + token + "\"";
}
else
{
out += token;
}
out += " ";
}
while( depth );
return out.c_str();
}
/*
=================
idLexer::ParseRestOfLine
parse the rest of the line
=================
*/
const char* idLexer::ParseRestOfLine( idStr& out )
{
idToken token;
out.Empty();
while( idLexer::ReadToken( &token ) )
{
if( token.linesCrossed )
{
idLexer::script_p = lastScript_p;
idLexer::line = lastline;
break;
}
if( out.Length() )
{
out += " ";
}
out += token;
}
return out.c_str();
}
/*
========================
idLexer::ParseCompleteLine
Returns a string up to the \n, but doesn't eat any whitespace at the beginning of the next line.
========================
*/
const char* idLexer::ParseCompleteLine( idStr& out )
{
idToken token;
const char* start;
start = script_p;
while( 1 )
{
// end of buffer
if( *script_p == 0 )
{
break;
}
if( *script_p == '\n' )
{
line++;
script_p++;
break;
}
script_p++;
}
out.Empty();
out.Append( start, script_p - start );
return out.c_str();
}
/*
================
idLexer::GetLastWhiteSpace
================
*/
int idLexer::GetLastWhiteSpace( idStr& whiteSpace ) const
{
whiteSpace.Clear();
for( const char* p = whiteSpaceStart_p; p < whiteSpaceEnd_p; p++ )
{
whiteSpace.Append( *p );
}
return whiteSpace.Length();
}
/*
================
idLexer::GetLastWhiteSpaceStart
================
*/
int idLexer::GetLastWhiteSpaceStart() const
{
return whiteSpaceStart_p - buffer;
}
/*
================
idLexer::GetLastWhiteSpaceEnd
================
*/
int idLexer::GetLastWhiteSpaceEnd() const
{
return whiteSpaceEnd_p - buffer;
}
/*
================
idLexer::Reset
================
*/
void idLexer::Reset()
{
// pointer in script buffer
idLexer::script_p = idLexer::buffer;
// pointer in script buffer before reading token
idLexer::lastScript_p = idLexer::buffer;
// begin of white space
idLexer::whiteSpaceStart_p = NULL;
// end of white space
idLexer::whiteSpaceEnd_p = NULL;
// set if there's a token available in idLexer::token
idLexer::tokenavailable = 0;
idLexer::line = 1;
idLexer::lastline = 1;
// clear the saved token
idLexer::token = "";
}
/*
================
idLexer::EndOfFile
================
*/
bool idLexer::EndOfFile()
{
return idLexer::script_p >= idLexer::end_p;
}
/*
================
idLexer::NumLinesCrossed
================
*/
int idLexer::NumLinesCrossed()
{
return idLexer::line - idLexer::lastline;
}
/*
================
idLexer::LoadFile
================
*/
int idLexer::LoadFile( const char* filename, bool OSPath )
{
idFile* fp;
idStr pathname;
int length;
char* buf;
if( idLexer::loaded )
{
idLib::common->Error( "idLexer::LoadFile: another script already loaded" );
return false;
}
if( !OSPath && ( baseFolder[0] != '\0' ) )
{
pathname = va( "%s/%s", baseFolder, filename );
}
else
{
pathname = filename;
}
if( OSPath )
{
fp = idLib::fileSystem->OpenExplicitFileRead( pathname );
}
else
{
fp = idLib::fileSystem->OpenFileRead( pathname );
}
if( !fp )
{
return false;
}
length = fp->Length();
buf = ( char* ) Mem_Alloc( length + 1, TAG_IDLIB_LEXER );
buf[length] = '\0';
fp->Read( buf, length );
idLexer::fileTime = fp->Timestamp();
idLexer::filename = fp->GetFullPath();
idLib::fileSystem->CloseFile( fp );
idLexer::buffer = buf;
idLexer::length = length;
// pointer in script buffer
idLexer::script_p = idLexer::buffer;
// pointer in script buffer before reading token
idLexer::lastScript_p = idLexer::buffer;
// pointer to end of script buffer
idLexer::end_p = &( idLexer::buffer[length] );
idLexer::tokenavailable = 0;
idLexer::line = 1;
idLexer::lastline = 1;
idLexer::allocated = true;
idLexer::loaded = true;
return true;
}
/*
================
idLexer::LoadMemory
================
*/
int idLexer::LoadMemory( const char* ptr, int length, const char* name, int startLine )
{
if( idLexer::loaded )
{
idLib::common->Error( "idLexer::LoadMemory: another script already loaded" );
return false;
}
idLexer::filename = name;
idLexer::buffer = ptr;
idLexer::fileTime = 0;
idLexer::length = length;
// pointer in script buffer
idLexer::script_p = idLexer::buffer;
// pointer in script buffer before reading token
idLexer::lastScript_p = idLexer::buffer;
// pointer to end of script buffer
idLexer::end_p = &( idLexer::buffer[length] );
idLexer::tokenavailable = 0;
idLexer::line = startLine;
idLexer::lastline = startLine;
idLexer::allocated = false;
idLexer::loaded = true;
return true;
}
/*
================
idLexer::FreeSource
================
*/
void idLexer::FreeSource()
{
#ifdef PUNCTABLE
if( idLexer::punctuationtable && idLexer::punctuationtable != default_punctuationtable )
{
Mem_Free( ( void* ) idLexer::punctuationtable );
idLexer::punctuationtable = NULL;
}
if( idLexer::nextpunctuation && idLexer::nextpunctuation != default_nextpunctuation )
{
Mem_Free( ( void* ) idLexer::nextpunctuation );
idLexer::nextpunctuation = NULL;
}
#endif //PUNCTABLE
if( idLexer::allocated )
{
Mem_Free( ( void* ) idLexer::buffer );
idLexer::buffer = NULL;
idLexer::allocated = false;
}
idLexer::tokenavailable = 0;
idLexer::token = "";
idLexer::loaded = false;
}
/*
================
idLexer::idLexer
================
*/
idLexer::idLexer()
{
idLexer::loaded = false;
idLexer::filename = "";
idLexer::flags = 0;
idLexer::SetPunctuations( NULL );
idLexer::allocated = false;
idLexer::fileTime = 0;
idLexer::length = 0;
idLexer::line = 0;
idLexer::lastline = 0;
idLexer::tokenavailable = 0;
idLexer::token = "";
idLexer::next = NULL;
idLexer::hadError = false;
}
/*
================
idLexer::idLexer
================
*/
idLexer::idLexer( int flags )
{
idLexer::loaded = false;
idLexer::filename = "";
idLexer::flags = flags;
idLexer::SetPunctuations( NULL );
idLexer::allocated = false;
idLexer::fileTime = 0;
idLexer::length = 0;
idLexer::line = 0;
idLexer::lastline = 0;
idLexer::tokenavailable = 0;
idLexer::token = "";
idLexer::next = NULL;
idLexer::hadError = false;
}
/*
================
idLexer::idLexer
================
*/
idLexer::idLexer( const char* filename, int flags, bool OSPath )
{
idLexer::loaded = false;
idLexer::flags = flags;
idLexer::SetPunctuations( NULL );
idLexer::allocated = false;
idLexer::token = "";
idLexer::next = NULL;
idLexer::hadError = false;
idLexer::LoadFile( filename, OSPath );
}
/*
================
idLexer::idLexer
================
*/
idLexer::idLexer( const char* ptr, int length, const char* name, int flags )
{
idLexer::loaded = false;
idLexer::flags = flags;
idLexer::SetPunctuations( NULL );
idLexer::allocated = false;
idLexer::token = "";
idLexer::next = NULL;
idLexer::hadError = false;
idLexer::LoadMemory( ptr, length, name );
}
/*
================
idLexer::~idLexer
================
*/
idLexer::~idLexer()
{
idLexer::FreeSource();
}
/*
================
idLexer::SetBaseFolder
================
*/
void idLexer::SetBaseFolder( const char* path )
{
idStr::Copynz( baseFolder, path, sizeof( baseFolder ) );
}
/*
================
idLexer::HadError
================
*/
bool idLexer::HadError() const
{
return hadError;
}