562 lines
14 KiB
C++
562 lines
14 KiB
C++
/*
|
||
* This file is part of the Colobot: Gold Edition source code
|
||
* Copyright (C) 2001-2014, Daniel Roux, EPSITEC SA & TerranovaTeam
|
||
* http://epsiteс.ch; http://colobot.info; http://github.com/colobot
|
||
*
|
||
* This program 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.
|
||
*
|
||
* This program 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 this program. If not, see http://gnu.org/licenses
|
||
*/
|
||
|
||
|
||
//////////////////////////////////////////////////////////////////
|
||
// Managing Tokens
|
||
// the text of a program is first transformed
|
||
// into a sequence of tokens for easy interpretation
|
||
// it will only treat the case as an error
|
||
// where there is an illegal character in a string
|
||
|
||
|
||
#include "CBot.h"
|
||
#include <cstdarg>
|
||
|
||
CBotStringArray CBotToken::m_ListKeyWords;
|
||
int CBotToken::m_ListIdKeyWords[200];
|
||
CBotStringArray CBotToken::m_ListKeyDefine;
|
||
long CBotToken::m_ListKeyNums[MAXDEFNUM];
|
||
|
||
//! contructors
|
||
CBotToken::CBotToken()
|
||
{
|
||
m_next = NULL;
|
||
m_prev = NULL;
|
||
m_type = TokenTypVar; // at the beginning a default variable type
|
||
m_IdKeyWord = -1;
|
||
}
|
||
|
||
CBotToken::CBotToken(const CBotToken* pSrc)
|
||
{
|
||
m_next = NULL;
|
||
m_prev = NULL;
|
||
|
||
m_Text.Empty();
|
||
m_Sep.Empty();
|
||
|
||
m_type = 0;
|
||
m_IdKeyWord = 0;
|
||
|
||
m_start = 0;
|
||
m_end = 0;
|
||
|
||
if ( pSrc != NULL )
|
||
{
|
||
|
||
m_type = pSrc->m_type;
|
||
m_IdKeyWord = pSrc->m_IdKeyWord;
|
||
|
||
m_Text = pSrc->m_Text;
|
||
m_Sep = pSrc->m_Sep;
|
||
|
||
m_start = pSrc->m_start;
|
||
m_end = pSrc->m_end;
|
||
}
|
||
}
|
||
|
||
CBotToken::CBotToken(const CBotString& mot, const CBotString& sep, int start, int end)
|
||
{
|
||
m_Text = mot; // word (mot) found as token
|
||
m_Sep = sep; // separator
|
||
m_next = NULL;
|
||
m_prev = NULL;
|
||
m_start = start;
|
||
m_end = end;
|
||
|
||
m_type = TokenTypVar; // at the beginning a default variable type
|
||
m_IdKeyWord = -1;
|
||
}
|
||
|
||
CBotToken::CBotToken(const char* mot, const char* sep)
|
||
{
|
||
m_Text = mot;
|
||
if ( sep != NULL ) m_Sep = sep;
|
||
m_next = NULL;
|
||
m_prev = NULL;
|
||
|
||
m_type = TokenTypVar; // at the beginning a default variable type
|
||
m_IdKeyWord = -1;
|
||
}
|
||
|
||
CBotToken::~CBotToken()
|
||
{
|
||
delete m_next; // recursive
|
||
m_next = NULL;
|
||
}
|
||
|
||
void CBotToken::Free()
|
||
{
|
||
m_ListKeyDefine.SetSize(0);
|
||
}
|
||
|
||
const CBotToken& CBotToken::operator=(const CBotToken& src)
|
||
{
|
||
delete m_next;
|
||
m_next = NULL;
|
||
m_prev = NULL;
|
||
|
||
m_Text = src.m_Text;
|
||
m_Sep = src.m_Sep;
|
||
|
||
m_type = src.m_type;
|
||
m_IdKeyWord = src.m_IdKeyWord;
|
||
|
||
m_start = src.m_start;
|
||
m_end = src.m_end;
|
||
return *this;
|
||
}
|
||
|
||
|
||
int CBotToken::GetType()
|
||
{
|
||
if (this == NULL) return 0;
|
||
if (m_type == TokenTypKeyWord) return m_IdKeyWord;
|
||
return m_type;
|
||
}
|
||
|
||
long CBotToken::GetIdKey()
|
||
{
|
||
return m_IdKeyWord;
|
||
}
|
||
|
||
CBotToken* CBotToken::GetNext()
|
||
{
|
||
if (this == NULL) return NULL;
|
||
return m_next;
|
||
}
|
||
|
||
CBotToken* CBotToken::GetPrev()
|
||
{
|
||
if (this == NULL) return NULL;
|
||
return m_prev;
|
||
}
|
||
|
||
void CBotToken::AddNext(CBotToken* p)
|
||
{
|
||
CBotToken* n = new CBotToken(p);
|
||
CBotToken* pt = this;
|
||
|
||
while ( pt->m_next != NULL ) pt = pt->m_next;
|
||
|
||
pt->m_next = n;
|
||
n->m_prev = pt;
|
||
}
|
||
|
||
|
||
CBotString& CBotToken::GetString()
|
||
{
|
||
return m_Text;
|
||
}
|
||
|
||
CBotString& CBotToken::GetSep()
|
||
{
|
||
return m_Sep;
|
||
}
|
||
|
||
void CBotToken::SetString(const char* name)
|
||
{
|
||
m_Text = name;
|
||
}
|
||
|
||
|
||
int CBotToken::GetStart()
|
||
{
|
||
if (this == NULL) return -1;
|
||
return m_start;
|
||
}
|
||
|
||
int CBotToken::GetEnd()
|
||
{
|
||
if (this == NULL) return -1;
|
||
return m_end;
|
||
}
|
||
|
||
void CBotToken::SetPos(int start, int end)
|
||
{
|
||
m_start = start;
|
||
m_end = end;
|
||
}
|
||
|
||
bool CharInList(const char c, const char* list)
|
||
{
|
||
int i = 0;
|
||
|
||
while (true)
|
||
{
|
||
if (c == list[i++]) return true;
|
||
if (list[i] == 0) return false;
|
||
}
|
||
}
|
||
|
||
bool Char2InList(const char c, const char cc, const char* list)
|
||
{
|
||
int i = 0;
|
||
|
||
while (true)
|
||
{
|
||
if (c == list[i++] &&
|
||
cc == list[i++]) return true;
|
||
|
||
if (list[i] == 0) return false;
|
||
}
|
||
}
|
||
|
||
static char sep1[] = " \r\n\t,:()[]{}-+*/=;><!~^|&%.";
|
||
static char sep2[] = " \r\n\t"; // only separators
|
||
static char sep3[] = ",:()[]{}-+*/=;<>!~^|&%."; // operational separators
|
||
static char num[] = "0123456789"; // point (single) is tested separately
|
||
static char hexnum[] = "0123456789ABCDEFabcdef";
|
||
static char nch[] = "\"\r\n\t"; // forbidden in chains
|
||
|
||
//static char* duo = "+=-=*=/===!=<=>=++--///**/||&&"; // double operators
|
||
|
||
// looking for the next token in a sentence
|
||
// do not start with separators
|
||
// which are made in the previous token
|
||
CBotToken* CBotToken::NextToken(char* &program, int& error, bool first)
|
||
{
|
||
CBotString mot; // the word which is found
|
||
CBotString sep; // separators that are after
|
||
char c;
|
||
bool stop = first;
|
||
|
||
if (*program == 0) return NULL;
|
||
|
||
c = *(program++); // next character
|
||
|
||
if (!first)
|
||
{
|
||
mot = c; // built the word
|
||
c = *(program++); // next character
|
||
|
||
// special case for strings
|
||
if ( mot[0] == '\"' )
|
||
{
|
||
while (c != 0 && !CharInList(c, nch))
|
||
{
|
||
if ( c == '\\' )
|
||
{
|
||
c = *(program++); // next character
|
||
if ( c == 'n' ) c = '\n';
|
||
if ( c == 'r' ) c = '\r';
|
||
if ( c == 't' ) c = '\t';
|
||
}
|
||
mot += c;
|
||
c = *(program++);
|
||
}
|
||
if ( c == '\"' )
|
||
{
|
||
mot += c; // string is complete
|
||
c = *(program++); // next character
|
||
}
|
||
stop = true;
|
||
}
|
||
|
||
// special case for numbers
|
||
if ( CharInList(mot[0], num ))
|
||
{
|
||
bool bdot = false; // found a point?
|
||
bool bexp = false; // found an exponent?
|
||
|
||
char* liste = num;
|
||
if (mot[0] == '0' && c == 'x') // hexadecimal value?
|
||
{
|
||
mot += c;
|
||
c = *(program++); // next character
|
||
liste = hexnum;
|
||
}
|
||
cw:
|
||
while (c != 0 && CharInList(c, liste))
|
||
{
|
||
cc: mot += c;
|
||
c = *(program++); // next character
|
||
}
|
||
if ( liste == num ) // not for hexadecimal
|
||
{
|
||
if ( !bdot && c == '.' ) { bdot = true; goto cc; }
|
||
if ( !bexp && ( c == 'e' || c == 'E' ) )
|
||
{
|
||
bexp = true;
|
||
mot += c;
|
||
c = *(program++); // next character
|
||
if ( c == '-' ||
|
||
c == '+' ) goto cc;
|
||
goto cw;
|
||
}
|
||
|
||
}
|
||
stop = true;
|
||
}
|
||
|
||
if (CharInList(mot[0], sep3)) // an operational separator?
|
||
{
|
||
CBotString motc = mot;
|
||
while (motc += c, c != 0 && GetKeyWords(motc)>0) // operand seeks the longest possible
|
||
{
|
||
mot += c; // build the word
|
||
c = *(program++); // next character
|
||
}
|
||
|
||
stop = true;
|
||
}
|
||
}
|
||
|
||
|
||
|
||
while (true)
|
||
{
|
||
if (stop || c == 0 || CharInList(c, sep1))
|
||
{
|
||
if (!first && mot.IsEmpty()) return NULL; // end of the analysis
|
||
bis:
|
||
while (CharInList(c, sep2))
|
||
{
|
||
sep += c; // after all the separators
|
||
c = *(program++);
|
||
}
|
||
if (c == '/' && *program == '/') // comment on the heap?
|
||
{
|
||
while( c != '\n' && c != 0 )
|
||
{
|
||
sep += c;
|
||
c = *(program++); // next character
|
||
}
|
||
goto bis;
|
||
}
|
||
|
||
if (c == '/' && *program == '*') // comment on the heap?
|
||
{
|
||
while( c != 0 && (c != '*' || *program != '/'))
|
||
{
|
||
sep += c;
|
||
c = *(program++); // next character
|
||
}
|
||
if ( c != 0 )
|
||
{
|
||
sep += c;
|
||
c = *(program++); // next character
|
||
sep += c;
|
||
c = *(program++); // next character
|
||
}
|
||
goto bis;
|
||
}
|
||
|
||
program--;
|
||
|
||
CBotToken* token = new CBotToken(mot, sep);
|
||
|
||
if (CharInList( mot[0], num )) token->m_type = TokenTypNum;
|
||
if (mot[0] == '\"') token->m_type = TokenTypString;
|
||
if (first) token->m_type = 0;
|
||
|
||
token->m_IdKeyWord = GetKeyWords(mot);
|
||
if (token->m_IdKeyWord > 0) token->m_type = TokenTypKeyWord;
|
||
else GetKeyDefNum(mot, token) ; // treats DefineNum
|
||
|
||
return token;
|
||
}
|
||
|
||
mot += c; // built the word
|
||
c = *(program++); // next character
|
||
}
|
||
}
|
||
|
||
CBotToken* CBotToken::CompileTokens(const char* program, int& error)
|
||
{
|
||
CBotToken *nxt, *prv, *tokenbase;
|
||
char* p = const_cast<char*> ( program);
|
||
int pos = 0;
|
||
|
||
error = 0;
|
||
prv = tokenbase = NextToken(p, error, true);
|
||
|
||
if (tokenbase == NULL) return NULL;
|
||
|
||
tokenbase->m_start = pos;
|
||
pos += tokenbase->m_Text.GetLength();
|
||
tokenbase->m_end = pos;
|
||
pos += tokenbase->m_Sep.GetLength();
|
||
|
||
char* pp = p;
|
||
while (NULL != (nxt = NextToken(p, error)))
|
||
{
|
||
prv->m_next = nxt; // added after
|
||
nxt->m_prev = prv;
|
||
prv = nxt; // advance
|
||
|
||
nxt->m_start = pos;
|
||
/* pos += nxt->m_Text.GetLength(); // chain may be shorter (BOA deleted)
|
||
nxt->m_end = pos;
|
||
pos += nxt->m_Sep.GetLength();*/
|
||
pos += (p - pp); // total size
|
||
nxt->m_end = pos - nxt->m_Sep.GetLength();
|
||
pp = p;
|
||
}
|
||
|
||
// adds a token as a terminator
|
||
// ( useful for the previous )
|
||
nxt = new CBotToken();
|
||
nxt->m_type = 0;
|
||
prv->m_next = nxt; // added after
|
||
nxt->m_prev = prv;
|
||
|
||
return tokenbase;
|
||
}
|
||
|
||
void CBotToken::Delete(CBotToken* pToken)
|
||
{
|
||
delete pToken;
|
||
}
|
||
|
||
|
||
// search if a word is part of the keywords
|
||
|
||
int CBotToken::GetKeyWords(const char* w)
|
||
{
|
||
int i;
|
||
int l = m_ListKeyWords.GetSize();
|
||
|
||
if (l == 0)
|
||
{
|
||
LoadKeyWords(); // takes the list for the first time
|
||
l = m_ListKeyWords.GetSize();
|
||
}
|
||
|
||
for (i = 0; i < l; i++)
|
||
{
|
||
if (m_ListKeyWords[i] == w) return m_ListIdKeyWords[ i ];
|
||
}
|
||
|
||
return -1;
|
||
}
|
||
|
||
bool CBotToken::GetKeyDefNum(const char* w, CBotToken* &token)
|
||
{
|
||
int i;
|
||
int l = m_ListKeyDefine.GetSize();
|
||
|
||
for (i = 0; i < l; i++)
|
||
{
|
||
if (m_ListKeyDefine[i] == w)
|
||
{
|
||
token->m_IdKeyWord = m_ListKeyNums[i];
|
||
token->m_type = TokenTypDef;
|
||
return true;
|
||
}
|
||
}
|
||
|
||
return false;
|
||
}
|
||
|
||
|
||
/// \todo Fixme Figure out how this should work.
|
||
|
||
// recreates the list of keywords and its IDs basing on some resources
|
||
// defines of TokenKey.. are in CBotDll.h
|
||
|
||
void CBotToken::LoadKeyWords()
|
||
{
|
||
CBotString s;
|
||
int i, n = 0;
|
||
|
||
i = TokenKeyWord; //start with keywords of the language
|
||
while (s.LoadString(i))
|
||
{
|
||
m_ListKeyWords.Add(s);
|
||
m_ListIdKeyWords[n++] = i++;
|
||
}
|
||
|
||
i = TokenKeyDeclare; //keywords of declarations
|
||
while (s.LoadString(i))
|
||
{
|
||
m_ListKeyWords.Add(s);
|
||
m_ListIdKeyWords[n++] = i++;
|
||
}
|
||
|
||
|
||
i = TokenKeyVal; //keywords of values
|
||
while (s.LoadString(i))
|
||
{
|
||
m_ListKeyWords.Add(s);
|
||
m_ListIdKeyWords[n++] = i++;
|
||
}
|
||
|
||
i = TokenKeyOp; //operators
|
||
while (s.LoadString(i))
|
||
{
|
||
m_ListKeyWords.Add(s);
|
||
m_ListIdKeyWords[n++] = i++;
|
||
}
|
||
}
|
||
|
||
bool CBotToken::DefineNum(const char* name, long val)
|
||
{
|
||
int i;
|
||
int l = m_ListKeyDefine.GetSize();
|
||
|
||
for (i = 0; i < l; i++)
|
||
{
|
||
if (m_ListKeyDefine[i] == name) return false;
|
||
}
|
||
if ( i == MAXDEFNUM ) return false;
|
||
|
||
m_ListKeyDefine.Add( name );
|
||
m_ListKeyNums[i] = val;
|
||
return true;
|
||
}
|
||
|
||
bool IsOfType(CBotToken* &p, int type1, int type2)
|
||
{
|
||
if (p->GetType() == type1 ||
|
||
p->GetType() == type2 )
|
||
{
|
||
p = p->GetNext();
|
||
return true;
|
||
}
|
||
return false;
|
||
}
|
||
// Same with any number of arguments
|
||
// There must be a zero as the last argument
|
||
bool IsOfTypeList(CBotToken* &p, int type1, ...)
|
||
{
|
||
int i = type1;
|
||
int max = 20;
|
||
int type = p->GetType();
|
||
|
||
va_list marker;
|
||
va_start( marker, type1 ); /* Initialize variable arguments. */
|
||
|
||
while (true)
|
||
{
|
||
if (type == i)
|
||
{
|
||
p = p->GetNext();
|
||
va_end( marker ); /* Reset variable arguments. */
|
||
return true;
|
||
}
|
||
if (--max == 0 || 0 == (i = va_arg( marker, int)))
|
||
{
|
||
va_end( marker ); /* Reset variable arguments. */
|
||
return false;
|
||
}
|
||
}
|
||
}
|
||
|