source/glsl/tokenizer.cpp

   1 #include <msp/core/raii.h>
   2 #include "glsl_error.h"
   3 #include "preprocessor.h"
   4 #include "syntax.h"
   5 #include "tokenizer.h"
   6
   7 using namespace std;
   8
   9 namespace Msp {
  10 namespace GL {
  11 namespace SL {
  12
  13 Tokenizer::Tokenizer():
  14         allow_preprocess(true),
  15         suppress_line_advance(false)
  16 {
  17         static string empty;
  18         iter = empty.begin();
  19         source_end = empty.end();
  20 }
  21
  22 void Tokenizer::begin(const string &src, const string &name)
  23 {
  24         iter = src.begin();
  25         source_end = src.end();
  26         location.name = name;
  27         location.line = 1;
  28         allow_preprocess = true;
  29         last_token.clear();
  30         next_tokens.clear();
  31 }
  32
  33 const string &Tokenizer::peek_token(unsigned index)
  34 {
  35         while(next_tokens.size()<=index)
  36                 next_tokens.push_back(parse_token_());
  37         return next_tokens[index];
  38 }
  39
  40 const string &Tokenizer::parse_token()
  41 {
  42         progress_mark = true;
  43
  44         if(!next_tokens.empty())
  45         {
  46                 last_token = next_tokens.front();
  47                 next_tokens.pop_front();
  48                 return last_token;
  49         }
  50
  51         return (last_token = parse_token_());
  52 }
  53
  54 void Tokenizer::expect(const string &token)
  55 {
  56         string parsed = parse_token();
  57         if(parsed!=token)
  58                 throw parse_error(location, parsed, format("'%s'", token));
  59 }
  60
  61 void Tokenizer::set_location(const Location &loc)
  62 {
  63         location = loc;
  64         suppress_line_advance = true;
  65 }
  66
  67 string Tokenizer::parse_token_()
  68 {
  69         while(1)
  70         {
  71                 skip_comment_and_whitespace();
  72                 bool allow_preproc = allow_preprocess;
  73                 allow_preprocess = false;
  74                 if(iter==source_end)
  75                         return string();
  76                 else if(allow_preproc && *iter=='#')
  77                 {
  78                         ++iter;
  79                         preprocess();
  80                 }
  81                 else if(isalpha(*iter) || *iter=='_')
  82                         return parse_identifier();
  83                 else if(isdigit(*iter))
  84                         return parse_number();
  85                 else if(*iter=='"')
  86                         return parse_string();
  87                 else if(*iter=='#' || *iter=='$' || *iter=='\'' || *iter=='@' || *iter=='\\' || *iter=='`')
  88                         throw syntax_error(location, string(1, *iter), "Invalid character in source");
  89                 else
  90                         return parse_other();
  91         }
  92 }
  93
  94 void Tokenizer::preprocess()
  95 {
  96         SetForScope<deque<string> > clear_tokens(next_tokens, deque<string>());
  97
  98         auto line_end = iter;
  99         for(; (line_end!=source_end && *line_end!='\n'); ++line_end) ;
 100         SetForScope<string::const_iterator> stop_at_line_end(source_end, line_end);
 101
 102         signal_preprocess.emit();
 103
 104         iter = line_end;
 105 }
 106
 107 string Tokenizer::parse_identifier()
 108 {
 109         string ident;
 110         while(iter!=source_end)
 111         {
 112                 if(isalnum(*iter) || *iter=='_')
 113                         ident += *iter++;
 114                 else
 115                         break;
 116         }
 117
 118         return ident;
 119 }
 120
 121 string Tokenizer::parse_number()
 122 {
 123         bool got_fract = false;
 124         string number;
 125         while(iter!=source_end)
 126         {
 127                 if(isdigit(*iter))
 128                         number += *iter++;
 129                 else if(!got_fract && *iter=='.')
 130                 {
 131                         number += *iter++;
 132                         got_fract = true;
 133                 }
 134                 else
 135                         break;
 136         }
 137
 138         bool require_digit = false;
 139         if(iter==source_end)
 140                 ;
 141         else if(*iter=='e' || *iter=='E')
 142         {
 143                 number += *iter++;
 144                 if(iter!=source_end && (*iter=='-' || *iter=='+'))
 145                         number += *iter++;
 146                 require_digit = true;
 147                 while(iter!=source_end)
 148                 {
 149                         if(isdigit(*iter))
 150                         {
 151                                 number += *iter++;
 152                                 require_digit = false;
 153                         }
 154                         else
 155                                 break;
 156                 }
 157                 if(!require_digit && iter!=source_end && *iter=='f')
 158                         number += *iter++;
 159         }
 160         else if(*iter=='u' || *iter=='f')
 161                 number += *iter++;
 162
 163         if(require_digit)
 164                 throw syntax_error(location, number, "Incomplete numeric literal");
 165         else if(iter!=source_end && (isalnum(*iter) || *iter=='_'))
 166                 throw syntax_error(location, number, "Garbage at end of numeric literal");
 167
 168         return number;
 169 }
 170
 171 string Tokenizer::parse_string()
 172 {
 173         string str(1, *iter++);
 174         bool escape = false;
 175         while(iter!=source_end)
 176         {
 177                 char c = *iter++;
 178                 str += c;
 179                 if(c=='\\')
 180                         escape = true;
 181                 else if(c=='"' && !escape)
 182                         break;
 183                 else
 184                         escape = false;
 185         }
 186
 187         return str;
 188 }
 189
 190 string Tokenizer::parse_other()
 191 {
 192         if(iter==source_end)
 193                 return string();
 194
 195         string token(1, *iter++);
 196         for(unsigned i=1; (i<3 && iter!=source_end); ++i)
 197         {
 198                 bool matched = false;
 199                 for(const Operator *j=Operator::operators; (!matched && j->type); ++j)
 200                 {
 201                         matched = (j->token[i]==*iter);
 202                         for(unsigned k=0; (matched && k<i && j->token[k]); ++k)
 203                                 matched = (j->token[k]==token[k]);
 204                 }
 205
 206                 if(!matched)
 207                         break;
 208
 209                 token += *iter++;
 210         }
 211
 212         return token;
 213 }
 214
 215 void Tokenizer::skip_comment_and_whitespace()
 216 {
 217         unsigned comment = 0;
 218         while(iter!=source_end)
 219         {
 220                 if(comment==0)
 221                 {
 222                         if(*iter=='/')
 223                                 comment = 1;
 224                         else if(!isspace(*iter))
 225                                 break;
 226                 }
 227                 else if(comment==1)
 228                 {
 229                         if(*iter=='/')
 230                                 comment = 2;
 231                         else if(*iter=='*')
 232                                 comment = 3;
 233                         else
 234                         {
 235                                 comment = 0;
 236                                 --iter;
 237                                 break;
 238                         }
 239                 }
 240                 else if(comment==2)
 241                 {
 242                         if(*iter=='\n')
 243                                 comment = 0;
 244                 }
 245                 else if(comment==3 && *iter=='*')
 246                         comment = 4;
 247                 else if(comment==4)
 248                 {
 249                         if(*iter=='/')
 250                                 comment = 0;
 251                         else if(*iter!='*')
 252                                 comment = 3;
 253                 }
 254
 255                 if(*iter=='\n')
 256                 {
 257                         if(!suppress_line_advance)
 258                                 ++location.line;
 259                         allow_preprocess = (comment<3);
 260                 }
 261
 262                 ++iter;
 263         }
 264
 265         suppress_line_advance = false;
 266 }
 267
 268 } // namespace SL
 269 } // namespace GL
 270 } // namespace Msp