Power Programming/powerprogramming1994.iso/progtool/turbopas/tptc17sc.arc/TPCSCAN.INC

Text File | 1988-03-26 | 20KB | 834 lines

(*
 * TPTC - Turbo Pascal to C translator
 *
 * (C) 1988 Samuel H. Smith (rev. 13-Feb-88)
 *
 *)


(********************************************************************)
(*
 * lexical scanner
 *
 *)

function numlit(n: integer): anystring;
var
   lit: string[6];

   {convert an integer into a c style numeric character literal}
   function digit(n: integer): char;
      (* convert an integer into a hex digit *)
   begin
      n := n and 15;
      if n > 9 then n := n + 7;
      digit := chr( n + ord('0') );
   end;

begin
   lit := '''\?''';

   case n of
     $07:   lit[3] := 'a';
     $08:   lit[3] := 'b';
     $09:   lit[3] := 't';
     $0a:   lit[3] := 'n';
     $0b:   lit[3] := 'v';
     $0c:   lit[3] := 'f';
     $0d:   lit[3] := 'r';

     32..126,128..254:
            lit := ''''+chr(n)+'''';

     else   begin
               lit := '''\x??''';
               lit[4] := digit(n shr 4);
               lit[5] := digit(n);
            end;
   end;

   numlit := lit;
   toktype := chars;
end;


(********************************************************************)
procedure getchar;
   {consume the current char and get the next one}
var
   stack: char;
begin
   if ofs(stack) < minstack then
      fatal('Out of stack space');

   while (srclevel > 0) and eof(srcfd[srclevel]) do
   begin
      if not linestart then putline;
      putln('/* TPTC: end of '+srcfiles[srclevel]+' */');

      if debug then writeln;
      writeln(^M,srcfiles[srclevel],'(',srclines[srclevel],')');

      close(srcfd[srclevel]);
      freemem(inbuf[srclevel],inbufsiz);

      dec(srclevel);
      statustime := 0;
   end;

   if eof(srcfd[srclevel]) then
      nextc := '.'
   else
      read(srcfd[srclevel], nextc);

   if nextc = ^J then
   begin
      inc(srclines[srclevel]);
      inc(srctotal);

      mark_time(curtime);
      if (curtime >= statustime) or debug then
      begin
         if debug then writeln;
         write(^M,srcfiles[srclevel],'(',srclines[srclevel],')');
         statustime := curtime+statrate;
         abortcheck;
      end;
   end;
end;


(********************************************************************)
function usec: char;
   {use up the current character(return it) and get
    the next one from the input stream}
var
   c: char;
begin
   c := nextc;
   getchar;
   usec := c;
end;


(********************************************************************)
function newc(n: string40): string40;
   {replace the current character with a different one and get the next
    character from the input stream}
var
   c: char;
begin
   c := nextc;
   getchar;
   newc := n;
end;


(********************************************************************)
procedure concat_tokens;
   {concatenate the next token and the current token}
var
   cur: string;
begin
   cur := ltok;
   ltok := nextc;
   toktype := unknown;
   scan_tok;

   ltok := copy(cur,1,length(cur)-1) + copy(ltok,2,255);
   ltok[1] := '"';
   ltok[length(ltok)] := '"';
   toktype := strng;
end;


(********************************************************************)
procedure scan_ident;
   {scan an identifier; output is ltok; nextc is first character following
    the identifier; toktype = identifier;  this is the protocol for all of
    the scan_xxxx procedures in the lexical analyzer}
begin

   toktype := unknown;
   ltok := '';

   repeat
      case nextc of
         'A'..'Z':
            begin
               if map_lower then
                  nextc := chr( ord(nextc)+32 );
               ltok := ltok + nextc;
               getchar;
            end;

         'a'..'z', '0'..'9', '_','@':
            ltok := ltok + usec;

         else
            toktype := identifier;
      end;

   until toktype = identifier;
end;



(********************************************************************)
procedure scan_preproc;
   {scan a tshell preprocessor directive;  same syntax as C already}
begin
   puts('#');

   repeat
      puts(nextc);
      getchar;
   until nextc = ^M;

   getchar;
   putline;
   toktype := unknown;
end;


(********************************************************************)
procedure scan_number;
   {scan a number;  this also processes #nnn character literals, which are
    converted into octal character literals.  imbedded periods are processed,
    and a special condition is noted for trailing periods.  this is needed
    for scanning the ".." keyword when used after numbers.  an ungetchar
    facility would be more general, but isn't needed anywhere else.
    in pascal/mt+, #nnn is translated into nnnL }
var
   hasdot:  boolean;
   charlit:   boolean;
   islong:  boolean;

begin
   hasdot := false;
   islong := false;
   charlit := false;
   toktype := number;

(* check for preprocessor directives, character literals or long literals *)
   if nextc = '#' then
   begin
      ltok := '';
      if mt_plus then
         islong := true
      else
         charlit := true;
   end;

   getchar;

(* check for preprocessor directives *)
   if tshell and charlit and (nextc >= 'a') and (nextc <= 'z') then
      scan_preproc
   else

   repeat
      case nextc of
         '$','0'..'9','a'..'f','A'..'F':
            ltok := ltok + usec;

         '.':
            if hasdot then
            begin
               if ltok[length(ltok)] = '.' then
               begin
                  ltok[0] := pred(ltok[0]);  {remove trailing ., part of ..}
                  if charlit then
                     ltok := numlit(atoi(ltok));
                  extradot := true;
               end;
               exit;
            end
            else

            begin
               hasdot := true;
               ltok := ltok + usec;
            end;

         else
            begin
               if charlit then
               begin
                  ltok := numlit(atoi(ltok));
                  if (nextc = '''') or (nextc = '^') or (nextc = '#') then
                     concat_tokens;
                  exit;
               end;

               if ltok[1] = '$' then
                  ltok := '0x' + copy(ltok,2,99);
               if islong then
                  ltok := ltok + 'L';
               exit;
            end;
      end;

   until true=false;
end;


(********************************************************************)
procedure scan_hat;
   {scan tokens starting with ^ - returns ^X as a character literal
    corresponding to the specified control character.  returns ^ident as
    an identifier with the leading ^ intact.  also scans ^. and ^[.}
var
   c: char;

begin
   getchar;

   if ((nextc = '.') or (nextc = '[')) and
      ((ptoktype = identifier) or (ptok = ']')) then
   begin
      ltok := '^' + usec;     {^. or ^[}
      exit;
   end;

   case nextc of
      '@','['..'`':
         ltok := usec;

      'A'..'Z','a'..'z':
         begin
            ltok := nextc;
            scan_ident;
         end;
       else
         exit;
   end;

   if length(ltok) = 1 then      {^c = control char}
   begin
      ltok := numlit( ord(upcase(ltok[1])) - ord('@') );
      if (nextc = '''') or (nextc = '^') or (nextc = '#') then
         concat_tokens;
   end
   else
      ltok := '^' + ltok;        {^ident = pointer to ident}

end;


(********************************************************************)
procedure scan_dot;
   {scans tokens starting with "."; knows about the 'extra dot' condition
    that comes up in number scanning.  returns a token of either '.' or '..'}
begin
   getchar;

   if (nextc = '.') or extradot then
   begin
      ltok := '..';
      extradot := false;
   end;

   if nextc = '.' then
      getchar;
end;


(********************************************************************)
procedure scan_string;
   {scans a literal string.  processes imbedded quotes ala pascal.  translates
    the string into a C string with the proper escapes on imbedded quotes.
    converts single character strings into character constants.  these are
    sometimes converted back to strings when the parser needs to}
begin
   toktype := unknown;
   ltok := '"';
   getchar;

   repeat
      case nextc of
         ^J,^M:
            begin
               error_message('Closing quote expected (scan_string)');
               toktype := strng;
            end;

         '''':
            begin
               getchar;     {consume the quote}

               if nextc = '''' then
                  ltok := ltok + usec
                  {double quotes are coded as a single quote}
               else

               begin        {end of string}
                  ltok := ltok + '"';
                  toktype := strng;
               end;
            end;

         '"':  ltok := ltok + newc('\"');
         '\':  ltok := ltok + newc('\\');

         else  ltok := ltok + usec;
      end;

   until toktype = strng;

   if length(ltok) = 3 then
   begin
      ltok[1] := '''';
      ltok[3] := '''';
      toktype := chars;
   end;

   if ltok = '"\""' then
   begin
      ltok := '''"''';
      toktype := chars;
   end
   else

   if (ltok = '"''"') or (ltok = '''''''') then
      ltok := '''\'''''
   else

   if (ltok = '"\\"') then
   begin
      ltok := '''\\''';
      toktype := chars;
   end;

   if (nextc = '^') or (nextc = '#') then
      concat_tokens;
end;


(********************************************************************)
procedure scan_pragma(var isinclude: anystring);
   {scans a turbo pascal compiler option and translates it into a comment.
    include directive is translated into the #include.
    returns with the first non-blank after the pragma}
var
   code: anystring;
   prag: anystring;
   arg:  anystring;

   procedure scanword(var dest: anystring);
   begin
      dest := '       ';  {insure dest[2] is initialized}
      dest := '';
      while true do
      case nextc of
         ' ', '*', '}', ',':
            exit;
         else
         begin
            dest := dest + upcase(nextc);
            getchar;
         end;
      end;
   end;

begin
   isinclude := '';

   repeat
      if nextc = ',' then
         newline;

      getchar;   {consume the $ or ,}

      {get the progma code}
      scanword(code);

      if nextc = ' ' then
      begin
         getchar;
         scanword(arg);
      end
      else
         arg := '';

      if code[2] = '+' then
         arg := 'ON'
      else
      if code[2] = '-' then
         arg := 'OFF';

      prag := '/* '+code[1]+'(' + arg + ')' + ' */';

      case code[1] of

         'D':  if code[2] = 'E' then
                  prag := '#define '+arg;

         'E':  if code[2] = 'N' then
                  prag := '#endif'
               else
               if code[2] = 'L' then
                  prag := '#else';

         'I':  if code[2] = ' ' then
               begin
                  if pos('.',arg) = 0 then
                     arg := arg + '.PAS';
                  prag := '#include "' + arg + '"   ';

                  if includeinclude then
                  begin
                     prag := '';
                     isinclude := arg;
                  end;
               end
               else

               if code[2] = 'F' then
               begin
                  if code[3] = 'N' then
                     prag := '#ifndef '+arg
                  else
                     prag := '#ifdef '+arg;
               end;

         'U':  if code[2] = 'N' then
                  prag := '#undef '+arg;

      end;

      puts(prag);
      puts('   ');

      while nextc = ' ' do
         getchar;

   until nextc <> ',';

end;


(********************************************************************)
procedure open_include(name: anystring);
begin
   if length(name) = 0 then exit;

   inc(srctotal);
   inc(objtotal);

   inc(srclevel);
   if srclevel > maxincl then
      fatal('Includes nested too deeply');

   srcfiles[srclevel] := name;
   srclines[srclevel] := 1;

   assign(srcfd[srclevel],name);
   {$I-} reset(srcfd[srclevel]); {$I+}
   if ioresult <> 0 then
   begin
      dec(srclevel);
      ltok := name;
      warning('Missing include file');
   end
   else

   begin
      if not linestart then putline;
      putln('/* TPTC: include '+name+' */');

      if maxavail-300 <= inbufsiz then
      begin
         ltok := name;
         fatal('Out of memory');
      end;

      getmem(inbuf[srclevel],inbufsiz);
      SetTextBuf(srcfd[srclevel],inbuf[srclevel]^,inbufsiz);
   end;

   if {quietmode and} not debug then
      write(^M,'':40,^M)
   else
      writeln;
   statustime := 0;
end;


(********************************************************************)
procedure scan_curlycomment;
   {processes a curly-brace enclosed comment}
var
   isinclude: anystring;

begin
   toktype := comment;
   getchar;   {consume the open comment}

   isinclude := '';
   if nextc = '$' then
      scan_pragma(isinclude);

   if nextc = '}' then
   begin
      getchar;
      open_include(isinclude);
      exit;
   end;

   if pass_comments then
      puts('  /* ');

   while nextc <> '}' do
   begin
      if pass_comments then
         puts(nextc);
      getchar;
   end;

   if pass_comments then
   begin
      puts(' */ ');
      if nospace then newline;
   end;

   getchar;   {consume the close comment}
   open_include(isinclude);
end;


(********************************************************************)
procedure scan_parencomment;
   {process a (* enclosed comment}
var
   isinclude: anystring;

begin
   toktype := comment;
   getchar;   {consume the *}

   isinclude := '';
   if nextc = '$' then
      scan_pragma(isinclude);

   if pass_comments then
      puts('/*');

   repeat
      if pass_comments then
         puts(nextc);

      if nextc = '*' then
      begin
         getchar;

         if nextc = ')' then
         begin
            getchar;
            if pass_comments then
            begin
               puts('/ ');
               if nospace then putline;
            end;
            open_include(isinclude);
            exit;
         end;
      end
      else
         getchar;

   until true=false;
end;


(********************************************************************)
procedure scan_blanks;
   {scan white space.  this procedure sometimes passes whitespace to the
    output.  it keeps track of the indentation of the current line so it
    can be used by newline}
var
   indent:        anystring;
   valid:         boolean;

begin
   linestart := false;
   indent := '';
   valid := false;

   repeat

      case nextc of
         ^J,^M:  begin
                    if (nospace = false) and (nextc = ^J) then
                       putline;

                    indent := '';
                    linestart := true;
                    getchar;
                 end;

         ' ',^I,^@,^L:
                 indent := indent + usec;

         '#':    if linestart and tshell then
                 begin
                    puts(indent);     {pass preprocessor directives}
                    indent := '';     {without change (single-line only)}

                    repeat
                       puts(nextc);
                       getchar;
                    until nextc = ^M;

                    getchar;
                    putline;
                 end
                 else
                    valid := true;

         else
                 valid := true;
      end;

   until valid;

   if linestart then
   begin
      spaces := indent;
      if nospace=false then
         puts(spaces);

      linestart := true;
   end;
end;


(********************************************************************)
procedure scan_tok;
   {scans the next lexical token; returns the token in ltok and toktype}
begin
   scan_blanks;

   toktype := unknown;
   ltok := nextc;

   case nextc of
      'a'..'z',
      '_', 'A'..'Z': scan_ident;

      '$':           scan_number;
      '0'..'9':      scan_number;

      '''':          scan_string;

      '^':           scan_hat;

      '#':           begin
                        scan_number;
                        if toktype = unknown then
                           scan_tok;         {in case of #directive}
                     end;


      '<':           begin
                        getchar;
                        if (nextc = '>') or (nextc = '=') then
                           ltok := '<' + usec;
                     end;

      '>':           begin
                        getchar;
                        if nextc = '=' then
                           ltok := '>' + usec;
                     end;

      ':':           begin
                        getchar;
                        if nextc = '=' then
                           ltok := ':' + usec;
                     end;

      '.':           scan_dot;

      '{':           scan_curlycomment;

      '(':           begin
                        getchar;
                        if nextc = '*' then
                           scan_parencomment;
                     end;

      else           getchar;   {consume the unknown char}
   end;
end;


(********************************************************************)
procedure gettok;
   {get the next input token;  this is the top level of the lexical analyzer.
    it returns ltok, tok(ltok in upper case), toktype.  it translates BEGIN
    and END into braces; it checks for statement and section keywords}
var
   i:             integer;

begin
   ptoktype := toktype;
   ptok := tok;
   cursym := nil;

   repeat
      scan_tok;
   until toktype <> comment;
   tok := ltok;

   if debug then write(' {',ltok,'}');

   if toktype = identifier then
   begin
      stoupper(tok);

      if tok = 'BEGIN' then
      begin
         tok := '{';
         ltok := tok;
         toktype := keyword;
      end
      else

      if tok = 'END' then
      begin
         tok := '}';
         ltok := tok;
         toktype := keyword;
      end;

      (* check for statement keywords *)
      i := 0;
      repeat
         inc(i);
         if tok[1] = keywords[i][1] then              {hack for speed}
         if length(tok) = length(keywords[i]) then
         if tok = keywords[i] then
            toktype := keyword;
      until (i = nkeywords) or (toktype = keyword);

      (* get symbol table information for this item *)
      cursym := locatesym(tok);
   end;
end;


(********************************************************************)
function usetok: string80;
   {return (use) and consume current token}
var
   tv: string80;
begin
   tv := ltok;
   gettok;
   usetok := tv;
end;