cweave.w

模拟器提供了一个简单易用的平台

    else if (xisspace(c)) continue; /* ignore spaces and tabs */
    if (c=='#' && loc==buffer+1) @<Raise preprocessor flag@>;
    mistake: @<Compress two-symbol operator@>@;
    return(c);
  }
}

@ Because preprocessor commands do not fit in with the rest of the syntax
of \CEE/,
we have to deal with them separately.  One solution is to enclose such
commands between special markers.  Thus, when a \.\# is seen as the
first character of a line, |get_next| returns a special code
|left_preproc| and raises a flag |preprocessing|.

We can use the same internal code number for |left_preproc| as we do
for |ord|, since |get_next| changes |ord| into a string.

@d left_preproc ord /* begins a preprocessor command */
@d right_preproc 0217 /* ends a preprocessor command */

@<Glob...@>=
boolean preprocessing=0; /* are we scanning a preprocessor command? */

@ @<Raise prep...@>= {
  preprocessing=1;
  @<Check if next token is |include|@>;
  return (left_preproc);
}

@ An additional complication is the freakish use of \.< and \.> to delimit
a file name in lines that start with \.{\#include}.  We must treat this file
name as a string.

@<Glob...@>=
boolean sharp_include_line=0; /* are we scanning a |#include| line? */

@ @<Check if next token is |include|@>=
while (loc<=buffer_end-7 && xisspace(*loc)) loc++;
if (loc<=buffer_end-6 && strncmp(loc,"include",7)==0) sharp_include_line=1;

@ When we get to the end of a preprocessor line,
we lower the flag and send a code |right_preproc|, unless
the last character was a \.\\.

@<Check if we're at...@>=
  while (loc==limit-1 && preprocessing && *loc=='\\')
    if (get_line()==0) return(new_section); /* still in preprocessor mode */
  if (loc>=limit && preprocessing) {
    preprocessing=sharp_include_line=0;
    return(right_preproc);
  }

@ The following code assigns values to the combinations \.{++},
\.{--}, \.{->}, \.{>=}, \.{<=}, \.{==}, \.{<<}, \.{>>}, \.{!=}, \.{\v\v}, and
\.{\&\&}, and to the \CPLUSPLUS/
combinations \.{...}, \.{::}, \.{.*} and \.{->*}.
The compound assignment operators (e.g., \.{+=}) are
treated as separate tokens.

@d compress(c) if (loc++<=limit) return(c)

@<Compress tw...@>=
switch(c) {
  case '/': if (*loc=='*') {compress(begin_comment);}
    else if (*loc=='/') compress(begin_short_comment); break;
  case '+': if (*loc=='+') compress(plus_plus); break;
  case '-': if (*loc=='-') {compress(minus_minus);}
    else if (*loc=='>') if (*(loc+1)=='*') {loc++; compress(minus_gt_ast);}
                        else compress(minus_gt); break;
  case '.': if (*loc=='*') {compress(period_ast);}
            else if (*loc=='.' && *(loc+1)=='.') {
              loc++; compress(dot_dot_dot);
            }
            break;
  case ':': if (*loc==':') compress(colon_colon); break;
  case '=': if (*loc=='=') compress(eq_eq); break;
  case '>': if (*loc=='=') {compress(gt_eq);}
    else if (*loc=='>') compress(gt_gt); break;
  case '<': if (*loc=='=') {compress(lt_eq);}
    else if (*loc=='<') compress(lt_lt); break;
  case '&': if (*loc=='&') compress(and_and); break;
  case '|': if (*loc=='|') compress(or_or); break;
  case '!': if (*loc=='=') compress(not_eq); break;
}

@ @<Get an identifier@>= {
  id_first=--loc;
  while (isalpha(*++loc) || isdigit(*loc) || isxalpha(*loc) || ishigh(*loc));
  id_loc=loc; return(identifier);
}

@ Different conventions are followed by \TEX/ and \CEE/ to express octal
and hexadecimal numbers; it is reasonable to stick to each convention
within its realm.  Thus the \CEE/ part of a \.{CWEB} file has octals
introduced by \.0 and hexadecimals by \.{0x}, but \.{CWEAVE} will print
with \TeX/ macros that the user can redefine to fit the context.
In order to simplify such macros, we replace some of the characters.

Notice that in this section and the next, |id_first| and |id_loc|
are pointers into the array |section_text|, not into |buffer|.

@<Get a constant@>= {
  id_first=id_loc=section_text+1;
  if (*(loc-1)=='\\') {*id_loc++='~';
  while (xisdigit(*loc)) *id_loc++=*loc++;} /* octal constant */
  else if (*(loc-1)=='0') {
    if (*loc=='x' || *loc=='X') {*id_loc++='^'; loc++;
      while (xisxdigit(*loc)) *id_loc++=*loc++;} /* hex constant */
    else if (xisdigit(*loc)) {*id_loc++='~';
      while (xisdigit(*loc)) *id_loc++=*loc++;} /* octal constant */
    else goto dec; /* decimal constant */
  }
  else { /* decimal constant */
    if (*(loc-1)=='.' && !xisdigit(*loc)) goto mistake; /* not a constant */
    dec: *id_loc++=*(loc-1);
    while (xisdigit(*loc) || *loc=='.') *id_loc++=*loc++;
    if (*loc=='e' || *loc=='E') { /* float constant */
      *id_loc++='_'; loc++;
      if (*loc=='+' || *loc=='-') *id_loc++=*loc++;
      while (xisdigit(*loc)) *id_loc++=*loc++;
    }
  }
  while (*loc=='u' || *loc=='U' || *loc=='l' || *loc=='L'
         || *loc=='f' || *loc=='F') {
    *id_loc++='$'; *id_loc++=toupper(*loc); loc++;
  }
  return(constant);
}

@ \CEE/ strings and character constants, delimited by double and single
quotes, respectively, can contain newlines or instances of their own
delimiters if they are protected by a backslash.  We follow this
convention, but do not allow the string to be longer than |longest_name|.

@<Get a string@>= {
  char delim = c; /* what started the string */
  id_first = section_text+1;
  id_loc = section_text;
  if (delim=='\'' && *(loc-2)=='@@') {*++id_loc='@@'; *++id_loc='@@';}
  *++id_loc=delim;
  if (delim=='L') { /* wide character constant */
    delim=*loc++; *++id_loc=delim;
  }
  if (delim=='<') delim='>'; /* for file names in |#include| lines */
  while (1) {
    if (loc>=limit) {
      if(*(limit-1)!='\\') {
        err_print("! String didn't end"); loc=limit; break;
@.String didn't end@>
      }
      if(get_line()==0) {
        err_print("! Input ended in middle of string"); loc=buffer; break;
@.Input ended in middle of string@>
      }
    }
    if ((c=*loc++)==delim) {
      if (++id_loc<=section_text_end) *id_loc=c;
      break;
    }
    if (c=='\\') if (loc>=limit) continue;
      else if (++id_loc<=section_text_end) {
        *id_loc = '\\'; c=*loc++;
      }
    if (++id_loc<=section_text_end) *id_loc=c;
  }
  if (id_loc>=section_text_end) {
    printf("\n! String too long: ");
@.String too long@>
    term_write(section_text+1,25);
    printf("..."); mark_error;
  }
  id_loc++;
  return(string);
}

@ After an \.{@@} sign has been scanned, the next character tells us
whether there is more work to do.

@<Get control code and possible section name@>= {
  c=*loc++;
  switch(ccode[(eight_bits)c]) {
    case translit_code: err_print("! Use @@l in limbo only"); continue;
@.Use @@l in limbo...@>
    case underline: xref_switch=def_flag; continue;
    case trace: tracing=c-'0'; continue;
    case xref_roman: case xref_wildcard: case xref_typewriter:
    case noop: case TeX_string: c=ccode[c]; skip_restricted(); return(c);
    case section_name:
      @<Scan the section name and make |cur_section| point to it@>;
    case verbatim: @<Scan a verbatim string@>;
    case ord: @<Get a string@>;
    default: return(ccode[(eight_bits)c]);
  }
}

@ The occurrence of a section name sets |xref_switch| to zero,
because the section name might (for example) follow \&{int}.

@<Scan the section name...@>= {
  char *k; /* pointer into |section_text| */
  cur_section_char=*(loc-1);
  @<Put section name into |section_text|@>;
  if (k-section_text>3 && strncmp(k-2,"...",3)==0)
        cur_section=section_lookup(section_text+1,k-3,1); /* 1 indicates a prefix */
  else cur_section=section_lookup(section_text+1,k,0);
  xref_switch=0; return(section_name);
}

@ Section names are placed into the |section_text| array with consecutive spaces,
tabs, and carriage-returns replaced by single spaces. There will be no
spaces at the beginning or the end. (We set |section_text[0]=' '| to facilitate
this, since the |section_lookup| routine uses |section_text[1]| as the first
character of the name.)

@<Set init...@>=section_text[0]=' ';

@ @<Put section name...@>=
k=section_text;
while (1) {
  if (loc>limit && get_line()==0) {
    err_print("! Input ended in section name");
@.Input ended in section name@>
    loc=buffer+1; break;
  }
  c=*loc;
  @<If end of name or erroneous control code, |break|@>;
  loc++; if (k<section_text_end) k++;
  if (xisspace(c)) {
    c=' '; if (*(k-1)==' ') k--;
  }
*k=c;
}
if (k>=section_text_end) {
  printf("\n! Section name too long: ");
@.Section name too long@>
  term_write(section_text+1,25);
  printf("..."); mark_harmless;
}
if (*k==' ' && k>section_text) k--;

@ @<If end of name...@>=
if (c=='@@') {
  c=*(loc+1);
  if (c=='>') {
    loc+=2; break;
  }
  if (ccode[(eight_bits)c]==new_section) {
    err_print("! Section name didn't end"); break;
@.Section name didn't end@>
  }
  if (c!='@@') {
    err_print("! Control codes are forbidden in section name"); break;
@.Control codes are forbidden...@>
  }
  *(++k)='@@'; loc++; /* now |c==*loc| again */
}

@ This function skips over a restricted context at relatively high speed.

@<Predecl...@>=
void skip_restricted();

@ @c
void
skip_restricted()
{
  id_first=loc; *(limit+1)='@@';
false_alarm:
  while (*loc!='@@') loc++;
  id_loc=loc;
  if (loc++>limit) {
    err_print("! Control text didn't end"); loc=limit;
@.Control text didn't end@>
  }
  else {
    if (*loc=='@@'&&loc<=limit) {loc++; goto false_alarm;}
    if (*loc++!='>')
      err_print("! Control codes are forbidden in control text");
@.Control codes are forbidden...@>
  }
}

@ At the present point in the program we
have |*(loc-1)==verbatim|; we set |id_first| to the beginning
of the string itself, and |id_loc| to its ending-plus-one location in the
buffer.  We also set |loc| to the position just after the ending delimiter.

@<Scan a verbatim string@>= {
  id_first=loc++; *(limit+1)='@@'; *(limit+2)='>';
  while (*loc!='@@' || *(loc+1)!='>') loc++;
  if (loc>=limit) err_print("! Verbatim string didn't end");
@.Verbatim string didn't end@>
  id_loc=loc; loc+=2;
  return (verbatim);
}

@** Phase one processing.
We now have accumulated enough subroutines to make it possible to carry out
\.{CWEAVE}'s first pass over the source file. If everything works right,
both phase one and phase two of \.{CWEAVE} will assign the same numbers to
sections, and these numbers will agree with what \.{CTANGLE} does.

The global variable |next_control| often contains the most recent output of
|get_next|; in interesting cases, this will be the control code that
ended a section or part of a section.

@<Global...@>=
eight_bits next_control; /* control code waiting to be acting upon */

@ The overall processing strategy in phase one has the following
straightforward outline.

@<Predecl...@>=
void phase_one();

@ @c
void
phase_one() {
  phase=1; reset_input(); section_count=0;
  skip_limbo(); change_exists=0;
  while (!input_has_ended)
    @<Store cross-reference data for the current section@>;
  changed_section[section_count]=change_exists;
    /* the index changes if anything does */
  phase=2; /* prepare for second phase */
  @<Print error messages about unused or undefined section names@>;
}

@ @<Store cross-reference data...@>=
{
  if (++section_count==max_sections) overflow("section number");
  changed_section[section_count]=changing;
     /* it will become 1 if any line changes */
  if (*(loc-1)=='*' && show_progress) {
    printf("*%d",section_count);
    update_terminal; /* print a progress report */
  }
  @<Store cross-references in the \TEX/ part of a section@>;
  @<Store cross-references in the definition part of a section@>;
  @<Store cross-references in the \CEE/ part of a section@>;
  if (changed_section[section_count]) change_exists=1;
}

@ The |C_xref| subroutine stores references to identifiers in
\CEE/ text material beginning with the current value of |next_control|
and continuing until |next_control| is `\.\{' or `\.{\v}', or until the next
``milestone'' is passed (i.e., |next_control>=format_code|). If
|next_control>=format_code| when |C_xref| is called, nothing will happen;
but if |next_control=='|'| upon entry, the procedure assumes that this is
the `\.{\v}' preceding \CEE/ text that is to be processed.

The parameter |spec_ctrl| is used to change this behavior. In most cases
|C_xref| is called with |spec_ctrl==ignore|, which triggers the default
processing described above. If |spec_ctrl==section_name|, section names will
be gobbled. This is used when \CEE/ text in the \TEX/ part or inside comments
is parsed: It allows for section names to appear in \pb, but these