freebsd-dev/gnu/usr.bin/groff/eqn/lex.cc
1995-01-15 08:10:00 +00:00

1161 lines
24 KiB
C++

// -*- C++ -*-
/* Copyright (C) 1989, 1990, 1991, 1992 Free Software Foundation, Inc.
Written by James Clark (jjc@jclark.com)
This file is part of groff.
groff 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 2, or (at your option) any later
version.
groff 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 groff; see the file COPYING. If not, write to the Free Software
Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. */
#include "eqn.h"
#include "eqn.tab.h"
#include "stringclass.h"
#include "ptable.h"
struct definition {
char is_macro;
char is_simple;
union {
int tok;
char *contents;
};
definition();
~definition();
};
definition::definition() : is_macro(1), is_simple(0)
{
contents = 0;
}
definition::~definition()
{
if (is_macro)
a_delete contents;
}
declare_ptable(definition)
implement_ptable(definition)
PTABLE(definition) macro_table;
static struct {
const char *name;
int token;
} token_table[] = {
{ "over", OVER },
{ "smallover", SMALLOVER },
{ "sqrt", SQRT },
{ "sub", SUB },
{ "sup", SUP },
{ "lpile", LPILE },
{ "rpile", RPILE },
{ "cpile", CPILE },
{ "pile", PILE },
{ "left", LEFT },
{ "right", RIGHT },
{ "to", TO },
{ "from", FROM },
{ "size", SIZE },
{ "font", FONT },
{ "roman", ROMAN },
{ "bold", BOLD },
{ "italic", ITALIC },
{ "fat", FAT },
{ "bar", BAR },
{ "under", UNDER },
{ "accent", ACCENT },
{ "uaccent", UACCENT },
{ "above", ABOVE },
{ "fwd", FWD },
{ "back", BACK },
{ "down", DOWN },
{ "up", UP },
{ "matrix", MATRIX },
{ "col", COL },
{ "lcol", LCOL },
{ "rcol", RCOL },
{ "ccol", CCOL },
{ "mark", MARK },
{ "lineup", LINEUP },
{ "space", SPACE },
{ "gfont", GFONT },
{ "gsize", GSIZE },
{ "define", DEFINE },
{ "sdefine", SDEFINE },
{ "ndefine", NDEFINE },
{ "tdefine", TDEFINE },
{ "undef", UNDEF },
{ "ifdef", IFDEF },
{ "include", INCLUDE },
{ "copy", INCLUDE },
{ "delim", DELIM },
{ "chartype", CHARTYPE },
{ "type", TYPE },
{ "vcenter", VCENTER },
{ "set", SET },
{ "opprime", PRIME },
{ "grfont", GRFONT },
{ "gbfont", GBFONT },
{ "split", SPLIT },
{ "nosplit", NOSPLIT },
{ "special", SPECIAL },
};
static struct {
const char *name;
const char *def;
} def_table[] = {
{ "ALPHA", "\\(*A" },
{ "BETA", "\\(*B" },
{ "CHI", "\\(*X" },
{ "DELTA", "\\(*D" },
{ "EPSILON", "\\(*E" },
{ "ETA", "\\(*Y" },
{ "GAMMA", "\\(*G" },
{ "IOTA", "\\(*I" },
{ "KAPPA", "\\(*K" },
{ "LAMBDA", "\\(*L" },
{ "MU", "\\(*M" },
{ "NU", "\\(*N" },
{ "OMEGA", "\\(*W" },
{ "OMICRON", "\\(*O" },
{ "PHI", "\\(*F" },
{ "PI", "\\(*P" },
{ "PSI", "\\(*Q" },
{ "RHO", "\\(*R" },
{ "SIGMA", "\\(*S" },
{ "TAU", "\\(*T" },
{ "THETA", "\\(*H" },
{ "UPSILON", "\\(*U" },
{ "XI", "\\(*C" },
{ "ZETA", "\\(*Z" },
{ "Alpha", "\\(*A" },
{ "Beta", "\\(*B" },
{ "Chi", "\\(*X" },
{ "Delta", "\\(*D" },
{ "Epsilon", "\\(*E" },
{ "Eta", "\\(*Y" },
{ "Gamma", "\\(*G" },
{ "Iota", "\\(*I" },
{ "Kappa", "\\(*K" },
{ "Lambda", "\\(*L" },
{ "Mu", "\\(*M" },
{ "Nu", "\\(*N" },
{ "Omega", "\\(*W" },
{ "Omicron", "\\(*O" },
{ "Phi", "\\(*F" },
{ "Pi", "\\(*P" },
{ "Psi", "\\(*Q" },
{ "Rho", "\\(*R" },
{ "Sigma", "\\(*S" },
{ "Tau", "\\(*T" },
{ "Theta", "\\(*H" },
{ "Upsilon", "\\(*U" },
{ "Xi", "\\(*C" },
{ "Zeta", "\\(*Z" },
{ "alpha", "\\(*a" },
{ "beta", "\\(*b" },
{ "chi", "\\(*x" },
{ "delta", "\\(*d" },
{ "epsilon", "\\(*e" },
{ "eta", "\\(*y" },
{ "gamma", "\\(*g" },
{ "iota", "\\(*i" },
{ "kappa", "\\(*k" },
{ "lambda", "\\(*l" },
{ "mu", "\\(*m" },
{ "nu", "\\(*n" },
{ "omega", "\\(*w" },
{ "omicron", "\\(*o" },
{ "phi", "\\(*f" },
{ "pi", "\\(*p" },
{ "psi", "\\(*q" },
{ "rho", "\\(*r" },
{ "sigma", "\\(*s" },
{ "tau", "\\(*t" },
{ "theta", "\\(*h" },
{ "upsilon", "\\(*u" },
{ "xi", "\\(*c" },
{ "zeta", "\\(*z" },
{ "max", "{type \"operator\" roman \"max\"}" },
{ "min", "{type \"operator\" roman \"min\"}" },
{ "lim", "{type \"operator\" roman \"lim\"}" },
{ "sin", "{type \"operator\" roman \"sin\"}" },
{ "cos", "{type \"operator\" roman \"cos\"}" },
{ "tan", "{type \"operator\" roman \"tan\"}" },
{ "sinh", "{type \"operator\" roman \"sinh\"}" },
{ "cosh", "{type \"operator\" roman \"cosh\"}" },
{ "tanh", "{type \"operator\" roman \"tanh\"}" },
{ "arc", "{type \"operator\" roman \"arc\"}" },
{ "log", "{type \"operator\" roman \"log\"}" },
{ "ln", "{type \"operator\" roman \"ln\"}" },
{ "exp", "{type \"operator\" roman \"exp\"}" },
{ "Re", "{type \"operator\" roman \"Re\"}" },
{ "Im", "{type \"operator\" roman \"Im\"}" },
{ "det", "{type \"operator\" roman \"det\"}" },
{ "and", "{roman \"and\"}" },
{ "if", "{roman \"if\"}" },
{ "for", "{roman \"for\"}" },
{ "sum", "{type \"operator\" vcenter size +5 \\(*S}" },
{ "prod", "{type \"operator\" vcenter size +5 \\(*P}" },
{ "int", "{type \"operator\" vcenter size +8 \\(is}" },
{ "union", "{type \"operator\" vcenter size +5 \\(cu}" },
{ "inter", "{type \"operator\" vcenter size +5 \\(ca}" },
{ "times", "type \"binary\" \\(mu" },
{ "ldots", "type \"inner\" { . . . }" },
{ "inf", "\\(if" },
{ "partial", "\\(pd" },
{ "nothing", "\"\"" },
{ "half", "{1 smallover 2}" },
{ "hat_def", "roman \"^\"" },
{ "hat", "accent { hat_def }" },
{ "dot_def", "back 15 \"\\v'-52M'.\\v'52M'\"" },
{ "dot", "accent { dot_def }" },
{ "dotdot_def", "back 25 \"\\v'-52M'..\\v'52M'\"" },
{ "dotdot", "accent { dotdot_def }" },
{ "tilde_def", "\"~\"" },
{ "tilde", "accent { tilde_def }" },
{ "utilde_def", "\"\\v'75M'~\\v'-75M'\"" },
{ "utilde", "uaccent { utilde_def }" },
{ "vec_def", "up 52 size -5 \\(->" },
{ "vec", "accent { vec_def }" },
{ "dyad_def", "up 52 size -5 {\\(<- back 60 \\(->}" },
{ "dyad", "accent { dyad_def }" },
{ "==", "type \"relation\" \\(==" },
{ "!=", "type \"relation\" \\(!=" },
{ "+-", "type \"binary\" \\(+-" },
{ "->", "type \"relation\" \\(->" },
{ "<-", "type \"relation\" \\(<-" },
{ "<<", "{ < back 20 < }" },
{ ">>", "{ > back 20 > }" },
{ "...", "type \"inner\" vcenter { . . . }" },
{ "prime", "'" },
{ "approx", "type \"relation\" \"\\(~=\"" },
{ "grad", "\\(gr" },
{ "del", "\\(gr" },
{ "cdot", "type \"binary\" vcenter ." },
{ "dollar", "$" },
};
void init_table(const char *device)
{
for (int i = 0; i < sizeof(token_table)/sizeof(token_table[0]); i++) {
definition *def = new definition;
def->is_macro = 0;
def->tok = token_table[i].token;
macro_table.define(token_table[i].name, def);
}
for (i = 0; i < sizeof(def_table)/sizeof(def_table[0]); i++) {
definition *def = new definition;
def->is_macro = 1;
def->contents = strsave(def_table[i].def);
def->is_simple = 1;
macro_table.define(def_table[i].name, def);
}
definition *def = new definition;
def->is_macro = 1;
def->contents = strsave("1");
macro_table.define(device, def);
}
class input {
input *next;
public:
input(input *p);
virtual ~input();
virtual int get() = 0;
virtual int peek() = 0;
virtual int get_location(char **, int *);
friend int get_char();
friend int peek_char();
friend int get_location(char **, int *);
friend void init_lex(const char *str, const char *filename, int lineno);
};
class file_input : public input {
FILE *fp;
char *filename;
int lineno;
string line;
const char *ptr;
int read_line();
public:
file_input(FILE *, const char *, input *);
~file_input();
int get();
int peek();
int get_location(char **, int *);
};
class macro_input : public input {
char *s;
char *p;
public:
macro_input(const char *, input *);
~macro_input();
int get();
int peek();
};
class top_input : public macro_input {
char *filename;
int lineno;
public:
top_input(const char *, const char *, int, input *);
~top_input();
int get();
int get_location(char **, int *);
};
class argument_macro_input: public input {
char *s;
char *p;
char *ap;
int argc;
char *argv[9];
public:
argument_macro_input(const char *, int, char **, input *);
~argument_macro_input();
int get();
int peek();
};
input::input(input *x) : next(x)
{
}
input::~input()
{
}
int input::get_location(char **, int *)
{
return 0;
}
file_input::file_input(FILE *f, const char *fn, input *p)
: input(p), lineno(0), ptr("")
{
fp = f;
filename = strsave(fn);
}
file_input::~file_input()
{
a_delete filename;
fclose(fp);
}
int file_input::read_line()
{
for (;;) {
line.clear();
lineno++;
for (;;) {
int c = getc(fp);
if (c == EOF)
break;
else if (illegal_input_char(c))
lex_error("illegal input character code %1", c);
else {
line += char(c);
if (c == '\n')
break;
}
}
if (line.length() == 0)
return 0;
if (!(line.length() >= 3 && line[0] == '.' && line[1] == 'E'
&& (line[2] == 'Q' || line[2] == 'N')
&& (line.length() == 3 || line[3] == ' ' || line[3] == '\n'
|| compatible_flag))) {
line += '\0';
ptr = line.contents();
return 1;
}
}
}
int file_input::get()
{
if (*ptr != '\0' || read_line())
return *ptr++ & 0377;
else
return EOF;
}
int file_input::peek()
{
if (*ptr != '\0' || read_line())
return *ptr;
else
return EOF;
}
int file_input::get_location(char **fnp, int *lnp)
{
*fnp = filename;
*lnp = lineno;
return 1;
}
macro_input::macro_input(const char *str, input *x) : input(x)
{
p = s = strsave(str);
}
macro_input::~macro_input()
{
a_delete s;
}
int macro_input::get()
{
if (p == 0 || *p == '\0')
return EOF;
else
return *p++ & 0377;
}
int macro_input::peek()
{
if (p == 0 || *p == '\0')
return EOF;
else
return *p & 0377;
}
top_input::top_input(const char *str, const char *fn, int ln, input *x)
: macro_input(str, x), lineno(ln)
{
filename = strsave(fn);
}
top_input::~top_input()
{
a_delete filename;
}
int top_input::get()
{
int c = macro_input::get();
if (c == '\n')
lineno++;
return c;
}
int top_input::get_location(char **fnp, int *lnp)
{
*fnp = filename;
*lnp = lineno;
return 1;
}
// Character representing $1. Must be illegal input character.
#define ARG1 14
argument_macro_input::argument_macro_input(const char *body, int ac,
char **av, input *x)
: input(x), argc(ac), ap(0)
{
for (int i = 0; i < argc; i++)
argv[i] = av[i];
p = s = strsave(body);
int j = 0;
for (i = 0; s[i] != '\0'; i++)
if (s[i] == '$' && s[i+1] >= '0' && s[i+1] <= '9') {
if (s[i+1] != '0')
s[j++] = ARG1 + s[++i] - '1';
}
else
s[j++] = s[i];
s[j] = '\0';
}
argument_macro_input::~argument_macro_input()
{
for (int i = 0; i < argc; i++)
a_delete argv[i];
a_delete s;
}
int argument_macro_input::get()
{
if (ap) {
if (*ap != '\0')
return *ap++ & 0377;
ap = 0;
}
if (p == 0)
return EOF;
while (*p >= ARG1 && *p <= ARG1 + 8) {
int i = *p++ - ARG1;
if (i < argc && argv[i] != 0 && argv[i][0] != '\0') {
ap = argv[i];
return *ap++ & 0377;
}
}
if (*p == '\0')
return EOF;
return *p++ & 0377;
}
int argument_macro_input::peek()
{
if (ap) {
if (*ap != '\0')
return *ap & 0377;
ap = 0;
}
if (p == 0)
return EOF;
while (*p >= ARG1 && *p <= ARG1 + 8) {
int i = *p++ - ARG1;
if (i < argc && argv[i] != 0 && argv[i][0] != '\0') {
ap = argv[i];
return *ap & 0377;
}
}
if (*p == '\0')
return EOF;
return *p & 0377;
}
static input *current_input = 0;
/* we insert a newline between input from different levels */
int get_char()
{
if (current_input == 0)
return EOF;
else {
int c = current_input->get();
if (c != EOF)
return c;
else {
input *tem = current_input;
current_input = current_input->next;
delete tem;
return '\n';
}
}
}
int peek_char()
{
if (current_input == 0)
return EOF;
else {
int c = current_input->peek();
if (c != EOF)
return c;
else
return '\n';
}
}
int get_location(char **fnp, int *lnp)
{
for (input *p = current_input; p; p = p->next)
if (p->get_location(fnp, lnp))
return 1;
return 0;
}
string token_buffer;
const int NCONTEXT = 4;
string context_ring[NCONTEXT];
int context_index = 0;
void flush_context()
{
for (int i = 0; i < NCONTEXT; i++)
context_ring[i] = "";
context_index = 0;
}
void show_context()
{
int i = context_index;
fputs(" context is\n\t", stderr);
for (;;) {
int j = (i + 1) % NCONTEXT;
if (j == context_index) {
fputs(">>> ", stderr);
put_string(context_ring[i], stderr);
fputs(" <<<", stderr);
break;
}
else if (context_ring[i].length() > 0) {
put_string(context_ring[i], stderr);
putc(' ', stderr);
}
i = j;
}
putc('\n', stderr);
}
void add_context(const string &s)
{
context_ring[context_index] = s;
context_index = (context_index + 1) % NCONTEXT;
}
void add_context(char c)
{
context_ring[context_index] = c;
context_index = (context_index + 1) % NCONTEXT;
}
void add_quoted_context(const string &s)
{
string &r = context_ring[context_index];
r = '"';
for (int i = 0; i < s.length(); i++)
if (s[i] == '"')
r += "\\\"";
else
r += s[i];
r += '"';
context_index = (context_index + 1) % NCONTEXT;
}
void init_lex(const char *str, const char *filename, int lineno)
{
while (current_input != 0) {
input *tem = current_input;
current_input = current_input->next;
delete tem;
}
current_input = new top_input(str, filename, lineno, 0);
flush_context();
}
void get_delimited_text()
{
char *filename;
int lineno;
int got_location = get_location(&filename, &lineno);
int start = get_char();
while (start == ' ' || start == '\t' || start == '\n')
start = get_char();
token_buffer.clear();
if (start == EOF) {
if (got_location)
error_with_file_and_line(filename, lineno,
"end of input while defining macro");
else
error("end of input while defining macro");
return;
}
for (;;) {
int c = get_char();
if (c == EOF) {
if (got_location)
error_with_file_and_line(filename, lineno,
"end of input while defining macro");
else
error("end of input while defining macro");
add_context(start + token_buffer);
return;
}
if (c == start)
break;
token_buffer += char(c);
}
add_context(start + token_buffer + start);
}
void interpolate_macro_with_args(const char *body)
{
char *argv[9];
int argc = 0;
for (int i = 0; i < 9; i++)
argv[i] = 0;
int level = 0;
int c;
do {
token_buffer.clear();
for (;;) {
c = get_char();
if (c == EOF) {
lex_error("end of input while scanning macro arguments");
break;
}
if (level == 0 && (c == ',' || c == ')')) {
if (token_buffer.length() > 0) {
token_buffer += '\0';
argv[argc] = strsave(token_buffer.contents());
}
// for `foo()', argc = 0
if (argc > 0 || c != ')' || i > 0)
argc++;
break;
}
token_buffer += char(c);
if (c == '(')
level++;
else if (c == ')')
level--;
}
} while (c != ')' && c != EOF);
current_input = new argument_macro_input(body, argc, argv, current_input);
}
/* If lookup flag is non-zero the token will be looked up to see
if it is macro. If it's 1, it will looked up to see if it's a token.
*/
int get_token(int lookup_flag = 0)
{
for (;;) {
int c = get_char();
while (c == ' ' || c == '\n')
c = get_char();
switch (c) {
case EOF:
add_context("end of input");
return 0;
case '"':
{
int quoted = 0;
token_buffer.clear();
for (;;) {
c = get_char();
if (c == EOF) {
lex_error("missing \"");
break;
}
else if (c == '\n') {
lex_error("newline before end of quoted text");
break;
}
else if (c == '"') {
if (!quoted)
break;
token_buffer[token_buffer.length() - 1] = '"';
quoted = 0;
}
else {
token_buffer += c;
quoted = quoted ? 0 : c == '\\';
}
}
}
add_quoted_context(token_buffer);
return QUOTED_TEXT;
case '{':
case '}':
case '^':
case '~':
case '\t':
add_context(c);
return c;
default:
{
int break_flag = 0;
int quoted = 0;
token_buffer.clear();
if (c == '\\')
quoted = 1;
else
token_buffer += c;
int done = 0;
while (!done) {
c = peek_char();
if (!quoted && lookup_flag != 0 && c == '(') {
token_buffer += '\0';
definition *def = macro_table.lookup(token_buffer.contents());
if (def && def->is_macro && !def->is_simple) {
(void)get_char(); // skip initial '('
interpolate_macro_with_args(def->contents);
break_flag = 1;
break;
}
token_buffer.set_length(token_buffer.length() - 1);
}
if (quoted) {
quoted = 0;
switch (c) {
case EOF:
lex_error("`\\' ignored at end of equation");
done = 1;
break;
case '\n':
lex_error("`\\' ignored because followed by newline");
done = 1;
break;
case '\t':
lex_error("`\\' ignored because followed by tab");
done = 1;
break;
case '"':
(void)get_char();
token_buffer += '"';
break;
default:
(void)get_char();
token_buffer += '\\';
token_buffer += c;
break;
}
}
else {
switch (c) {
case EOF:
case '{':
case '}':
case '^':
case '~':
case '"':
case ' ':
case '\t':
case '\n':
done = 1;
break;
case '\\':
(void)get_char();
quoted = 1;
break;
default:
(void)get_char();
token_buffer += char(c);
break;
}
}
}
if (break_flag || token_buffer.length() == 0)
break;
if (lookup_flag != 0) {
token_buffer += '\0';
definition *def = macro_table.lookup(token_buffer.contents());
token_buffer.set_length(token_buffer.length() - 1);
if (def) {
if (def->is_macro) {
current_input = new macro_input(def->contents, current_input);
break;
}
else if (lookup_flag == 1) {
add_context(token_buffer);
return def->tok;
}
}
}
add_context(token_buffer);
return TEXT;
}
}
}
}
void do_include()
{
int t = get_token(2);
if (t != TEXT && t != QUOTED_TEXT) {
lex_error("bad filename for include");
return;
}
token_buffer += '\0';
const char *filename = token_buffer.contents();
errno = 0;
FILE *fp = fopen(filename, "r");
if (fp == 0) {
lex_error("can't open included file `%1'", filename);
return;
}
current_input = new file_input(fp, filename, current_input);
}
void ignore_definition()
{
int t = get_token();
if (t != TEXT) {
lex_error("bad definition");
return;
}
get_delimited_text();
}
void do_definition(int is_simple)
{
int t = get_token();
if (t != TEXT) {
lex_error("bad definition");
return;
}
token_buffer += '\0';
const char *name = token_buffer.contents();
definition *def = macro_table.lookup(name);
if (def == 0) {
def = new definition;
macro_table.define(name, def);
}
else if (def->is_macro) {
a_delete def->contents;
}
get_delimited_text();
token_buffer += '\0';
def->is_macro = 1;
def->contents = strsave(token_buffer.contents());
def->is_simple = is_simple;
}
void do_undef()
{
int t = get_token();
if (t != TEXT) {
lex_error("bad undef command");
return;
}
token_buffer += '\0';
macro_table.define(token_buffer.contents(), 0);
}
void do_gsize()
{
int t = get_token(2);
if (t != TEXT && t != QUOTED_TEXT) {
lex_error("bad argument to gsize command");
return;
}
token_buffer += '\0';
if (!set_gsize(token_buffer.contents()))
lex_error("invalid size `%1'", token_buffer.contents());
}
void do_gfont()
{
int t = get_token(2);
if (t != TEXT && t != QUOTED_TEXT) {
lex_error("bad argument to gfont command");
return;
}
token_buffer += '\0';
set_gfont(token_buffer.contents());
}
void do_grfont()
{
int t = get_token(2);
if (t != TEXT && t != QUOTED_TEXT) {
lex_error("bad argument to grfont command");
return;
}
token_buffer += '\0';
set_grfont(token_buffer.contents());
}
void do_gbfont()
{
int t = get_token(2);
if (t != TEXT && t != QUOTED_TEXT) {
lex_error("bad argument to gbfont command");
return;
}
token_buffer += '\0';
set_gbfont(token_buffer.contents());
}
void do_space()
{
int t = get_token(2);
if (t != TEXT && t != QUOTED_TEXT) {
lex_error("bad argument to space command");
return;
}
token_buffer += '\0';
char *ptr;
long n = strtol(token_buffer.contents(), &ptr, 10);
if (n == 0 && ptr == token_buffer.contents())
lex_error("bad argument `%1' to space command");
else
set_space(int(n));
}
void do_ifdef()
{
int t = get_token();
if (t != TEXT) {
lex_error("bad ifdef");
return;
}
token_buffer += '\0';
definition *def = macro_table.lookup(token_buffer.contents());
int result = def && def->is_macro && !def->is_simple;
get_delimited_text();
if (result) {
token_buffer += '\0';
current_input = new macro_input(token_buffer.contents(), current_input);
}
}
void do_delim()
{
int c = get_char();
while (c == ' ' || c == '\n')
c = get_char();
int d;
if (c == EOF || (d = get_char()) == EOF)
lex_error("end of file while reading argument to `delim'");
else {
if (c == 'o' && d == 'f' && peek_char() == 'f') {
(void)get_char();
start_delim = end_delim = '\0';
}
else {
start_delim = c;
end_delim = d;
}
}
}
void do_chartype()
{
int t = get_token(2);
if (t != TEXT && t != QUOTED_TEXT) {
lex_error("bad chartype");
return;
}
token_buffer += '\0';
string type = token_buffer;
t = get_token();
if (t != TEXT && t != QUOTED_TEXT) {
lex_error("bad chartype");
return;
}
token_buffer += '\0';
set_char_type(type.contents(), strsave(token_buffer.contents()));
}
void do_set()
{
int t = get_token(2);
if (t != TEXT && t != QUOTED_TEXT) {
lex_error("bad set");
return;
}
token_buffer += '\0';
string param = token_buffer;
t = get_token();
if (t != TEXT && t != QUOTED_TEXT) {
lex_error("bad set");
return;
}
token_buffer += '\0';
int n;
if (sscanf(&token_buffer[0], "%d", &n) != 1) {
lex_error("bad number `%1'", token_buffer.contents());
return;
}
set_param(param.contents(), n);
}
int yylex()
{
for (;;) {
int tk = get_token(1);
switch(tk) {
case UNDEF:
do_undef();
break;
case SDEFINE:
do_definition(1);
break;
case DEFINE:
do_definition(0);
break;
case TDEFINE:
if (!nroff)
do_definition(0);
else
ignore_definition();
break;
case NDEFINE:
if (nroff)
do_definition(0);
else
ignore_definition();
break;
case GSIZE:
do_gsize();
break;
case GFONT:
do_gfont();
break;
case GRFONT:
do_grfont();
break;
case GBFONT:
do_gbfont();
break;
case SPACE:
do_space();
break;
case INCLUDE:
do_include();
break;
case IFDEF:
do_ifdef();
break;
case DELIM:
do_delim();
break;
case CHARTYPE:
do_chartype();
break;
case SET:
do_set();
break;
case QUOTED_TEXT:
case TEXT:
token_buffer += '\0';
yylval.str = strsave(token_buffer.contents());
// fall through
default:
return tk;
}
}
}
void lex_error(const char *message,
const errarg &arg1,
const errarg &arg2,
const errarg &arg3)
{
char *filename;
int lineno;
if (!get_location(&filename, &lineno))
error(message, arg1, arg2, arg3);
else
error_with_file_and_line(filename, lineno, message, arg1, arg2, arg3);
}
void yyerror(const char *s)
{
char *filename;
int lineno;
if (!get_location(&filename, &lineno))
error(s);
else
error_with_file_and_line(filename, lineno, s);
show_context();
}