syn.c

这是一个同样来自贝尔实验室的和UNIX有着渊源的操作系统, 其简洁的设计和实现易于我们学习和理解

/*
 * shell parser (C version)
 */

#include "sh.h"
#include "c_test.h"

struct nesting_state {
	int	start_token;	/* token than began nesting (eg, FOR) */
	int	start_line;	/* line nesting began on */
};

static void	yyparse		ARGS((void));
static struct op *pipeline	ARGS((int cf));
static struct op *andor		ARGS((void));
static struct op *c_list	ARGS((int multi));
static struct ioword *synio	ARGS((int cf));
static void	musthave	ARGS((int c, int cf));
static struct op *nested	ARGS((int type, int smark, int emark));
static struct op *get_command	ARGS((int cf));
static struct op *dogroup	ARGS((void));
static struct op *thenpart	ARGS((void));
static struct op *elsepart	ARGS((void));
static struct op *caselist	ARGS((void));
static struct op *casepart	ARGS((int endtok));
static struct op *function_body	ARGS((char *name, int ksh_func));
static char **	wordlist	ARGS((void));
static struct op *block		ARGS((int type, struct op *t1, struct op *t2,
				      char **wp));
static struct op *newtp		ARGS((int type));
static void	syntaxerr	ARGS((const char *what))
						GCC_FUNC_ATTR(noreturn);
static void	nesting_push ARGS((struct nesting_state *save, int tok));
static void	nesting_pop ARGS((struct nesting_state *saved));
static int	assign_command ARGS((char *s));
static int	inalias ARGS((struct source *s));
#ifdef KSH
static int	dbtestp_isa ARGS((Test_env *te, Test_meta meta));
static const char *dbtestp_getopnd ARGS((Test_env *te, Test_op op,
					int do_eval));
static int	dbtestp_eval ARGS((Test_env *te, Test_op op, const char *opnd1,
				const char *opnd2, int do_eval));
static void	dbtestp_error ARGS((Test_env *te, int offset, const char *msg));
#endif /* KSH */

static	struct	op	*outtree; /* yyparse output */

static struct nesting_state nesting;	/* \n changed to ; */

static	int	reject;		/* token(cf) gets symbol again */
static	int	symbol;		/* yylex value */

#define	REJECT	(reject = 1)
#define	ACCEPT	(reject = 0)
#define	token(cf) \
	((reject) ? (ACCEPT, symbol) : (symbol = yylex(cf)))
#define	tpeek(cf) \
	((reject) ? (symbol) : (REJECT, symbol = yylex(cf)))

static void
yyparse()
{
	int c;

	ACCEPT;

	outtree = c_list(source->type == SSTRING);
	c = tpeek(0);
	if (c == 0 && !outtree)
		outtree = newtp(TEOF);
	else if (c != '\n' && c != 0)
		syntaxerr((char *) 0);
}

static struct op *
pipeline(cf)
	int cf;
{
	register struct op *t, *p, *tl = NULL;

	t = get_command(cf);
	if (t != NULL) {
		while (token(0) == '|') {
			if ((p = get_command(CONTIN)) == NULL)
				syntaxerr((char *) 0);
			if (tl == NULL)
				t = tl = block(TPIPE, t, p, NOWORDS);
			else
				tl = tl->right = block(TPIPE, tl->right, p, NOWORDS);
		}
		REJECT;
	}
	return (t);
}

static struct op *
andor()
{
	register struct op *t, *p;
	register int c;

	t = pipeline(0);
	if (t != NULL) {
		while ((c = token(0)) == LOGAND || c == LOGOR) {
			if ((p = pipeline(CONTIN)) == NULL)
				syntaxerr((char *) 0);
			t = block(c == LOGAND? TAND: TOR, t, p, NOWORDS);
		}
		REJECT;
	}
	return (t);
}

static struct op *
c_list(multi)
	int multi;
{
	register struct op *t = NULL, *p, *tl = NULL;
	register int c;
	int have_sep;

	while (1) {
		p = andor();
		/* Token has always been read/rejected at this point, so
		 * we don't worry about what flags to pass token()
		 */
		c = token(0);
		have_sep = 1;
		if (c == '\n' && (multi || inalias(source))) {
			if (!p) /* ignore blank lines */
				continue;
		} else if (!p)
			break;
		else if (c == '&' || c == COPROC)
			p = block(c == '&' ? TASYNC : TCOPROC,
				  p, NOBLOCK, NOWORDS);
		else if (c != ';')
			have_sep = 0;
		if (!t)
			t = p;
		else if (!tl)
			t = tl = block(TLIST, t, p, NOWORDS);
		else
			tl = tl->right = block(TLIST, tl->right, p, NOWORDS);
		if (!have_sep)
			break;
	}
	REJECT;
	return t;
}

static struct ioword *
synio(cf)
	int cf;
{
	register struct ioword *iop;
	int ishere;

	if (tpeek(cf) != REDIR)
		return NULL;
	ACCEPT;
	iop = yylval.iop;
	ishere = (iop->flag&IOTYPE) == IOHERE;
	musthave(LWORD, ishere ? HEREDELIM : 0);
	if (ishere) {
		iop->delim = yylval.cp;
		if (*ident != 0) /* unquoted */
			iop->flag |= IOEVAL;
		if (herep >= &heres[HERES])
			yyerror("too many <<'s\n");
		*herep++ = iop;
	} else
		iop->name = yylval.cp;
	return iop;
}

static void
musthave(c, cf)
	int c, cf;
{
	if ((token(cf)) != c)
		syntaxerr((char *) 0);
}

static struct op *
nested(type, smark, emark)
	int type, smark, emark;
{
	register struct op *t;
	struct nesting_state old_nesting;

	nesting_push(&old_nesting, smark);
	t = c_list(TRUE);
	musthave(emark, KEYWORD|ALIAS);
	nesting_pop(&old_nesting);
	return (block(type, t, NOBLOCK, NOWORDS));
}

static struct op *
get_command(cf)
	int cf;
{
	register struct op *t;
	register int c, iopn = 0, syniocf;
	struct ioword *iop, **iops;
	XPtrV args, vars;
	struct nesting_state old_nesting;

	iops = (struct ioword **) alloc(sizeofN(struct ioword *, NUFILE+1),
					ATEMP);
	XPinit(args, 16);
	XPinit(vars, 16);

	syniocf = KEYWORD|ALIAS;
	switch (c = token(cf|KEYWORD|ALIAS|VARASN)) {
	  default:
		REJECT;
		afree((void*) iops, ATEMP);
		XPfree(args);
		XPfree(vars);
		return NULL; /* empty line */

	  case LWORD:
	  case REDIR:
		REJECT;
		syniocf &= ~(KEYWORD|ALIAS);
		t = newtp(TCOM);
		t->lineno = source->line;
		while (1) {
			cf = (t->u.evalflags ? ARRAYVAR : 0)
			     | (XPsize(args) == 0 ? ALIAS|VARASN : CMDWORD);
			switch (tpeek(cf)) {
			  case REDIR:
				if (iopn >= NUFILE)
					yyerror("too many redirections\n");
				iops[iopn++] = synio(cf);
				break;

			  case LWORD:
				ACCEPT;
				/* the iopn == 0 and XPsize(vars) == 0 are
				 * dubious but at&t ksh acts this way
				 */
				if (iopn == 0 && XPsize(vars) == 0
				    && XPsize(args) == 0
				    && assign_command(ident))
					t->u.evalflags = DOVACHECK;
				if ((XPsize(args) == 0 || Flag(FKEYWORD))
				    && is_wdvarassign(yylval.cp))
					XPput(vars, yylval.cp);
				else
					XPput(args, yylval.cp);
				break;

			  case '(':
				/* Check for "> foo (echo hi)", which at&t ksh
				 * allows (not POSIX, but not disallowed)
				 */
				afree(t, ATEMP);
				if (XPsize(args) == 0 && XPsize(vars) == 0) {
					ACCEPT;
					goto Subshell;
				}
				/* Must be a function */
				if (iopn != 0 || XPsize(args) != 1
				    || XPsize(vars) != 0)
					syntaxerr((char *) 0);
				ACCEPT;
				/*(*/
				musthave(')', 0);
				t = function_body(XPptrv(args)[0], FALSE);
				goto Leave;

			  default:
				goto Leave;
			}
		}
	  Leave:
		break;

	  Subshell:
	  case '(':
		t = nested(TPAREN, '(', ')');
		break;

	  case '{': /*}*/
		t = nested(TBRACE, '{', '}');
		break;

#ifdef KSH
	  case MDPAREN:
	  {
		static const char let_cmd[] = { CHAR, 'l', CHAR, 'e',
						CHAR, 't', EOS };
		/* Leave KEYWORD in syniocf (allow if (( 1 )) then ...) */
		t = newtp(TCOM);
		t->lineno = source->line;
		ACCEPT;
		XPput(args, wdcopy(let_cmd, ATEMP));
		musthave(LWORD,LETEXPR);
		XPput(args, yylval.cp);
		break;
	  }
#endif /* KSH */

#ifdef KSH
	  case DBRACKET: /* [[ .. ]] */
		/* Leave KEYWORD in syniocf (allow if [[ -n 1 ]] then ...) */
		t = newtp(TDBRACKET);
		ACCEPT;
		{
			Test_env te;

			te.flags = TEF_DBRACKET;
			te.pos.av = &args;
			te.isa = dbtestp_isa;
			te.getopnd = dbtestp_getopnd;
			te.eval = dbtestp_eval;
			te.error = dbtestp_error;

			test_parse(&te);
		}
		break;
#endif /* KSH */

	  case FOR:
	  case SELECT:
		t = newtp((c == FOR) ? TFOR : TSELECT);
		musthave(LWORD, ARRAYVAR);
		if (!is_wdvarname(yylval.cp, TRUE))
			yyerror("%s: bad identifier\n",
				c == FOR ? "for" : "select");
		t->str = str_save(ident, ATEMP);
		nesting_push(&old_nesting, c);
		t->vars = wordlist();
		t->left = dogroup();
		nesting_pop(&old_nesting);
		break;

	  case WHILE:
	  case UNTIL:
		nesting_push(&old_nesting, c);
		t = newtp((c == WHILE) ? TWHILE : TUNTIL);
		t->left = c_list(TRUE);
		t->right = dogroup();
		nesting_pop(&old_nesting);
		break;

	  case CASE:
		t = newtp(TCASE);
		musthave(LWORD, 0);
		t->str = yylval.cp;
		nesting_push(&old_nesting, c);
		t->left = caselist();
		nesting_pop(&old_nesting);
		break;

	  case IF:
		nesting_push(&old_nesting, c);
		t = newtp(TIF);
		t->left = c_list(TRUE);
		t->right = thenpart();
		musthave(FI, KEYWORD|ALIAS);
		nesting_pop(&old_nesting);
		break;

	  case BANG:
		syniocf &= ~(KEYWORD|ALIAS);
		t = pipeline(0);
		if (t == (struct op *) 0)
			syntaxerr((char *) 0);
		t = block(TBANG, NOBLOCK, t, NOWORDS);
		break;

	  case TIME:
		syniocf &= ~(KEYWORD|ALIAS);
		t = pipeline(0);
		t = block(TTIME, t, NOBLOCK, NOWORDS);
		break;

	  case FUNCTION:
		musthave(LWORD, 0);
		t = function_body(yylval.cp, TRUE);
		break;
	}

	while ((iop = synio(syniocf)) != NULL) {
		if (iopn >= NUFILE)
			yyerror("too many redirections\n");
		iops[iopn++] = iop;
	}

	if (iopn == 0) {
		afree((void*) iops, ATEMP);
		t->ioact = NULL;
	} else {
		iops[iopn++] = NULL;
		iops = (struct ioword **) aresize((void*) iops,
					sizeofN(struct ioword *, iopn), ATEMP);
		t->ioact = iops;
	}

	if (t->type == TCOM || t->type == TDBRACKET) {
		XPput(args, NULL);
		t->args = (char **) XPclose(args);
		XPput(vars, NULL);
		t->vars = (char **) XPclose(vars);
	} else {
		XPfree(args);
		XPfree(vars);
	}

	return t;
}

static struct op *
dogroup()
{
	register int c;
	register struct op *list;

	c = token(CONTIN|KEYWORD|ALIAS);
	/* A {...} can be used instead of do...done for for/select loops
	 * but not for while/until loops - we don't need to check if it
	 * is a while loop because it would have been parsed as part of
	 * the conditional command list...
	 */
	if (c == DO)
		c = DONE;
	else if (c == '{')
		c = '}';
	else
		syntaxerr((char *) 0);
	list = c_list(TRUE);
	musthave(c, KEYWORD|ALIAS);
	return list;
}

static struct op *
thenpart()
{
	register struct op *t;

	musthave(THEN, KEYWORD|ALIAS);
	t = newtp(0);
	t->left = c_list(TRUE);
	if (t->left == NULL)
		syntaxerr((char *) 0);
	t->right = elsepart();
	return (t);
}

static struct op *
elsepart()
{
	register struct op *t;

	switch (token(KEYWORD|ALIAS|VARASN)) {
	  case ELSE:
		if ((t = c_list(TRUE)) == NULL)
			syntaxerr((char *) 0);
		return (t);

	  case ELIF:
		t = newtp(TELIF);
		t->left = c_list(TRUE);
		t->right = thenpart();
		return (t);

	  default:
		REJECT;
	}
	return NULL;