1 /***********************************************************
2 Copyright 1991-1995 by Stichting Mathematisch Centrum, Amsterdam,
7 Permission to use, copy, modify, and distribute this software and its
8 documentation for any purpose and without fee is hereby granted,
9 provided that the above copyright notice appear in all copies and that
10 both that copyright notice and this permission notice appear in
11 supporting documentation, and that the names of Stichting Mathematisch
12 Centrum or CWI or Corporation for National Research Initiatives or
13 CNRI not be used in advertising or publicity pertaining to
14 distribution of the software without specific, written prior
17 While CWI is the initial source for this software, a modified version
18 is made available by the Corporation for National Research Initiatives
19 (CNRI) at the Internet address ftp://ftp.python.org.
21 STICHTING MATHEMATISCH CENTRUM AND CNRI DISCLAIM ALL WARRANTIES WITH
22 REGARD TO THIS SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES OF
23 MERCHANTABILITY AND FITNESS, IN NO EVENT SHALL STICHTING MATHEMATISCH
24 CENTRUM OR CNRI BE LIABLE FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL
25 DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR
26 PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER
27 TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR
28 PERFORMANCE OF THIS SOFTWARE.
30 ******************************************************************/
32 /* Parser implementation */
34 /* For a description, see the comments at end of this file */
36 /* XXX To do: error recovery */
38 #include "pgenheaders.h"
48 extern int Py_DebugFlag
;
49 #define D(x) if (!Py_DebugFlag); else x
57 static void s_reset
Py_PROTO((stack
*));
63 s
->s_top
= &s
->s_base
[MAXSTACK
];
66 #define s_empty(s) ((s)->s_top == &(s)->s_base[MAXSTACK])
68 static int s_push
Py_PROTO((stack
*, dfa
*, node
*));
76 register stackentry
*top
;
77 if (s
->s_top
== s
->s_base
) {
78 fprintf(stderr
, "s_push: parser stack overflow\n");
83 top
->s_parent
= parent
;
90 static void s_pop
Py_PROTO((stack
*));
97 Py_FatalError("s_pop: parser stack underflow -- FATAL");
101 #else /* !Py_DEBUG */
103 #define s_pop(s) (s)->s_top++
108 /* PARSER CREATION */
111 PyParser_New(g
, start
)
118 PyGrammar_AddAccelerators(g
);
119 ps
= PyMem_NEW(parser_state
, 1);
123 ps
->p_tree
= PyNode_New(start
);
124 if (ps
->p_tree
== NULL
) {
128 s_reset(&ps
->p_stack
);
129 (void) s_push(&ps
->p_stack
, PyGrammar_FindDFA(g
, start
), ps
->p_tree
);
137 /* NB If you want to save the parse tree,
138 you must set p_tree to NULL before calling delparser! */
139 PyNode_Free(ps
->p_tree
);
144 /* PARSER STACK OPERATIONS */
146 static int shift
Py_PROTO((stack
*, int, char *, int, int));
149 shift(s
, type
, str
, newstate
, lineno
)
157 if (PyNode_AddChild(s
->s_top
->s_parent
, type
, str
, lineno
) == NULL
) {
158 fprintf(stderr
, "shift: no mem in addchild\n");
161 s
->s_top
->s_state
= newstate
;
165 static int push
Py_PROTO((stack
*, int, dfa
*, int, int));
168 push(s
, type
, d
, newstate
, lineno
)
176 n
= s
->s_top
->s_parent
;
178 if (PyNode_AddChild(n
, type
, (char *)NULL
, lineno
) == NULL
) {
179 fprintf(stderr
, "push: no mem in addchild\n");
182 s
->s_top
->s_state
= newstate
;
183 return s_push(s
, d
, CHILD(n
, NCH(n
)-1));
189 static int classify
Py_PROTO((grammar
*, int, char *));
192 classify(g
, type
, str
)
197 register int n
= g
->g_ll
.ll_nlabels
;
200 register char *s
= str
;
201 register label
*l
= g
->g_ll
.ll_label
;
203 for (i
= n
; i
> 0; i
--, l
++) {
204 if (l
->lb_type
== NAME
&& l
->lb_str
!= NULL
&&
205 l
->lb_str
[0] == s
[0] &&
206 strcmp(l
->lb_str
, s
) == 0) {
207 D(printf("It's a keyword\n"));
214 register label
*l
= g
->g_ll
.ll_label
;
216 for (i
= n
; i
> 0; i
--, l
++) {
217 if (l
->lb_type
== type
&& l
->lb_str
== NULL
) {
218 D(printf("It's a token we know\n"));
224 D(printf("Illegal token\n"));
229 PyParser_AddToken(ps
, type
, str
, lineno
)
230 register parser_state
*ps
;
237 D(printf("Token %s/'%s' ... ", _PyParser_TokenNames
[type
], str
));
239 /* Find out which label this token is */
240 ilabel
= classify(ps
->p_grammar
, type
, str
);
244 /* Loop until the token is shifted or an error occurred */
246 /* Fetch the current dfa and state */
247 register dfa
*d
= ps
->p_stack
.s_top
->s_dfa
;
248 register state
*s
= &d
->d_state
[ps
->p_stack
.s_top
->s_state
];
250 D(printf(" DFA '%s', state %d:",
251 d
->d_name
, ps
->p_stack
.s_top
->s_state
));
253 /* Check accelerator */
254 if (s
->s_lower
<= ilabel
&& ilabel
< s
->s_upper
) {
255 register int x
= s
->s_accel
[ilabel
- s
->s_lower
];
258 /* Push non-terminal */
259 int nt
= (x
>> 8) + NT_OFFSET
;
260 int arrow
= x
& ((1<<7)-1);
261 dfa
*d1
= PyGrammar_FindDFA(
263 if (push(&ps
->p_stack
, nt
, d1
,
264 arrow
, lineno
) < 0) {
265 D(printf(" MemError: push\n"));
268 D(printf(" Push ...\n"));
272 /* Shift the token */
273 if (shift(&ps
->p_stack
, type
, str
,
275 D(printf(" MemError: shift.\n"));
278 D(printf(" Shift.\n"));
279 /* Pop while we are in an accept-only state */
280 while (s
= &d
->d_state
281 [ps
->p_stack
.s_top
->s_state
],
282 s
->s_accept
&& s
->s_narcs
== 1) {
283 D(printf(" Direct pop.\n"));
285 if (s_empty(&ps
->p_stack
)) {
286 D(printf(" ACCEPT.\n"));
289 d
= ps
->p_stack
.s_top
->s_dfa
;
296 /* Pop this dfa and try again */
298 D(printf(" Pop ...\n"));
299 if (s_empty(&ps
->p_stack
)) {
300 D(printf(" Error: bottom of stack.\n"));
306 /* Stuck, report syntax error */
307 D(printf(" Error.\n"));
330 printf("%s", PyGrammar_LabelRepr(&l
));
331 if (ISNONTERMINAL(TYPE(n
))) {
333 for (i
= 0; i
< NCH(n
); i
++) {
336 dumptree(g
, CHILD(n
, i
));
352 if (ISNONTERMINAL(TYPE(n
))) {
353 for (i
= 0; i
< NCH(n
); i
++)
354 showtree(g
, CHILD(n
, i
));
356 else if (ISTERMINAL(TYPE(n
))) {
357 printf("%s", _PyParser_TokenNames
[TYPE(n
)]);
358 if (TYPE(n
) == NUMBER
|| TYPE(n
) == NAME
)
359 printf("(%s)", STR(n
));
371 printf("Parse tree:\n");
372 dumptree(ps
->p_grammar
, ps
->p_tree
);
375 showtree(ps
->p_grammar
, ps
->p_tree
);
378 printf("Listing:\n");
379 PyNode_ListTree(ps
->p_tree
);
383 #endif /* Py_DEBUG */
390 The parser's interface is different than usual: the function addtoken()
391 must be called for each token in the input. This makes it possible to
392 turn it into an incremental parsing system later. The parsing system
393 constructs a parse tree as it goes.
395 A parsing rule is represented as a Deterministic Finite-state Automaton
396 (DFA). A node in a DFA represents a state of the parser; an arc represents
397 a transition. Transitions are either labeled with terminal symbols or
398 with non-terminals. When the parser decides to follow an arc labeled
399 with a non-terminal, it is invoked recursively with the DFA representing
400 the parsing rule for that as its initial state; when that DFA accepts,
401 the parser that invoked it continues. The parse tree constructed by the
402 recursively called parser is inserted as a child in the current parse tree.
404 The DFA's can be constructed automatically from a more conventional
405 language description. An extended LL(1) grammar (ELL(1)) is suitable.
406 Certain restrictions make the parser's life easier: rules that can produce
407 the empty string should be outlawed (there are other ways to put loops
408 or optional parts in the language). To avoid the need to construct
409 FIRST sets, we can require that all but the last alternative of a rule
410 (really: arc going out of a DFA's state) must begin with a terminal
413 As an example, consider this grammar:
415 expr: term (OP term)*
416 term: CONSTANT | '(' expr ')'
418 The DFA corresponding to the rule for expr is:
420 ------->.---term-->.------->
425 The parse tree generated for the input a+b is:
427 (expr: (term: (NAME: a)), (OP: +), (term: (NAME: b)))