2
/* Parser implementation */
4
/* For a description, see the comments at end of this file */
6
/* XXX To do: error recovery */
9
#include "pgenheaders.h"
18
extern int Py_DebugFlag;
19
#define D(x) if (!Py_DebugFlag); else x
27
static void s_reset(stack *);
32
s->s_top = &s->s_base[MAXSTACK];
35
#define s_empty(s) ((s)->s_top == &(s)->s_base[MAXSTACK])
38
s_push(register stack *s, dfa *d, node *parent)
40
register stackentry *top;
41
if (s->s_top == s->s_base) {
42
fprintf(stderr, "s_push: parser stack overflow\n");
47
top->s_parent = parent;
55
s_pop(register stack *s)
58
Py_FatalError("s_pop: parser stack underflow -- FATAL");
64
#define s_pop(s) (s)->s_top++
72
PyParser_New(grammar *g, int start)
77
PyGrammar_AddAccelerators(g);
78
ps = (parser_state *)PyMem_MALLOC(sizeof(parser_state));
82
#ifdef PY_PARSER_REQUIRES_FUTURE_KEYWORD
85
ps->p_tree = PyNode_New(start);
86
if (ps->p_tree == NULL) {
90
s_reset(&ps->p_stack);
91
(void) s_push(&ps->p_stack, PyGrammar_FindDFA(g, start), ps->p_tree);
96
PyParser_Delete(parser_state *ps)
98
/* NB If you want to save the parse tree,
99
you must set p_tree to NULL before calling delparser! */
100
PyNode_Free(ps->p_tree);
105
/* PARSER STACK OPERATIONS */
108
shift(register stack *s, int type, char *str, int newstate, int lineno, int col_offset)
112
err = PyNode_AddChild(s->s_top->s_parent, type, str, lineno, col_offset);
115
s->s_top->s_state = newstate;
120
push(register stack *s, int type, dfa *d, int newstate, int lineno, int col_offset)
124
n = s->s_top->s_parent;
126
err = PyNode_AddChild(n, type, (char *)NULL, lineno, col_offset);
129
s->s_top->s_state = newstate;
130
return s_push(s, d, CHILD(n, NCH(n)-1));
137
classify(parser_state *ps, int type, char *str)
139
grammar *g = ps->p_grammar;
140
register int n = g->g_ll.ll_nlabels;
143
register char *s = str;
144
register label *l = g->g_ll.ll_label;
146
for (i = n; i > 0; i--, l++) {
147
if (l->lb_type != NAME || l->lb_str == NULL ||
148
l->lb_str[0] != s[0] ||
149
strcmp(l->lb_str, s) != 0)
151
#ifdef PY_PARSER_REQUIRES_FUTURE_KEYWORD
152
if (ps->p_flags & CO_FUTURE_PRINT_FUNCTION &&
153
s[0] == 'p' && strcmp(s, "print") == 0) {
154
break; /* no longer a keyword */
157
D(printf("It's a keyword\n"));
163
register label *l = g->g_ll.ll_label;
165
for (i = n; i > 0; i--, l++) {
166
if (l->lb_type == type && l->lb_str == NULL) {
167
D(printf("It's a token we know\n"));
173
D(printf("Illegal token\n"));
177
#ifdef PY_PARSER_REQUIRES_FUTURE_KEYWORD
179
future_hack(parser_state *ps)
181
node *n = ps->p_stack.s_top->s_parent;
185
/* from __future__ import ..., must have at least 4 children */
190
if (STR(ch) == NULL || strcmp(STR(ch), "from") != 0)
193
if (NCH(ch) == 1 && STR(CHILD(ch, 0)) &&
194
strcmp(STR(CHILD(ch, 0)), "__future__") != 0)
197
/* ch can be a star, a parenthesis or import_as_names */
198
if (TYPE(ch) == STAR)
200
if (TYPE(ch) == LPAR)
203
for (i = 0; i < NCH(ch); i += 2) {
205
if (NCH(cch) >= 1 && TYPE(CHILD(cch, 0)) == NAME) {
206
char *str_ch = STR(CHILD(cch, 0));
207
if (strcmp(str_ch, FUTURE_WITH_STATEMENT) == 0) {
208
ps->p_flags |= CO_FUTURE_WITH_STATEMENT;
209
} else if (strcmp(str_ch, FUTURE_PRINT_FUNCTION) == 0) {
210
ps->p_flags |= CO_FUTURE_PRINT_FUNCTION;
211
} else if (strcmp(str_ch, FUTURE_UNICODE_LITERALS) == 0) {
212
ps->p_flags |= CO_FUTURE_UNICODE_LITERALS;
217
#endif /* future keyword */
220
PyParser_AddToken(register parser_state *ps, register int type, char *str,
221
int lineno, int col_offset, int *expected_ret)
226
D(printf("Token %s/'%s' ... ", _PyParser_TokenNames[type], str));
228
/* Find out which label this token is */
229
ilabel = classify(ps, type, str);
233
/* Loop until the token is shifted or an error occurred */
235
/* Fetch the current dfa and state */
236
register dfa *d = ps->p_stack.s_top->s_dfa;
237
register state *s = &d->d_state[ps->p_stack.s_top->s_state];
239
D(printf(" DFA '%s', state %d:",
240
d->d_name, ps->p_stack.s_top->s_state));
242
/* Check accelerator */
243
if (s->s_lower <= ilabel && ilabel < s->s_upper) {
244
register int x = s->s_accel[ilabel - s->s_lower];
247
/* Push non-terminal */
248
int nt = (x >> 8) + NT_OFFSET;
249
int arrow = x & ((1<<7)-1);
250
dfa *d1 = PyGrammar_FindDFA(
252
if ((err = push(&ps->p_stack, nt, d1,
253
arrow, lineno, col_offset)) > 0) {
254
D(printf(" MemError: push\n"));
257
D(printf(" Push ...\n"));
261
/* Shift the token */
262
if ((err = shift(&ps->p_stack, type, str,
263
x, lineno, col_offset)) > 0) {
264
D(printf(" MemError: shift.\n"));
267
D(printf(" Shift.\n"));
268
/* Pop while we are in an accept-only state */
269
while (s = &d->d_state
270
[ps->p_stack.s_top->s_state],
271
s->s_accept && s->s_narcs == 1) {
272
D(printf(" DFA '%s', state %d: "
275
ps->p_stack.s_top->s_state));
276
#ifdef PY_PARSER_REQUIRES_FUTURE_KEYWORD
277
if (d->d_name[0] == 'i' &&
283
if (s_empty(&ps->p_stack)) {
284
D(printf(" ACCEPT.\n"));
287
d = ps->p_stack.s_top->s_dfa;
294
#ifdef PY_PARSER_REQUIRES_FUTURE_KEYWORD
295
if (d->d_name[0] == 'i' &&
296
strcmp(d->d_name, "import_stmt") == 0)
299
/* Pop this dfa and try again */
301
D(printf(" Pop ...\n"));
302
if (s_empty(&ps->p_stack)) {
303
D(printf(" Error: bottom of stack.\n"));
309
/* Stuck, report syntax error */
310
D(printf(" Error.\n"));
312
if (s->s_lower == s->s_upper - 1) {
313
/* Only one possible expected token */
314
*expected_ret = ps->p_grammar->
315
g_ll.ll_label[s->s_lower].lb_type;
330
dumptree(grammar *g, node *n)
340
printf("%s", PyGrammar_LabelRepr(&l));
341
if (ISNONTERMINAL(TYPE(n))) {
343
for (i = 0; i < NCH(n); i++) {
346
dumptree(g, CHILD(n, i));
354
showtree(grammar *g, node *n)
360
if (ISNONTERMINAL(TYPE(n))) {
361
for (i = 0; i < NCH(n); i++)
362
showtree(g, CHILD(n, i));
364
else if (ISTERMINAL(TYPE(n))) {
365
printf("%s", _PyParser_TokenNames[TYPE(n)]);
366
if (TYPE(n) == NUMBER || TYPE(n) == NAME)
367
printf("(%s)", STR(n));
375
printtree(parser_state *ps)
378
printf("Parse tree:\n");
379
dumptree(ps->p_grammar, ps->p_tree);
382
showtree(ps->p_grammar, ps->p_tree);
385
printf("Listing:\n");
386
PyNode_ListTree(ps->p_tree);
390
#endif /* Py_DEBUG */
397
The parser's interface is different than usual: the function addtoken()
398
must be called for each token in the input. This makes it possible to
399
turn it into an incremental parsing system later. The parsing system
400
constructs a parse tree as it goes.
402
A parsing rule is represented as a Deterministic Finite-state Automaton
403
(DFA). A node in a DFA represents a state of the parser; an arc represents
404
a transition. Transitions are either labeled with terminal symbols or
405
with non-terminals. When the parser decides to follow an arc labeled
406
with a non-terminal, it is invoked recursively with the DFA representing
407
the parsing rule for that as its initial state; when that DFA accepts,
408
the parser that invoked it continues. The parse tree constructed by the
409
recursively called parser is inserted as a child in the current parse tree.
411
The DFA's can be constructed automatically from a more conventional
412
language description. An extended LL(1) grammar (ELL(1)) is suitable.
413
Certain restrictions make the parser's life easier: rules that can produce
414
the empty string should be outlawed (there are other ways to put loops
415
or optional parts in the language). To avoid the need to construct
416
FIRST sets, we can require that all but the last alternative of a rule
417
(really: arc going out of a DFA's state) must begin with a terminal
420
As an example, consider this grammar:
422
expr: term (OP term)*
423
term: CONSTANT | '(' expr ')'
425
The DFA corresponding to the rule for expr is:
427
------->.---term-->.------->
432
The parse tree generated for the input a+b is:
434
(expr: (term: (NAME: a)), (OP: +), (term: (NAME: b)))