struct production {
unsigned short precedence;
enum assoc assoc;
+ char line_like;
## production fields
};
struct grammar {
listed and may be inherited by any production which uses the symbol. A
production inherits from the last symbol which has a precedence.
+The symbols on the first precedence line have the lowest precedence.
+Subsequent lines introduce symbols with higher precedence.
+
###### grammar fields
struct text current_type;
int type_isref;
goto abort;
}
vs = sym_find(g, tk.txt);
- if (vs->type != Virtual) {
- err = "symbol after $$ must be virtual";
+ if (vs->num == TK_newline)
+ p.line_like = 1;
+ else if (vs->num == TK_out)
+ p.line_like = 2;
+ else if (vs->precedence == 0) {
+ err = "symbol after $$ must have precedence";
goto abort;
+ } else {
+ p.precedence = vs->precedence;
+ p.assoc = vs->assoc;
}
- p.precedence = vs->precedence;
- p.assoc = vs->assoc;
tk = token_next(state);
}
if (tk.num == TK_open) {
the complete item, if there is one. We ignore the possibility of
there being two and don't (currently) handle precedence in such
grammars. When completing a grammar we ignore any item where DOT is
-followed by a terminal with a precedence lower (numerically higher)
-than that for the itemset. Unless the terminal has right
-associativity, we also ignore items where the terminal has the same
-precedence. The result is that unwanted items are still in the
-itemset, but the terminal doesn't get into the go to set, so the item
-is ineffective.
+followed by a terminal with a precedence lower than that for the
+itemset. Unless the terminal has right associativity, we also ignore
+items where the terminal has the same precedence. The result is that
+unwanted items are still in the itemset, but the terminal doesn't get
+into the go to set, so the item is ineffective.
###### complete itemset
for (i = 0; i < is->items.cnt; i++) {
struct symbol *s;
struct symset LA = INIT_SYMSET;
unsigned short sn = 0;
+ struct symset LAnl = INIT_SYMSET;
+ unsigned short snnl = 0;
if (is->min_prefix == 0 ||
(bs > 0 && bs < is->min_prefix))
continue;
s = pr->body[bs];
if (s->precedence && is->precedence &&
- is->precedence < s->precedence)
+ is->precedence > s->precedence)
/* This terminal has a low precedence and
* shouldn't be shifted
*/
}
sn = save_set(g, LA);
LA = set_find(g, sn);
+ if (symset_find(&LA, TK_newline))
+ symset_add(&LAnl, TK_newline, 0);
+ snnl = save_set(g, LAnl);
+ LAnl = set_find(g, snnl);
}
/* Add productions for this symbol */
int itm = item_num(p2, 0);
int pos = symset_find(&is->items, itm);
if (pos < 0) {
- symset_add(&is->items, itm, sn);
+ if (g->productions[p2]->line_like)
+ symset_add(&is->items, itm, snnl);
+ else
+ symset_add(&is->items, itm, sn);
/* Will have re-ordered, so start
* from beginning again */
i = -1;
} else if (type >= LALR) {
struct symset ss = set_find(g, is->items.data[pos]);
struct symset tmp = INIT_SYMSET;
+ struct symset *la = &LA;
+ if (g->productions[p2]->line_like)
+ la = &LAnl;
symset_union(&tmp, &ss);
- if (symset_union(&tmp, &LA)) {
+ if (symset_union(&tmp, la)) {
is->items.data[pos] = save_set(g, tmp);
i = -1;
- }else
+ } else
symset_free(tmp);
}
}
pos = symset_find(&newitemset, pr->head->num);
if (bp + 1 == pr->body_size &&
pr->precedence > 0 &&
- (precedence == 0 ||
- pr->precedence < precedence)) {
+ pr->precedence > precedence) {
// new itemset is reducible and has a precedence.
precedence = pr->precedence;
assoc = pr->assoc;
printf(" [%d%s]", s->precedence,
assoc_names[s->assoc]);
}
+ if (pr->line_like == 1)
+ printf(" $$NEWLINE");
+ else if (pr->line_like)
+ printf(" $$OUT");
printf("\n");
}
which maps terminals to items that could be reduced when the terminal
is in look-ahead. We report when we get conflicts between the two.
-As a special case, if we find a SHIFT/REDUCE conflict, where a
-terminal that could be shifted is in the lookahead set of some
-reducable item, then set check if the reducable item also have
-`TK_newline` in its lookahead set. If it does, then a newline will
-force and reduction, but anything else can reasonably be shifts, so
-that isn't really a conflict. Such apparent conflicts do not get
-reported. This will not affect a "tradtional" grammar that does not
-include newlines as token.
+As a special case, if we find a SHIFT/REDUCE conflict, on the NEWLINE
+terminal, we ignore it. NEWLINES are handled specially with its own
+rules for when to shift and when to reduce. Conflicts are expected,
+but handled internally.
static int conflicts_slr(struct grammar *g, enum grammar_type type)
{
int p = item_prod(itm);
int bp = item_index(itm);
struct production *pr = g->productions[p];
+ struct symbol *s;
- if (bp < pr->body_size &&
- pr->body[bp]->type == Terminal) {
- /* shiftable */
- int sym = pr->body[bp]->num;
- if (symset_find(&shifts, sym) < 0)
- symset_add(&shifts, sym, itm);
- }
+ if (bp >= pr->body_size ||
+ pr->body[bp]->type != Terminal)
+ /* not shiftable */
+ continue;
+
+ s = pr->body[bp];
+ if (s->precedence && is->precedence)
+ /* Precedence resolves this, so no conflict */
+ continue;
+
+ if (symset_find(&shifts, s->num) < 0)
+ symset_add(&shifts, s->num, itm);
}
/* Now look for reductions and conflicts */
for (j = 0; j < is->items.cnt; j++) {
int k;
for (k = 0; k < la.cnt; k++) {
int pos = symset_find(&shifts, la.syms[k]);
- if (pos >= 0 && symset_find(&la, TK_newline) < 0) {
+ if (pos >= 0 && la.syms[k] != TK_newline) {
printf(" State %d has SHIFT/REDUCE conflict on ", i);
- prtxt(g->symtab[la.syms[k]]->name);
+ cnt++;
+ prtxt(g->symtab[la.syms[k]]->name);
printf(":\n");
report_item(g, shifts.data[pos]);
report_item(g, itm);
- cnt++;
}
pos = symset_find(&reduce, la.syms[k]);
if (pos < 0) {
###### parser_generate
- static void gen_parser(FILE *f, struct grammar *g, char *file, char *name)
+ static void gen_parser(FILE *f, struct grammar *g, char *file, char *name,
+ struct code_node *pre_reduce)
{
gen_known(f, g);
gen_non_term(f, g);
gen_goto(f, g);
gen_states(f, g);
- gen_reduce(f, g, file);
+ gen_reduce(f, g, file, pre_reduce);
gen_free(f, g);
fprintf(f, "#line 0 \"gen_parser\"\n");
### Known words table
The known words table is simply an array of terminal symbols.
-The table of nonterminals used for tracing is a similar array.
+The table of nonterminals used for tracing is a similar array. We
+include virtual symbols in the table of non_terminals to keep the
+numbers right.
###### functions
for (i = TK_reserved;
i < g->num_syms;
i++)
- if (g->symtab[i]->type == Nonterminal)
+ if (g->symtab[i]->type != Terminal)
fprintf(f, "\t\"%.*s\",\n", g->symtab[i]->name.len,
g->symtab[i]->name.txt);
fprintf(f, "};\n\n");
short reduce_prod;
short reduce_size;
short reduce_sym;
- short starts_line;
+ char starts_line;
+ char newline_only;
short min_prefix;
};
}
if (prod >= 0)
- fprintf(f, "\t[%d] = { %d, goto_%d, %d, %d, %d, %d, %d },\n",
+ fprintf(f, "\t[%d] = { %d, goto_%d, %d, %d, %d, %d, %d, %d },\n",
i, is->go_to.cnt, i, prod,
g->productions[prod]->body_size,
g->productions[prod]->head->num,
- is->starts_line, is->min_prefix);
+ is->starts_line,
+ g->productions[prod]->line_like,
+ is->min_prefix);
else
- fprintf(f, "\t[%d] = { %d, goto_%d, -1, -1, -1, %d, %d },\n",
+ fprintf(f, "\t[%d] = { %d, goto_%d, -1, -1, -1, %d, 0, %d },\n",
i, is->go_to.cnt, i,
is->starts_line, is->min_prefix);
}
fputs("\n", f);
for (i = 0; i < p->body_size; i++) {
if (p->body[i]->struct_name.txt &&
- p->body[i]->isref &&
- used[i])
+ used[i]) {
// assume this has been copied out
- fprintf(f, "\t\t*(void**)body[%d] = NULL;\n", i);
+ if (p->body[i]->isref)
+ fprintf(f, "\t\t*(void**)body[%d] = NULL;\n", i);
+ else
+ fprintf(f, "\t\tmemset(body[%d], 0, sizeof(struct %.*s));\n", i, p->body[i]->struct_name.len, p->body[i]->struct_name.txt);
+ }
}
free(used);
}
###### functions
- static void gen_reduce(FILE *f, struct grammar *g, char *file)
+ static void gen_reduce(FILE *f, struct grammar *g, char *file,
+ struct code_node *code)
{
int i;
- fprintf(f, "#line 0 \"gen_reduce\"\n");
+ fprintf(f, "#line 1 \"gen_reduce\"\n");
fprintf(f, "static int do_reduce(int prod, void **body, struct token_config *config, void *ret)\n");
fprintf(f, "{\n");
fprintf(f, "\tint ret_size = 0;\n");
+ if (code)
+ code_node_print(f, code, file);
+ fprintf(f, "#line 4 \"gen_reduce\"\n");
fprintf(f, "\tswitch(prod) {\n");
for (i = 0; i < g->production_count; i++) {
struct production *p = g->productions[i];
struct code_node *hdr = NULL;
struct code_node *code = NULL;
struct code_node *gram = NULL;
+ struct code_node *pre_reduce = NULL;
for (s = table; s; s = s->next) {
struct text sec = s->section;
if (tag && !strip_tag(&sec, tag))
code = s->code;
else if (text_is(sec, "grammar"))
gram = s->code;
+ else if (text_is(sec, "reduce"))
+ pre_reduce = s->code;
else {
fprintf(stderr, "Unknown content section: %.*s\n",
s->section.len, s->section.txt);
if (f) {
if (code)
code_node_print(f, code, infile);
- gen_parser(f, g, infile, name);
+ gen_parser(f, g, infile, name, pre_reduce);
fclose(f);
} else {
fprintf(stderr, "Cannot create %s.c\n",
continue;
}
force_reduce:
- if (states[tos->state].reduce_prod >= 0) {
+ if (states[tos->state].reduce_prod >= 0 &&
+ states[tos->state].newline_only &&
+ tk->num != TK_newline && tk->num != TK_eof && tk->num != TK_out) {
+ /* Anything other than newline in an error as this
+ * production must end at EOL
+ */
+ } else if (states[tos->state].reduce_prod >= 0) {
void **body;
void *res;
const struct state *nextstate = &states[tos->state];
# calc: grammar
- $LEFT * /
$LEFT + -
+ $LEFT * /
Session -> Session Line
| Line
code / ocean / blobdiff