(not a structure) and means that the reference is being moved out, so
it will not automatically be freed.
+Symbols that are left-recursive are a little special. These are symbols
+that both the head of a production and the first body symbol of the same
+production. They are problematic when they appear in other productions
+elsewhere than at the end, and when indenting is used to describe
+structure. In this case there is no way for a smaller indent to ensure
+the left-recursive symbol cannot be extended. When it appears at the
+end of a production, that production can be reduced to ensure the symbol
+isn't extended. So we record left-recursive symbols while reading the
+grammar, and produce a warning when reporting the grammar if they are
+found in an unsuitable place.
+
+A symbol that is only left recursive in a production where it is
+followed by newline does not cause these problems because the newline
+will effectively terminate it.
+
While building productions we will need to add to an array which needs to
grow dynamically.
###### symbol fields
int first_production;
+ int left_recursive;
###### functions
static char *parse_production(struct grammar *g,
}
tk = token_next(state);
}
+ if (p.body_size >= 2 &&
+ p.body[0] == p.head &&
+ p.body[1]->num != TK_newline)
+ p.head->left_recursive = 1;
+
if (tk.num != TK_newline && tk.num != TK_eof) {
err = "stray tokens at end of line";
goto abort;
snum++;
}
g->first_nonterm = snum;
+ for (s = g->syms; s; s = s->next)
+ if (s->num < 0 && s->type != Virtual) {
+ s->num = snum;
+ snum++;
+ }
for (s = g->syms; s; s = s->next)
if (s->num < 0) {
s->num = snum;
if (g->follow)
report_follow(g);
report_itemsets(g);
- return report_conflicts(g, type);
+ return report_conflicts(g, type) + report_left_recursive(g);
}
Firstly we have the complete list of symbols, together with the
return cnt;
}
+
+### Reporting non-final left-recursive symbols.
+
+Left recursive symbols are a problem for parses that honour indentation
+when they appear other than at the end of the production. So we need to
+report these when asked.
+
+###### functions
+
+ static int report_left_recursive(struct grammar *g)
+ {
+ int p;
+ int bad_left_recursive = 0;
+
+ for (p = 0; p < g->production_count; p++) {
+ struct production *pr = g->productions[p];
+ int sn;
+
+ for (sn = 0; sn < pr->body_size - 1; sn++) {
+ struct symbol *s = pr->body[sn];
+
+ if (s->left_recursive == 1 &&
+ s != pr->head) {
+ if (!bad_left_recursive) {
+ bad_left_recursive = 1;
+ printf("Misplaced left recursive symbols.\n");
+ }
+ printf(" ");
+ prtxt(s->name);
+ printf(" in production [%d]\n", p);
+ s->left_recursive = 2;
+ }
+ }
+ }
+ return bad_left_recursive;
+ }
+
## Generating the parser
The exported part of the parser is the `parse_XX` function, where the name
for (i = TK_reserved;
i < g->num_syms;
i++)
- if (g->symtab[i]->type != Terminal)
+ if (g->symtab[i]->type == Nonterminal)
fprintf(f, "\t\"%.*s\",\n", g->symtab[i]->name.len,
g->symtab[i]->name.txt);
fprintf(f, "};\n\n");
We return whatever `asn` was returned by reducing production zero.
If we can neither shift nor reduce we have an error to handle. We pop
-single entries off the stack until we can shift the `TK_error` symbol, then
-drop input tokens until we find one we can shift into the new error state.
+single entries off the stack until we can shift the `TK_error` symbol,
+then drop input tokens until we find one we can shift into the new error
+state. We need to ensure that something is dropped or shifted after an
+error, or we could get into an infinite loop, only shifting in
+`TK_error`, then reducing. So we track if there has been a shift since
+the last error, and if not the next error always discards one token.
When we find `TK_in` and `TK_out` tokens which report indents we need
to handle them directly as the grammar cannot express what we want to
Here the NEWLINE will be shifted because nothing can be reduced until
the `if` is seen.
-When, during error handling, we discard token read in, we want to keep
+When during error handling we discard tokens read in, we want to keep
discarding until we see one that is recognised. If we had a full set
-of LR(1) grammar states, this will mean looking in the look-ahead set,
+of LR(1) grammar states, this would mean looking in the look-ahead set,
but we don't keep a full look-ahead set. We only record the subset
-that leads to SHIFT. We can, however, deduce the look-ahead set but
+that leads to SHIFT. We can, however, deduce the look-ahead set by
looking at the SHIFT subsets for all states that we can get to by
reducing zero or more times. So we need a little function which
checks if a given token is in any of these look-ahead sets.
struct parser p = { 0 };
struct token *tk = NULL;
int accepted = 0;
+ int shift_since_err = 1;
void *ret = NULL;
shift(&p, TK_eof, 0, 1, NULL, states);
goto force_reduce;
}
if (shift(&p, tk->num, 0, tk->num == TK_newline, tk, states)) {
+ shift_since_err = 1;
tk = NULL;
parser_trace_action(trace, "Shift");
continue;
// no state accepted TK_error
break;
}
+ if (!shift_since_err) {
+ /* must discard at least one token to avoid
+ * infinite loop.
+ */
+ if (tk->num == TK_eof)
+ break;
+ free(tk);
+ tk = tok_copy(token_next(tokens));
+ }
+ shift_since_err = 0;
tos = &p.stack[p.tos-1];
while (!in_lookahead(tk, states, tos->state) &&
tk->num != TK_eof) {
free(tk);
tk = tok_copy(token_next(tokens));
+ shift_since_err = 1;
if (tk->num == TK_in)
indents += 1;
if (tk->num == TK_out) {
code / ocean / blobdiff